hydrometeorology of tropical montane cloud forests ...hydrometeorology of tropical montane cloud...

33
HYDROLOGICAL PROCESSES Hydrol. Process. (2010) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.7974 Hydrometeorology of tropical montane cloud forests: emerging patterns L. A. Bruijnzeel, 1 * Mark Mulligan 2 and Frederick N. Scatena 3 1 Faculty of Earth and Life Sciences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands 2 Environmental Monitoring and Modelling Research Group, Department of Geography, King’s College London, Strand, London WC2R 2LS, UK 3 Department of Earth & Environmental Science, Hayden Hall, University of Pennsylvania, 240 South 33rd Street, Philadelphia, PA 19104, USA Abstract: Tropical montane cloud forests (TMCF) typically experience conditions of frequent to persistent fog. On the basis of the altitudinal limits between which TMCF generally occur (800–3500 m.a.s.l. depending on mountain size and distance to coast) their current areal extent is estimated at ¾215 000 km 2 or 6Ð6% of all montane tropical forests. Alternatively, on the basis of remotely sensed frequencies of cloud occurrence, fog-affected forest may occupy as much as 2Ð21 Mkm 2 . Four hydrologically distinct montane forest types may be distinguished, viz. lower montane rain forest below the cloud belt (LMRF), tall lower montane cloud forest (LMCF), upper montane cloud forest (UMCF) of intermediate stature and a group that combines stunted sub-alpine cloud forest (SACF) and ‘elfin’ cloud forest (ECF). Average throughfall to precipitation ratios increase from 0Ð72 š 0Ð07 in LMRF (n D 15) to 0Ð81 š 0Ð11 in LMCF (n D 23), to 1Ð0 š 0Ð27 (n D 18) and 1Ð04 š 0Ð25 (n D 8) in UMCF and SACF–ECF, respectively. Average stemflow fractions increase from LMRF to UMCF and ECF, whereas leaf area index (LAI) and annual evapotranspiration (ET) decrease along the same sequence. Although the data sets for UMCF (n D 3) and ECF (n D 2) are very limited, the ET from UMCF (783 š 112 mm) and ECF (547 š 25 mm) is distinctly lower than that from LMCF (1188 š 239 mm, n D 9) and LMRF (1280 š 72 mm; n D 7). Field-measured annual ‘cloud-water’ interception (CWI) totals determined with the wet-canopy water budget method (WCWB) vary widely between locations and range between 22 and 1990 mm (n D 15). Field measured values also tend to be much larger than modelled amounts of fog interception, particularly at exposed sites. This is thought to reflect a combination of potential model limitations, a mismatch between the scale at which the model was applied (1 ð 1 km) and the scale of the measurements (small plots), as well as the inclusion of near-horizontal wind-driven precipitation in the WCWB-based estimate of CWI. Regional maps of modelled amounts of fog interception across the tropics are presented, showing major spatial variability. Modelled contributions by CWI make up less than 5% of total precipitation in wet areas to more than 75% in low-rainfall areas. Catchment water yields typically increase from LMRF to UMCF and SACF–ECF reflecting concurrent increases in incident precipitation and decreases in evaporative losses. The conversion of LMCF (or LMRF) to pasture likely results in substantial increases in water yield. Changes in water yield after UMCF conversion are probably modest due to trade-offs between concurrent changes in ET and CWI. General circulation model (GCM)-projected rates of climatic drying under SRES greenhouse gas scenarios to the year 2050 are considered to have a profound effect on TMCF hydrological functioning and ecology, although different GCMs produce different and sometimes opposing results. Whilst there have been substantial increases in our understanding of the hydrological processes operating in TMCF, additional research is needed to improve the quantification of occult precipitation inputs (CWI and wind-driven precipitation), and to better understand the hydrological impacts of climate- and land-use change. Copyright 2010 John Wiley & Sons, Ltd. KEY WORDS cloud forest; cloud-water interception; fog; evaporation; rainfall interception; stemflow; throughfall; transpiration; wind-driven rain Received 2 November 2010; Accepted 3 December 2010 INTRODUCTION Tropical montane cloud forests (TMCF) are typically found in foggy, wet and often windy environments whose ecological and hydrological functioning have puzzled and challenged investigators for decades. Apart from being * Correspondence to: L. A. Bruijnzeel, Faculty of Earth and Life Sci- ences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. E-mail: [email protected] This paper is partly derived from a chapter previously published as Bruijnzeel LA, Kappelle M, Mulligan M, Scatena FN. 2010. Tropical montane cloud forests: state of knowledge and sustainability perspec- tives in a changing world. In Tropical Montane Cloud Forests. Science for Conservation and Management, Bruijnzeel LA, Scatena FN, Hamil- ton LS (eds). Cambridge University Press: Cambridge, UK; 691–740 (www.cambridge.org/9780521760355). amongst the world’s most valuable terrestrial ecosystems in terms of species richness and levels of endemism [see Bruijnzeel et al. (2010a,b) for a recent overview], headwater areas with TMCF also provide a stable supply of high-quality water that is indispensable for maintaining irrigation, hydro-electric power generation and drinking water (Zadroga, 1981; Brown et al., 1996; Tognetti et al., 2010). Although cloud forests are often referred to as a single category, it is helpful to distinguish between (1) tall-statured lower montane cloud forest (LMCF), (2) upper montane cloud forest (UMCF) of intermediate stature and (3) stunted sub-alpine (SACF) and ‘elfin’ cloud forests (ECF). The rationale for making such a distinction lies in the wetter and cooler conditions Copyright 2010 John Wiley & Sons, Ltd.

Upload: others

Post on 01-Aug-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROLOGICAL PROCESSESHydrol. Process. (2010)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/hyp.7974

Hydrometeorology of tropical montane cloud forests:emerging patterns†

L. A. Bruijnzeel,1* Mark Mulligan2 and Frederick N. Scatena3

1 Faculty of Earth and Life Sciences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands2 Environmental Monitoring and Modelling Research Group, Department of Geography, King’s College London, Strand, London WC2R

2LS, UK3 Department of Earth & Environmental Science, Hayden Hall, University of Pennsylvania, 240 South 33rd Street, Philadelphia, PA 19104, USA

Abstract:

Tropical montane cloud forests (TMCF) typically experience conditions of frequent to persistent fog. On the basis of thealtitudinal limits between which TMCF generally occur (800–3500 m.a.s.l. depending on mountain size and distance to coast)their current areal extent is estimated at ¾215 000 km2 or 6Ð6% of all montane tropical forests. Alternatively, on the basis ofremotely sensed frequencies of cloud occurrence, fog-affected forest may occupy as much as 2Ð21 Mkm2. Four hydrologicallydistinct montane forest types may be distinguished, viz. lower montane rain forest below the cloud belt (LMRF), tall lowermontane cloud forest (LMCF), upper montane cloud forest (UMCF) of intermediate stature and a group that combines stuntedsub-alpine cloud forest (SACF) and ‘elfin’ cloud forest (ECF). Average throughfall to precipitation ratios increase from0Ð72 š 0Ð07 in LMRF (n D 15) to 0Ð81 š 0Ð11 in LMCF (n D 23), to 1Ð0 š 0Ð27 (n D 18) and 1Ð04 š 0Ð25 (n D 8) in UMCFand SACF–ECF, respectively. Average stemflow fractions increase from LMRF to UMCF and ECF, whereas leaf area index(LAI) and annual evapotranspiration (ET) decrease along the same sequence. Although the data sets for UMCF (n D 3) andECF (n D 2) are very limited, the ET from UMCF (783 š 112 mm) and ECF (547 š 25 mm) is distinctly lower than thatfrom LMCF (1188 š 239 mm, n D 9) and LMRF (1280 š 72 mm; n D 7). Field-measured annual ‘cloud-water’ interception(CWI) totals determined with the wet-canopy water budget method (WCWB) vary widely between locations and range between22 and 1990 mm (n D 15). Field measured values also tend to be much larger than modelled amounts of fog interception,particularly at exposed sites. This is thought to reflect a combination of potential model limitations, a mismatch between thescale at which the model was applied (1 ð 1 km) and the scale of the measurements (small plots), as well as the inclusion ofnear-horizontal wind-driven precipitation in the WCWB-based estimate of CWI. Regional maps of modelled amounts of foginterception across the tropics are presented, showing major spatial variability. Modelled contributions by CWI make up lessthan 5% of total precipitation in wet areas to more than 75% in low-rainfall areas. Catchment water yields typically increasefrom LMRF to UMCF and SACF–ECF reflecting concurrent increases in incident precipitation and decreases in evaporativelosses. The conversion of LMCF (or LMRF) to pasture likely results in substantial increases in water yield. Changes inwater yield after UMCF conversion are probably modest due to trade-offs between concurrent changes in ET and CWI.General circulation model (GCM)-projected rates of climatic drying under SRES greenhouse gas scenarios to the year 2050are considered to have a profound effect on TMCF hydrological functioning and ecology, although different GCMs producedifferent and sometimes opposing results. Whilst there have been substantial increases in our understanding of the hydrologicalprocesses operating in TMCF, additional research is needed to improve the quantification of occult precipitation inputs (CWIand wind-driven precipitation), and to better understand the hydrological impacts of climate- and land-use change. Copyright 2010 John Wiley & Sons, Ltd.

KEY WORDS cloud forest; cloud-water interception; fog; evaporation; rainfall interception; stemflow; throughfall; transpiration;wind-driven rain

Received 2 November 2010; Accepted 3 December 2010

INTRODUCTION

Tropical montane cloud forests (TMCF) are typicallyfound in foggy, wet and often windy environments whoseecological and hydrological functioning have puzzled andchallenged investigators for decades. Apart from being

* Correspondence to: L. A. Bruijnzeel, Faculty of Earth and Life Sci-ences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, TheNetherlands. E-mail: [email protected]† This paper is partly derived from a chapter previously published asBruijnzeel LA, Kappelle M, Mulligan M, Scatena FN. 2010. Tropicalmontane cloud forests: state of knowledge and sustainability perspec-tives in a changing world. In Tropical Montane Cloud Forests. Sciencefor Conservation and Management, Bruijnzeel LA, Scatena FN, Hamil-ton LS (eds). Cambridge University Press: Cambridge, UK; 691–740(www.cambridge.org/9780521760355).

amongst the world’s most valuable terrestrial ecosystemsin terms of species richness and levels of endemism[see Bruijnzeel et al. (2010a,b) for a recent overview],headwater areas with TMCF also provide a stable supplyof high-quality water that is indispensable for maintainingirrigation, hydro-electric power generation and drinkingwater (Zadroga, 1981; Brown et al., 1996; Tognetti et al.,2010). Although cloud forests are often referred to asa single category, it is helpful to distinguish between(1) tall-statured lower montane cloud forest (LMCF),(2) upper montane cloud forest (UMCF) of intermediatestature and (3) stunted sub-alpine (SACF) and ‘elfin’cloud forests (ECF). The rationale for making sucha distinction lies in the wetter and cooler conditions

Copyright 2010 John Wiley & Sons, Ltd.

Page 2: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

generally encountered as one moves from the lowermontane to the upper montane and sub-alpine belts, andwhich are known to affect the hydrological and ecologicalfunctioning of the respective forest types (Grubb, 1977;Silver et al., 1999; Bruijnzeel, 2001; Gerold et al., 2008;Benner et al., 2010; Roman et al., 2010).

The wet and generally remote and difficult terrainof the world’s TMCF has not only made them hydro-logically and ecologically unique but also given themsome de facto protection in the past compared to tropi-cal forests situated in more accessible areas. However, inthe late 1970s and early 1980s, it became apparent thatin many parts of the world TMCF were rapidly beingconverted and in need of more formal forms of pro-tection (LaBastille and Pool, 1978; Stadtmuller, 1987).Indeed, between 1981 and 1990, montane forests acrossthe tropics were being lost at a faster rate than low-land tropical forests (1Ð1% vs 0Ð8% year�1, respectively;Doumenge et al., 1995). Two recent inventories estimatedthat around the year 2000 about 45–55% of all cloud-affected forests located between 23Ð5 °N and 35 °S hadbeen converted to other forms of land use (Mulligan,2010; Scatena et al., 2010). Conversions to agriculturaland grazing lands, excessive timber harvesting, invasionsby exotic species, road ingressions and various types ofdevelopment have been identified as threats to TMCFin all regions, whereas mining, fire, forest clearing fordrug cultivation and other activities like golf courses orcommunication facilities can be locally important (Hamil-ton et al., 1995; Bruijnzeel and Hamilton, 2000; Kap-pelle and Brown, 2001; Bubb et al., 2004; Hemp, 2005a;Asbjornsen and Garnica-Sanchez, 2010; Mulligan, 2010).In recent years, climatic warming and drying relatedto global or regional climate change have become anincreasingly important factor that can potentially threatenTMCF hydrological functioning (Lawton et al., 2001;Hemp, 2005a; Ray et al., 2006), in addition to havinga devastating effect on particularly vulnerable plant andanimal groups like mosses and amphibians (Pounds et al.,1999, 2006; Nadkarni and Solano, 2002; Williams et al.,2003).

Whilst it is broadly recognized that all of these threatscan impact the hydrological functioning of headwaterareas with TMCF, the scientific information required toquantify these impacts and to help manage these uniquebut vulnerable ecosystems was largely lacking until com-paratively recently. In 1993, the First International Sym-posium on TMCF was held in San Juan, Puerto Rico, theproceedings of which (Hamilton et al., 1995) containedthe first overview of what was known hydrologicallyof TMCF at the time (Bruijnzeel and Proctor, 1995)as well as one of the first physically based studies ofcloud-water interception (CWI) in a TMCF setting (Juvikand Nullet, 1995a). Certain aspects of CWI and TMCFhydrology have been considered at a series of Confer-ences on Fog and Fog Collection (held every three yearssince 1998; Schemenauer and Bridgman, 1998; Scheme-nauer and Puxbaum, 2001; Rautenbach and Oliver, 2004;Biggs and Cereceda, 2007; Climatology Working Group,

2010). Arguably, however, the San Juan Symposiummarked the start of increased research activity in thefields of TMCF hydrology, hydrometeorology and eco-physiology. Thus, whilst Bruijnzeel and Proctor (1995)were able to list only eight studies of crown drip andoccult precipitation in TMCF environments, plus a meresix studies estimating overall evaporation loss throughindirect methods and none quantifying transpiration ratesin TMCF or the impact of TMCF conversion on stream-flow amounts and seasonal distribution, at the follow-upSymposium on Science for the Conservation and Man-agement of TMCF held in 2004 in Waimea, Hawai‘i(Bruijnzeel et al., 2010a), some 25 presentations reportedon hydrometeorological and plant physiological work thathad been conducted since 1993. Quantitative evidence onthe effects of TMCF conversion to pasture, as well as onthe impacts of climatic variability and change were givenin another ten presentations.

The presence of cloud forest is widely assumed toincrease streamflow volumes, not only because of theextra amounts of water captured from passing fog,beyond that provided by precipitation, but also becauseof reduced evaporative losses under the prevailing lowradiation levels and high atmospheric humidity (cf.Zadroga, 1981; Calvo, 1986; Jarvis and Mulligan, 2010).In addition, the forest helps to reduce the number ofshallow landslides and prevents surface erosion, therebymaintaining better water quality (Sidle et al., 2006;cf. Bruijnzeel, 2004). Such considerations lie at theheart of many ‘payment for ecosystem services’ (PES)schemes in which downstream users pay a certain fee for(mostly hydrological) services rendered by cloud forestto compensate upstream forest owners who conserve theircloud forests instead of converting them to economicallymore profitable forms of land use such as grazingor cropping (Pagiola, 2002; Rodriguez-Zuniga, 2003).Given the great pressure on the world’s remaining cloudforests, and the growing recognition of their value astreasure houses of biodiversity and as providers of high-quality water, an array of PES-initiatives aimed at TMCFconservation has emerged in recent years (Asquith andWunder, 2008; Munoz-Pina et al., 2008; Porras et al.,2008; Garriguata and Balvanera, 2009; Tognetti et al.,2010). Needless to say, such PES schemes and land-and forest managers and policy-makers in general needto determine which cloud forests under their jurisdictionprovide the best water supplies (and to whom), which arethe most vulnerable to climate change or most threatenedby encroachment, and what are the hydrological impactsassociated with forest conversion or climate change. Inshort, there is a great need for site-specific informationon TMCF hydrological functioning for incorporation intoconservation and management plans at various spatialscales (Bruijnzeel et al., 2010b).

After first defining the various types of cloudforests and exploring their global distribution, thisarticle summarizes the currently available knowl-edge on the hydrometeorology of TMCF and provides

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 3: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Table I. Summary of key structural characteristics marking the chief tropical (montane) forest types distinguished in the presentpaper (based on Frahm and Gradstein, 1991; Whitmore, 1998)

Forest formationa LERF LMRF/LMCF UMRF SACF

Canopy height 25–45 m 15–33 m 1Ð5–18 m 1Ð5–9 mEmergent trees Up to 67 m tall Often absent, up to 37 m Usually absent, up

to 26 mUsually absent, up

to 15 mCompound leaves Abundant Occasional Rare AbsentPrincipal leaf size classb Mesophyllous Meso-/notophyllous Microphyllous Nanophyllous

Leaf drip-tips Abundant Present Rare or absent AbsentButtresses Frequent and large Uncommon, and small Usually absent AbsentCauliflory Frequent Rare Absent AbsentBig woody climbers Abundant Usually absent Absent AbsentBole climbers Often abundant Frequent to abundant Very few AbsentVascular epiphytes Frequent Abundant Frequent Very rareNon-vascular epiphytes

(mosses, liverworts)Occasional Occasional/Abundant

<10%/25–50%Usually abundant

70–80%Abundant >80%

a LERF, lowland evergreen rain forest; LMRF/LMCF, lower montane rain/cloud forest; UMCF, upper montane cloud forest; SACF, sub-alpine cloudforest.b Leaf sizes according to the Raunkiaer classification system: mesophyllous, 4500–18 225 mm2; notophyllous, 2025–4500 mm2; microphyllous,225–2025 mm2; nanophyllous, <225 mm2.

state-of-knowledge answers to the following questionsthat are considered to be of key importance:

ž What are the best ways to determine precipitation andcloud-water inputs, both at the point scale and at largerscales in these challenging environments with theircomplex topography and (often) strong winds?

ž What are the amounts of CWI, net precipitation (i.e.throughfall C stemflow) and evaporation losses typi-cally associated with different types of TMCF?

ž Are there predictable regional patterns in CWI?ž How are annual and seasonal water yields (and thus

ecosystem services) affected by converting differenttypes of TMCF to pasture, annual crops or coffeeplantations?

ž How are different types of TMCF affected by climaticwarming and a reduction in precipitation?

This article expands upon the information presented onhydrological processes that was included in a recent state-of-knowledge review of the biodiversity, biophysicalfunctioning and conservation and management issues ofTMCF (Bruijnzeel et al., 2010b). In addition, it presentsthe first regional maps of modelled fog interception acrossthe tropics, both in terms of absolute amounts and as apercentage of total water inputs (i.e. rainfall plus fog).

DEFINITIONS OF TROPICAL MONTANE CLOUDFOREST

Despite the widely acknowledged biological and hydro-logical values of TMCF, their definition—and there-fore the delineation of their spatial distribution—hasremained a persistent challenge (Stadtmuller, 1987; Cam-panella, 1995; Hamilton et al., 1995; Ashton, 2003; Bach,2004; Bubb et al., 2004; Hemp, 2005b; Martin et al.,2007; Meyer, 2010; Mulligan, 2010). Historically, the

problem was confounded by a myriad of imprecise, over-lapping and, at times, contradictory definitions of TMCFand as many as 35 different names have been used totypify ‘cloud forests’ in the past (Stadtmuller, 1987; cf.Meyer, 2010). However, it is possible to distinguish threegeneral TMCF types based on forest structural character-istics (Grubb, 1977), degree of mossiness (Frahm andGradstein, 1991) and contrasts in the fraction of net pre-cipitation reaching the forest floor (Bruijnzeel, 2001):viz. (1) LMCF, (2) UMCF and (3) SACF–ECF withinthe widely adopted definition of cloud forests as ‘foreststhat are frequently covered in cloud or mist’ (Stadtmuller,1987; Hamilton et al., 1995). The presently used classifi-cation (Table I) thus recognizes the important influence oftemperature and humidity levels on montane forest zona-tion. A brief description of the respective forest types isgiven below.

With increasing elevation on wet tropical mountains,distinct changes occur in forest appearance and struc-ture. Generally, the tall and often buttressed trees of themulti-storied lowland evergreen rain forest (main canopyheight 25–45 m, with emergents reported up to 67 m)gradually give way to lower montane forests. With anaverage canopy height of up to 33 m, lower montaneforest can still be quite impressive. The large buttressesand climbers that are so abundant in the lowland for-est all but disappear whilst epiphytes (orchids, ferns,bromeliads) on branches and stems become more numer-ous with elevation (Whitmore, 1998). The change fromlowland to lower montane forest is normally observedat the elevation where the average minimum tempera-ture drops below 18 °C and many lowland tree speciesare displaced by a floristically different assemblage ofmontane species (Kitayama, 1995). On large equatorialinland mountains this transition usually occurs at an alti-tude of 1200–1500 m.a.s.l. and coincides roughly withthe start of intermittent fog occurrence (Kitayama, 1995).

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 4: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

As elevation continues to increase, the trees become grad-ually smaller, moss cover on the stems increases from<10% to 25–50%, and an LMCF is observed. Withfurther increases in elevation, there is usually a veryclear change from relatively tall (15–33 m) LMCF toa distinctly shorter-statured (¾2–18 m) and much moremossy (70–80% bryophytic stem cover) UMCF (Frahmand Gradstein, 1991). Although the two forest types arenot separated by a distinct thermal threshold, there islittle doubt that the transition from LMCF to UMCF coin-cides with the level where cloud condensation becomesmost persistent (Grubb and Whitmore, 1966; Wolf, 1993;Kitayama, 1995; cf. Schawe et al., 2010). On large moun-tains in equatorial regions away from the ocean thisbelt of persistent cloud typically occurs at elevationsof ¾2000–3000 m.a.s.l. (Grubb, 1977; Kitayama, 1995;Hemp, 2010). However, on small oceanic islands and oncoastal mountains the change from LMCF to UMCF mayoccur at much lower altitudes (Table II). Mosses also startto cover rocks and fallen trunks on the soil surface in theUMCF zone. With increasing exposure to wind and pre-cipitation, tree stems become increasingly crooked andgnarled, and bamboos often replace palms as dominantundergrowth species (Kappelle, 1995). A third major typeof TMCF occurs around the elevation where the averagemaximum temperature falls below 10 °C. Here, UMCFgives way to still smaller-statured (1Ð5–9 m) and more

species-poor SACF (or scrub) (Kitayama, 1995; Bach,2004; Hemp, 2005b). This forest type is characterizednot only by its low stature and gnarled appearance butalso by even smaller leaves, and a comparative absenceof epiphytes (Table I). Mosses may remain abundant andfog incidence is still a paramount feature though (Frahmand Gradstein, 1991; Schawe et al., 2010). On largeequatorial mountains the transition to SACF is generallyobserved at elevations between 2800 and 3200 m.a.s.l.(Kitayama, 1995; Hemp, 2010; Schawe et al., 2010).Therefore, SACF is found only on the highest moun-tains, mostly in Latin America and Papua New Guinea,where it may extend to ¾3900 m.a.s.l. (Whitmore, 1998;cf. Hemp, 2010). However, as discussed more fullybelow, equally short-statured forest (usually referred toas ECF ), may also be found at much lower elevations(<1000 m.a.s.l.). Whilst structurally similar to SACF,ECF are distinctly different floristically. As such, ECFmay occur at a wide range of elevations (Stadtmuller,1987).

CLIMATE AND THE GLOBAL DISTRIBUTIONOF TMCF

General considerations

Although the different types of TMCF described inTable I are encountered in many regions, there is much

Table II. Regional groupings of tropical montane cloud forest and associated lower and upper altitudinal limits of the cloud belt(after Scatena et al., 2010)

Regional groupings Lower and upper altitudinal limits of cloudforest used in the analysis (m.a.s.l.).

Mexico west of the Isthmus of Tehuantepec 2200–3500Mexico east of the Isthmus of Tehuantepec and Central and South America 1800–3500Caribbean islands and coastal Venezuelan mountains Above 660Atlantic coast of Brazil Above 700Galapagos islands and coastal Ecuadorian mountains Above 400Hawai‘i 1000–3000African Atlantic coast countries 1000–3500Ethiopia and Albertine Rift mountains, including DR Congo 2000–3500Kenya and Northern Tanzania 1500–3500Southern Tanzania, Malawi, Mozambique and Reunion 1000–3500Madagascar 1300–3500Philippines 1200–3500New Guinea 2000–3500Continental South-East Asia, Borneo, Sumatra, Java and Sulawesi 1500–3500South India, Sri Lanka, other insular Asia-Pacific islands, northern Australia 1000–3500

Table III. Estimated area of tropical montane cloud forest (km2) and as a percentage of all tropical (montane) forest per continentalregion

Region All tropicalforest (km2)

Estimated areaof TMCF (km2)

TMCF as % ofall tropical forest

Tropical montaneforest (km2)

TMCF as % of alltropical montane forest

Regional distribution(%) of TMCF

Americas 7 762 359 87 626 1Ð1 1 150 588 7Ð6 40Ð8Africa 4 167 546 34 328 0Ð8 544 664 6Ð3 16Ð0Asia 3 443 330 92 676 2Ð7 1 562 023 5Ð9 43Ð2Global total 15 373 235 214 630 1Ð4 3 257 275 6Ð6 100Ð0

Areas of tropical (montane) forest based on Iremonger et al. (1997) and Kapos et al. (2000). The calculations of Kapos et al. (2000) of the areas ofthe world’s montane forest included altitudinal ranges from 300 m to >4500 m.a.s.l.

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 5: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Figure 1. Modelled distribution of cloud-affected tropical montane forests, with UNEP-WCMC listed cloud forest sites indicated in red. The colourscale indicates the approximate fractional cover of forest within the 1-km pixel

variation in the actual elevations at which the cloudforests occur (Table II) and in their spatial extent indifferent continental regions (Table III and Figure 1).In general, the distributions depend on the upper andlower bounds of the cloud belt (Table II) and on theglobal, regional and local factors that influence cloudformation. As stated previously, the transition fromLMCF to UMCF—as well as the thickness of the cloudforest belt itself—is primarily governed by the levelof persistent cloud condensation (Grubb and Whitmore,1966; Frahm and Gradstein, 1991; Kitayama, 1995). Thelatter, in turn, is determined by the moisture content and

temperature of the atmosphere such that the more humidthe uplifted air, the lower will be the altitude at whichit condenses (Foster, 2010). With increasing distancefrom the ocean, the air tends to be less humid and willrequire lower temperatures, and thus higher elevations, toreach condensation. Consequently, the associated cloudbase, and thus the presence of TMCF, will occur at ahigher elevation as one is moving away from the ocean.Similarly, for a given atmospheric moisture content, thecondensation point is reached more rapidly for coolair than for warm air (Foster, 2010). Hence, at greaterdistance from the equator, the average temperature—and

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 6: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

therefore the altitude at which condensation and TMCFoccur—will be lower (Nullet and Juvik, 1994; Jarvis andMulligan, 2010).

In addition to the elevation of the cloud base, thedistribution and extension of the TMCF belt is alsogoverned by the upper limit of cloud formation, whichis also influenced by global-scale atmospheric circulationfeatures, such as the Hadley cell. In the latter, heated airrises to great elevations in the equatorial zone, and flowspolewards and eastwards in the upper atmosphere as itcools. The cool dry air then descends in a broad belt in theouter tropics and sub-tropics from where it returns to theequator. This subsidence reaches its maximum expressionat oceanic sub-tropical high-pressure centres and alongthe eastern margins of the oceanic basins. As the airdescends and warms, it forms a temperature inversionthat separates the moist layer of surface air (being cooledwhilst rising) from the drier descending air above. Thisso-called ‘trade wind inversion’ (TWI) forms a surfacethat generally rises towards the equator and from east towest across the oceans (Riehl, 1979). Over the easternPacific Ocean, the TWI occurs at only a few hundredmetres above sea level, for example, off the coast ofsouthern California. It rises to about 2200 m near Hawai‘i(Cao et al., 2007) and dissipates in the equatorial westernPacific (Nullet and Juvik, 1994). The consequences ofthe TWI for the occurrence of the upper boundary ofTMCF are profound and are another reason why thevegetation zonation on mountains situated away fromthe equator tends to be compressed. For instance, somewindward slopes in the Hawaiian archipelago receivemore than 6000 mm of rain year�1 below the inversionlayer. However, above the inversion, montane cloudforest suddenly gives way to dry sub-alpine scrub becausethe inversion prevents clouds moving upward and bringmoisture to those areas (Kitayama and Muller-Dombois,1994a,b; cf. Loope and Giambelluca, 1998).

Superimposed on these global-scale moisture and tem-perature gradients are more local processes influencingthe temperature of the air column and thus the ‘start-ing point’ for air subject to cooling by lifting. Theseinclude the influence of offshore sea surface tempera-tures, land–sea interactions involving the coastal plain,the size of a mountain and its orientation and expo-sure to the prevailing winds (Malkus, 1955; Van Steenis,1972; Stadtmuller, 1987; Jarvis and Mulligan, 2010). Theinteractions of these local and regional influences on thedistribution of TMCF can be quite pronounced. The sheermass of large mountains exposed to intense radiation dur-ing cloudless periods is believed to raise the temperatureof the overlying air sufficiently to decrease the lapserate and enable plants to extend their altitudinal range.This effect is commonly referred to as the ‘mass ele-vation’ or ‘telescoping’ effect and has been recognizedfor many decades (Schroter, 1926; Van Steenis, 1972;Whitmore, 1998). More recent research has indicated thatlow-statured, mossy forests occurring at relatively lowelevations (<1000 m.a.s.l.) on small coastal mountains inthe tropics (cf. Table II) can be attributed to persistently

high soil moisture levels rather than to a steeper tem-perature lapse rate (Bruijnzeel et al., 2010b; cf. Romanet al., 2010). Further support for this contention comesfrom the observation that stunting is generally most pro-nounced in areas with high rainfall (Van Steenis, 1972;Bruijnzeel et al., 1993).

Cloud forest areal extent and regional distribution

As a result of the complex interactions describedabove, mapping of actual or potential areas with TMCFoccurrence at the pan-tropical scale is a major challenge(Bubb et al., 2004; Scatena et al., 2010). A map of theglobal distribution of TMCF cannot be compiled fromexisting national forest assessments because—with theexception of Mexico (Vazquez-Garcıa, 1995) and thePhilippines (Penafiel, 1995)—national vegetation clas-sification schemes do not commonly or consistentlyinclude a separate ‘cloud forest’ category. Thus, differ-ent approaches are needed. Elevation has been used asa proxy for the climatic (temperature, rainfall, fog inci-dence) and edaphic (soil water status, acidity) conditionsthat tend to be associated with ‘cloud forest’ (e.g. Cam-panella, 1995; Bubb et al., 2004; cf. Table II). Scatenaet al. (2010) combined global forest cover assessmentswith global elevation data and the regionally-specific alti-tudinal limits of cloud forests listed in Table II to deriveestimates of the potential (i.e. assuming the entire areabetween the specified altitudinal limits to be forested)and actual areas of TMCF (all categories lumped). Thepotential area of TMCF was estimated to be in the orderof 380 000 km2 (or 2Ð5% of the world’s tropical for-est area), whereas the actual area (in the year 2000)was estimated at ¾215 000 km2 (or 1Ð4% of the world’stropical forests and 6Ð6% of all tropical montane forest;Table III). Taking these estimates of actual and poten-tial TMCF occurrence at face value would imply thatroughly 56% of the original forest still remains and thatsome 165 000 km2 of cloud forest have been converted.As to their relative distribution, ¾43% of all remainingcloud forests can be found in Asia (including northernAustralia) and Oceania, 41% in the Americas (includingthe Hawaiian archipelago) and 16% in Africa (Table III).

Most recently, TMCF distribution has also beenmapped using a hydro-climatic definition of cloudforest as ‘forest under frequent and/or persistent expo-sure to ground-level cloud (i.e. fog)’ in combinationwith new tropics-wide data sets on satellite-observedcloud frequency, climate and land cover (Mulligan,2010). Arguably, knowledge of the distribution of thesehydro-climatically defined cloud-affected forests (CAF)may be more relevant in a hydrological context thanthe distribution of altitude-based TMCF. The hydro-climatic approach to mapping TMCF has yielded a muchlarger global extent of cloud forest cover (2Ð21 Mkm2)than the previously discussed altitude-based estimates(215 000 km2), suggesting that CAF may represent asmuch as 14Ð2% of all tropical forests (Figure 1). Thesame data also suggest that ¾45% of the original CAF

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 7: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

still remains and that some 2Ð72 Mkm2 of CAF havebeen converted (Mulligan, 2010). Whilst there is rea-son to believe that the hydro-climatic approach slightlyover-estimates the area with TMCF [see detailed discus-sion by Mulligan (2010)], the results were tested witha high level of success against the more than 560 trop-ical sites listed by WCMC-UNEP as having confirmedcloud forest presence (Aldrich et al., 1997). The best fitbetween actual and modelled cloud forest presence wasobtained when using a threshold value for ground-levelcloud occurrence (i.e. fog) of at least 70% of the time(Mulligan, 2010; Figure 1). It is recognized that this isa relatively high level of fog occurrence, but the use ofeither higher or lower values of fog frequency resulted insignificant reductions in the proportion of observed cloudforests being correctly modelled as CAF [see Mulliganand Burke (2005b) for details on the model’s sensitivity].

CLOUD FOREST HYDROMETEOROLOGY

General climatic conditions

The more than 560 tropical sites with confirmed cloudforest presence (albeit unspecified in terms of cloudforest type; Aldrich et al., 1997; Figure 1) represent awide range of climatic conditions (rainfall and temper-ature, wind) and landscape settings (altitude, exposure,mountain size, distance to sea, bedrock geology). Jarvisand Mulligan (2010) employed spatial data sets derivedfrom the WorldClim data-base (Hijmans et al., 2004), todescribe the climate at 477 cloud forest sites as identi-fied by UNEP-WCMC. Further, comparisons were madebetween the climate of cloud forest sites and that of ran-domly generated sites covering forested areas throughoutthe montane tropics, with the aim of identifying the cli-matic variables most important in distinguishing TMCFfrom other tropical forests. TMCFs were found to be wet-ter (by 184 mm year�1 on average), cooler (by 4Ð2 °C onaverage) and less seasonally variable than other mon-tane forests. The most statistically significant differencesin climate between TMCFs and other montane forestswere: maximum temperature, mean temperature, rainfalland rainfall seasonality (in order of significance). Cloudforests also tend to be located closer to the coast (particu-larly in Asia) and at higher altitudes than montane forestsnot affected by cloud. Furthermore, cloud forests occupymore topographically exposed areas than do other mon-tane forests. Interestingly, cloud forest sites in Africa tendto be drier (average annual rainfall <1500 mm) and moreseasonal than their counterparts in Asia (¾2150 mm ofrain on average), whereas the cloud forests of the LatinAmerican and Caribbean region experience a wide rangeof rainfall conditions (Jarvis and Mulligan, 2010).

Total water inputs and net precipitation

Proper quantification of the total amounts of waterinputs received by TMCF and the relative contribu-tions by ‘ordinary’ (i.e. assumed vertical) precipitation,

near-horizontal wind-driven drizzle (drop sizes rangingbetween 100 and 500 µm) and rain (drop sizes >500 µm;Pruppacher and Klett, 1978), and CWI is fundamental toassessing the hydrological importance of intact and con-verted cloud forest areas. Because it is difficult to distin-guish drizzle from rain in precipitation records, the term‘precipitation’ is used in this article to denote either. Like-wise, the term wind-driven precipitation (WDR) refersto either form of near-horizontal precipitation, whereasthe term ‘occult’ precipitation (HP) is used to denotethe sum of CWI and WDR without making a distinctionbetween the two (cf. Frumau et al., 2010a). The impor-tance of occult contributions is illustrated by the resultsobtained by several early studies from Central Amer-ica that arguably contributed greatly to the reputationof TMCFs as suppliers of high amounts of streamflowthroughout the year. Zadroga (1981) compared the rain-fall and streamflow regimes for two groups of catchmentsin northern Costa Rica, one located on the (wetter) wind-ward Atlantic side of the Continental Divide and theother on the (drier) leeward Pacific side. Annual stream-flow from the Pacific catchments amounted to 34% ofthe rainfall and showed a clear seasonal flow patternthat followed that for rainfall, whereas annual streamflowfrom the Atlantic catchments roughly equalled rainfall(102%), and even exceeded rainfall inputs for seven outof 12 months (Figure 2). Whilst acknowledging that thehigh runoff coefficient derived for the Atlantic catch-ments was partly due to underestimation of rainfall inputsin the higher, rainier parts of the catchments that lackedrainfall measurement stations, Zadroga (1981) attributedthe very high streamflows primarily to unmeasured inputsof CWI. He also emphasized the fact that months withexcess streamflow over precipitation coincided with thedominant occurrence of moisture-laden clouds broughtin from the Caribbean by the trade winds. In addi-tion, Zadroga recognized that evaporative losses fromthese fog-ridden slopes should be low. These contentionswere subsequently confirmed by measurements of rain-fall, streamflow and climatically based estimates of evap-otranspiration for another Atlantic catchment located fur-ther south in Costa Rica (Calvo, 1986). Although bothof these early investigations must be considered ‘blackbox’ studies that did not quantify the underlying hydro-logical processes, further support came from comparativeobservations of rainfall and throughfall (TF) in Atlanticcloud forests in Puerto Rico (Weaver 1972), Costa Rica(Caceres, 1981) and Honduras (Stadtmuller and Agudelo,1990). These studies indicated that annual TF at exposedlocations could attain values of as much as 110–180%of measured rainfall.

To what extent are these early observations exemplaryfor the hydrological behaviour of TMCF in general?And how reliable are such direct comparisons of rain-fall and TF in view of such potentially disturbing factorsas wind-induced precipitation losses around rain gauges(e.g. Førland et al., 1996; Yang et al., 1998; Nespor andSevruk, 1999) and the effect of inclined precipitationfalling onto steeply sloping terrain as opposed onto a

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 8: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Figure 2. Contrasting rainfall and streamflow regimes for catchments situated on the Atlantic and Pacific slopes of northern Costa Rica (afterZadroga, 1981)

horizontal gauge orifice (Sharon, 1980; Herwitz and Slye,1992), relative to unmeasured contributions by CWI orWDR? Earlier reviews of the hydrometeorological litera-ture on TMCF (Bruijnzeel and Proctor, 1995; Bruijnzeel,2001, 2005) lacked sufficient data for a meta-analysisbut the proliferation of local studies of net precipitation(or at least of TF)—many of which are reported byBruijnzeel et al. (2010a)—now allows an analysis ofsome of these questions and whether different types ofTMCF do indeed exhibit different net precipitation frac-tions.

Table IV lists net precipitation data for lower montanerain forests that are little or not affected by fog and lowcloud (LMRF, n D 15), tall LMCF subject to moderatefog incidence (n D 23), UMCF of intermediate staturesubject to frequent fog incidence (n D 18) and stuntedSACF and ECF (n D 8). Figure 3 shows the averageamounts of rainfall (P), the fraction of rainfall becomingTF, and the leaf area index (LAI) for the respectivemontane forest types, whereas Figure 4 shows scatterplots of annual TF versus P at individual study sitesgrouped per forest type.

On the basis of the more than 60 local studies listedin Table IV, the following patterns emerge: (1) averageLAI values per forest type decrease from 5Ð54 š 1Ð81in LMRF through LMCF (4Ð67 š 1Ð11) and UMCF(3Ð96 š 1Ð25) to 3Ð10 š 1Ð21 in SACF–ECF; (2) P atSACF–ECF sites tends to be higher on average thanat sites representing the other three forest categories forwhich differences between groups were comparativelysmall and (3) averaged ratios of TF to P increase steadilyfrom LMRF to SACF, viz. from 72 š 7% (SD) in LMRF,to 81 š 11% in LMCF, 100 š 27% in UMCF, and 104 š25% in SACF–ECF (Figure 3).

Rigorous comparisons of the statistical differences inTF/P between the different forest types is limited by thesmall and uneven sample sizes as well as by differencesin the sampling methodologies used in the studies of

both TF and P. Nevertheless, comparisons of means andmedians using t-tests, Mann–Whitney rank sum testsand analysis of variance (ANOVA) where appropriate,do support the patterns observed in Figure 3. Moreover,there are significant differences (at p D 0Ð05) betweenthe means or medians of TF/P for UMCF and bothLMCF and LMRF (but not SACF–ECF). The medianTF/P value for LMCF is also significantly higher thanthe median value for LMRF.

Comparison of the slopes of the P versus TF graphsper forest type (Figure 4) also indicates that TF exceedsprecipitation as measured in the open at SACF–ECFsites, whereas the two are nearly equal at UMCF sites.To these TF fractions the fraction of P reaching the forestfloor as stemflow (SF) should be added. Unfortunately,not all studies of net precipitation have measured SFbut values observed in LMRF and LMCF are typicallyvery low (<1–2% of P), although occasionally somewhathigher values have been reported (up to 8Ð8%, Table IV).However, SF fractions obtained for UMCF vary widely(0Ð1–30Ð5%)—possibly reflecting variations in stem den-sity and degree of mossiness—with an overall averageclose to 10% of rainfall (median 3%); the correspond-ing range and average SF for SACF–ECF are 3–18%and 6Ð7%, respectively (Table IV). Thus, whilst overallamounts of precipitation reaching the forest floor are onlyslightly enhanced by adding the contributions by SF inLMRF and LMCF (to 73% and 83%, respectively), aver-age TF fractions for UMCF and SACF–ECF may beraised by ¾9% and ¾7%, respectively, after inclusion ofthe stemflow component.

The rather low TF fraction observed for manyLMRF (Table IV) is often attributed to the high epi-phyte loading of these forests (e.g. Cavelier et al.,1997; Ataroff, 1998; Fleischbein et al., 2005). Inter-estingly, although both LMCF and UMCF typicallyhave a much higher epiphytic biomass than LMRF(Frahm and Gradstein, 1991; Wolf, 1993; Kohler et al.,

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 9: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Tabl

eIV

.T

hrou

ghfa

ll(T

F),

stem

flow

(SF)

and

appa

rent

rain

fall

inte

rcep

tion

(Ei)

frac

tion

s(%

ofin

cide

ntpr

ecip

itat

ion)

and

appa

rent

clou

d-w

ater

inte

rcep

tion

(CW

I,m

mye

ar�1

;c

Dva

lue

corr

ecte

dfo

rw

ind

and

topo

grap

hic

effe

cts)

asm

easu

red

indi

ffer

ent

type

sof

trop

ical

mon

tane

rain

fore

st

Loc

atio

nan

dfo

rest

type

Ele

vati

on(m

.a.s

.l.)

MA

Pa

(mm

)L

AI/

H�/

(m)

TF

SF(%

ofP)

Ei

CW

I(m

mye

ar�1

)R

emar

ks

Mon

tane

rain

fore

sts

little

affe

cted

byfo

gB

oliv

ia,

Yun

gasb

1850

2310

�/20

79—

<21

—10

rovi

ngga

uges

,3

mon

ths,

wee

kly

Col

ombi

a,C

ordi

ller

aC

entr

alc

1700

3150

�/35

76—

<24

—6

rovi

ngga

uges

,15

mon

ths,

wee

kly

1950

2200

�/35

85—

<15

—id

emE

cuad

ord

1950

2320

7Ð3/2

567

1Ð132

—T

rans

ect

2on

ly;

15–

36fix

edga

uges

,36

mon

ths,

wee

kly

Idem

e19

6020

80�/

2771

—<

29—

31fix

edga

uges

,12

mon

ths,

wee

kly

Cos

taR

ica,

Mon

teve

rdef

1200

2500

3Ð5/2

972

—<

28—

5fix

edga

uges

,12

mon

ths,

biw

eekl

yC

osta

Ric

a,Ta

lam

anca

g29

0028

107Ð7

/35

702Ð0

28—

UM

RF,

30fix

edga

uges

,46

wee

ks,

wee

kly

Gua

tem

alah

2100

2500

—65

—<

35—

36fix

edga

uges

,44

wee

ks,

wee

kly;

Plik

ely

over

estim

ated

sinc

em

easu

red

at25

50m

;fo

rest

grad

ing

into

LM

CF

Indo

nesi

a(S

ulaw

esi)

i10

4229

006Ð4

/23

700Ð3

30—

Site

NF4

,30

fixed

gaug

es,

15–

17ev

ents

Mal

aysi

a,Pe

nins

ular

j16

0023

00�/

1462

2Ð236

—20

rovi

ngga

uges

,68

wee

ks,

2–

3da

ysPa

nam

ak12

0036

80�/

3063

0Ð437

—50

trou

ghs

(0Ð05

m2),

12m

onth

s,da

ily

Papu

aN

ewG

uine

al25

0038

005Ð5

/30

67<

132

—32

fixed

gaug

es,

8m

onth

s,bi

wee

kly

Peru

,C

entr

alC

ordi

ller

am24

7022

202Ð9

/20

700Ð1

30—

20tr

ough

s(0

Ð1m

2),

54w

eeks

,w

eekl

yTa

nzan

ia,

Usa

mba

raM

tsn,

o15

0012

30�/

3579

120

—8

fixed

gage

s,30

mon

ths,

daily

Low

erm

onta

necl

oud

fore

sts

Aus

tral

ia,

SEQ

ueen

slan

dp10

0013

505Ð0

/35

902

834

325

trou

ghs,

31m

onth

s,2

–3

wee

kly

Aus

tral

ia,

NQ

ld,

UB

site

q,r

1050

2985

4Ð1/2

579

8Ð812

904

c4

–6

trou

ghs

(2Ð4–

3Ð6m

2),

25m

onth

s,co

ntin

uous

reco

rdin

gId

em,

Mt

Lew

issi

tesq,

r11

0033

154Ð5

/32

773

2416

0c

idem

,23

mo

1160

2610

4Ð1/1

277

617

401

cid

em,

35m

oB

oliv

ia,

Yun

gasb

2600

3970

�/18

74—

<26

—10

rovi

ngga

uges

,3

mon

ths,

wee

kly,

fore

stgr

adin

gin

toU

MC

FC

hina

,Y

unna

n,A

ilao

Mnt

.s25

0019

305Ð8

/25

872

11—

12fix

edga

uges

,24

mon

ths,

1–

3ð/d

ayId

em,

Xis

huan

gban

nat

750

1485

5Ð0/3

678

4Ð817

89(d

rip)

6fix

edco

llec

tors

(0Ð6

m2),

36m

onth

s,ev

ent-

base

dC

olom

bia,

Cen

tral

Cor

dill

erau,

v25

5021

15�/

2387

Ð5—

12Ð5

—20

fixed

gaug

es,

12m

onth

s,w

eekl

yC

osta

Ric

a,M

onte

verd

eW

indw

ard

fore

stf

1200

6390

3Ð4/2

870

—<

30—

5fix

edga

uges

,12

mon

ths,

biw

eekl

yL

eew

ard

fore

stw

,f15

0025

203Ð6

/25

65—

<35

—20

fixed

gaug

es,

12m

onth

s,1

–3

days

Cos

taR

ica,

Mon

tede

los

Oli

vosx

Win

dwar

dfo

rest

1500

3300

90Ð5

—�/

40<

9Ð5—

20tr

ough

s(0

Ð1m

2),

12m

onth

s,da

ily

Idem

1500

3300

�/25

106

—�7

—Id

emE

cuad

ore

2030

2140

5Ð7/2

085

114

—31

fixed

gaug

es,

12m

onth

s,w

eekl

yId

eme

2175

2500

4Ð9/1

591

18

—Id

emG

uate

mal

ay(w

indw

ard)

2200

2600

�/�

81<

118

—3

–6

fixed

coll

ecto

rs(0

Ð22m

2),

12m

onth

s,2ð

/wee

kH

awai

i,M

aui

(lee

war

d)z

1219

1010

�/2Ð5

88—

3116

6c

4fix

edco

llec

tors

(0Ð07

m2

each

),75

1da

ys,

Hon

dura

sy90

0–

1400

>35

00�/

�95

—<

5—

3–

6fix

edco

llec

tors

(0Ð22

m2),

12m

onth

s,2ð

/wee

k?In

done

sia

(WJa

va)aa

1750

3300

�/�

80—

<20

—4

coll

ecto

rsof

1m

2,

12m

onth

s,w

eekl

yM

exic

o,V

erac

ruzab

2100

3000

6Ð6/2

782

Ð5—

1754

4ro

ving

coll

ecto

rs(1

Ð2m

2ea

ch),

12m

onth

s,co

ntin

uous

reco

rdin

g

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Sticky Note
Table IV. Line 4: (1) delete '1950' in LH-column and (2) replace '-/35' by '1950 in 3rd line of 3rd column (MAP).
Page 10: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Tabl

eIV

.(C

onti

nued

)

Loc

atio

nan

dfo

rest

type

Ele

vati

on(m

.a.s

.l.)

MA

Pa

(mm

)L

AI/

H�/

(m)

TF

SF(%

ofP)

Ei

CW

I(m

mye

ar�1

)R

emar

ks

Tanz

ania

,M

tK

ilim

anja

roac

2250

2480

�/40

82<

117

—10

fixed

gaug

es,

12m

onth

s,w

eekl

yId

em,

dry

year

1960

�/40

70<

129

—Id

emT

hail

and,

Kog

Maad

,ae

1265

2085

4Ð0/2

589

1Ð59Ð5

114

30fix

edga

uges

,48

mon

ths,

daily

Ven

ezue

la,

San

Eus

ebio

af23

0015

75�/

2779

Ð51

19Ð5

—30

fixed

gaug

es,

12m

onth

s,1

–2ð

/wee

kV

enez

uela

,L

aM

ucuy

ag23

5031

252Ð9

/25

540Ð2

46—

6fix

edco

llect

ors

(0Ð6

m2),

12m

onth

s,w

eekl

yU

pper

mon

tane

clou

dfo

rest

sA

ustr

alia

,Q

ueen

slan

d,B

Kq,

r15

6053

003Ð3

/811

714

�31

1190

c4

–6

trou

ghs

(tot

alli

ng2Ð4

–3Ð6

m2),

21m

onth

s,co

ntin

uous

reco

rdin

gC

olom

bia,

Cen

tral

Cor

dill

erac

3000

1700

�/�

89—

<11

—6

rovi

ngga

uges

,15

mon

ths,

wee

kly

Idem

,C

entr

alC

ordi

ller

au,v

3370

1455

�/22

87Ð5

0Ð112

Ð5—

20fix

edga

uges

,12

mon

ths,

wee

kly

Idem

,Z

ipaq

uira

ah32

5016

153Ð6

/15

892

9—

20ro

ving

gaug

es,

24m

onth

s,w

eekl

yC

osta

Ric

a,M

onte

verd

ef,ai

1500

4310

3Ð4/1

011

8—

�18

—5

fixed

gaug

es,

12m

onth

s,bi

wee

kly

Idem

aj,a

k14

5060

003Ð0

/22

108

1�9

1300

c30

rovi

ngga

uges

,12

mon

ths,

dail

yG

uate

mal

ay(w

indw

ard)

2400

2500

�/�

113

1�1

4—

3–

6fix

edco

llec

tors

(0Ð22

m2),

12m

onth

s,2ð

/wee

kG

uate

mal

ah(w

indw

ard)

2550

2600

�/�

881

11—

36fix

edga

uges

,44

wee

ks,

wee

kly

Haw

aiiz

(win

dwar

d)19

5127

00�/

613

1—

�31

1212

c4

fixed

colle

ctor

s(0

Ð07m

2ea

ch),

64da

ysof

cont

inuo

usre

cord

ing

Hon

dura

sal17

9515

00�/

�4

fixed

colle

ctor

s(0

Ð25m

2),

12m

onth

s,da

ily

Con

cave

slop

e94

—6

—C

onve

xsl

ope

129

—�2

9—

Rid

geto

p17

9—

�79

—Ja

mai

ca,

PMul

lfo

rest

am18

1028

505Ð0

/12

7312

14—

12ro

ving

gaug

es,

12m

onth

s,3

–4

days

;C1

reco

rdin

gtr

ough

Mal

aysi

a,Pe

nins

ular

j20

3021

15�/

7Ð564

30Ð5

5Ð5—

20ro

ving

gage

s,68

wee

ks,

wee

kly

Peru

,C

entr

alC

ordi

ller

am28

1527

502Ð5

/14

92Ð5

<0Ð1

7Ð5—

20fix

edco

llec

tors

(0Ð11

5m

2),

12m

onth

s,w

eekl

yPh

ilipp

ines

,L

uzon

an22

0039

10�/

986

122

—16

fixed

coll

ecto

rs(0

Ð041

m2),

60m

onth

s,da

ily

Puer

toR

ico

Palm

fore

stao

900

4500

4Ð6/1

169

265

—20

rovi

ngco

llec

tors

(0Ð02

54m

2),

6m

onth

s,da

ily

Elfi

nan

dsu

b-al

pine

clou

dfo

rest

sB

oliv

ia,

Yun

gasb

3050

5150

�/10

75—

<25

—10

rovi

ngga

uges

,3

mon

ths,

wee

kly

Haw

aiiap

2600

500

�/8

75—

<25

222

fixed

coll

ecto

rs(0

Ð125

m2),

12m

onth

s,co

ntin

uous

reco

rdin

gJa

mai

ca,

MM

orFo

rest

am,a

q18

2528

554Ð1

/860

ar18

22—

12ro

ving

gaug

es,

12m

onth

s,3

–4

days

;C1

reco

rdin

gtr

ough

Puer

toR

ico,

Eas

tPe

akE

lfin,

win

dwar

das,a

t10

0054

002Ð0

/2Ð5

115

5�2

0—

60fix

edco

llec

tors

,8

mon

ths,

dail

yId

emri

dge

1015

4800

�/�

125

9Ð5�3

5—

Idem

Idem

,le

ewar

d93

060

00�/

�96

31

—Id

emE

lfin,

win

dwar

dau10

0054

002Ð0

/2Ð5

165

5Ð710

770

(fog

only

)10

45c

20ro

ving

gaug

es,

44da

ys,

2xda

y�1

Idem

,sh

elte

redav

970

4500

2Ð1/3

108

(5)

�13

400

(fog

only

)51

1c

45ro

ving

&22

fixed

gaug

es,

97da

ys,

twic

e/w

eek;

C3tr

ough

s(1

Ð0m

2)

Reu

nion

,Fr

ance

aw21

5030

00�/

312

4—

�24

—8

fixed

gaug

es,

8m

onth

s,co

ntin

uous

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Sticky Note
After Puerto Rico, pls insert a comma (i.e. before Palm forest). Also, bottom line of table: place an accent on the e in Reunion pls.
Page 11: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTSSp

ain,

La

Gom

eraax

1290

660

4Ð2/9

126

2Ð8�2

9—

2fix

edco

llec

tors

(0Ð20

m2),

19m

onth

s,co

ntin

uous

aM

ean

annu

alpr

ecip

itatio

n.b

Ger

old

etal

.(2

008)

.c

Vis

(198

6).

dFl

eisc

hbei

net

al.

(200

6).

eO

eske

ret

al.

(201

0).

fH

ager

and

Doh

renb

usch

(201

0).

gH

olsc

her

etal

.(2

004)

.h

Hol

der

(200

4).

iD

ietz

etal

.(2

006)

.jK

umar

an(2

008)

.k

Cav

elie

ret

al.

(199

7).

lE

dwar

ds(1

982)

.m

Gom

ez-P

eral

taet

al.

(200

8).

nL

undg

ren

and

Lun

dgre

n(1

979)

.o

Pocs

(198

0).

pH

utle

yet

al.

(199

7).

qM

cJan

net

etal

.(2

007a

).rM

cJan

net

etal

.(2

007b

).s

Liu

etal

.(2

002)

.tL

iuet

al.

(200

7).

uV

enek

laas

and

Van

Ek

(199

0).

vV

enek

laas

etal

.(1

990)

.w

Cla

rket

al.

(199

8).

xFa

llas

(200

2).

yB

row

net

al.

(199

6).

zG

iam

bellu

caet

al.

(201

0b).

aaG

ongg

rijp

(194

1).

abH

olw

erda

etal

.(2

010c

).ac

Schr

umpf

etal

.(2

010)

.ad

Tana

kaet

al.

(200

5).

aeTa

naka

etal

.(2

010)

.af

Stei

nhar

dt(1

979)

.ag

Ata

roff

and

Rad

a(2

000)

.ah

Tobo

n(2

009)

.ai

Nad

karn

i(1

984)

.aj

K.F

.A.

Frum

au(p

erso

nal

com

mun

icat

ion)

.ak

Alth

ough

clas

sifie

das

LM

CF

onth

eba

sis

offo

rest

stat

ure

alon

e,th

isfo

rest

isco

nsid

ered

tobe

have

asan

UM

CF

due

tove

ryw

etan

dw

indy

cond

ition

s;re

lativ

ely

tall

stat

ure

may

beex

plai

ned

byw

ell-

drai

ned

volc

anic

soils

desp

iteex

cess

ive

rain

fall.

alSt

adtm

ulle

ran

dA

gude

lo(1

990)

.am

Haf

kens

chei

det

al.

(200

2).

anM

aman

teo

and

Ver

acio

n(1

985)

.ao

Hol

wer

da(2

005)

.ap

Juvi

kan

dN

ulle

t(1

995a

).aq

Tann

er(1

980)

.ar

Prob

ably

unde

rest

imat

ed.

asW

eave

r(1

972)

.at

Wea

ver

etal

.(1

986)

.au

Hol

wer

daet

al.

(200

6).

avH

olw

erda

etal

.(2

010b

).aw

Gab

riel

and

Jauz

e(2

008)

.ax

Gar

cıa-

Sant

osan

dB

ruijn

zeel

(201

0).

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Sticky Note
Since Palm forest was moved up one line (Puerto Rico, page 10) and MMor forest went up one line (Jamaica, page 10) it is surprising to see that Garcia-Santos is not part of the table on page 10 and instead is placed on its own at the top of page 11. Is there a possibility to include this study on page 10???
Page 12: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Figure 3. Mean and median annual rainfall, relative amounts of TF (% of rainfall), and LAI for LMRF, LMCF, UMCF and SACF–ECF. Box plotsdenote 25th and 75th percentiles. Individual data are listed in Table IV

Figure 4. Linear regression relationships linking amounts of TF and P in LMRF, LMCF, UMCF and SACF–ECF. Individual data are listed inTable IV

2007; Table I), this apparently does not have a reduc-ing effect on the magnitude of the TF fraction in therespective types of cloud forests (Figure 3; Table IV).On the basis of measurement and modelling of thewater dynamics of epiphytes and bryophytes in amontane forest in Costa Rica not subject to muchfog, Holscher et al. (2004) concluded that—whilst the

potential (total) water storage capacity of the epiphytesand mosses was very high—their actual (effective)capacity to accommodate additional precipitation or cloudwater was modest at best because the epiphytes tendto retain already stored water for long periods (cf.Kohler et al., 2010; Oesker et al., 2010; Tobon et al.,2010a).

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 13: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Although the high TF fractions obtained for UMCFand SACF–ECF (Figure 3) must partly reflect occultcontributions by CWI and WDR captured by the canopy,but missed by a standard-type rain gauge (Kerfoot, 1968;cf. Zadroga, 1981; see also the section on QuantifyingCWI), it is pertinent to note that wind speeds at exposedUMCF and SACF–ECF sites tend to be much higher thanthose experienced lower down on the slopes by the moresheltered LMCF and LMRF (Holwerda, 2005; Motzer,2005; Bendix et al., 2008). Therefore, rainfall amountsmeasured at exposed, windy sites are in danger of being(grossly) underestimated and need to be corrected for(1) wind losses around the gauge (Førland et al., 1996;Yang et al., 1998; Nespor and Sevruk, 1999) and (2) theinteractive effects of inclined precipitation falling onto(steeply) sloping terrain as opposed to a horizontallypositioned gauge orifice (Sharon, 1980; Herwitz and Slye,1992; cf. Frumau et al., 2010a). To date, only a handfulof studies have applied such corrections (e.g. Holwerdaet al., 2006; McJannet et al., 2007a,b; Holwerda et al.,2010b; Garcıa-Santos and Bruijnzeel, 2010; Giambellucaet al., 2010a,b). However, the magnitude of the correctioncan be substantial (>20%)—depending on the localsituation. As such, the higher average TF fractionsderived for UMCF and SACF–ECF are likely to be duein part to underestimation of incident precipitation ina number of cases (see, e.g. Prada et al. (2009) for arecent example), rather than merely attributable to occultcontributions missed by the rain gauge. However, the lackof information on wind speeds for many studies listed inTable IV presently does not permit a quantification of theeffect for all but a limited number of sites (see also thesection on Modelling regional patterns of CWI).

Several powerful modelling approaches of variousdegrees of complexity are available to estimate inputsby inclined precipitation and WDR at the landscapescale (e.g. Blocken et al., 2005, 2006; Mulligan andBurke, 2005a). However, validation of model predictionsthrough field measurements remains a challenge, largelybecause of the difficulty of adequately measuring precip-itation inputs under conditions of strong wind in complexterrain (Blocken et al., 2006). Arguably, the sphericalrain gauges of Chang and Flannery (2001) and the‘potential rain gauge’ of Frumau et al. (2010b)—whichallows quantification of the angle of incidence of theprecipitation—may prove useful in this regard. How-ever, both devices have received only limited testingthus far (Chang and Harrison, 2005; cf. Holwerda et al.,2010a).

Local topography, including slope gradient, aspect andexposure also exerts a major influence on atmosphericinputs, as demonstrated by the high small-scale variabil-ity in relative and absolute amounts of TF vis a vistopographical position reported by Weaver (1972) forECF in Puerto Rico (see also Holwerda et al., 2006,2010b), by Juvik and Ekern (1978) in Hawai‘i (cf. DeLayand Giambelluca, 2010; Giambelluca et al., 2010a,b), byStadtmuller and Agudelo (1990) in Honduras, by Hagerand Dohrenbusch (2010) in northern Costa Rica, and

by Garcıa-Santos (2007) in La Gomera (Canary Islands;cf. Marzol-Jaen et al., 2010). Future work will need tomore explicitly quantify inputs from WDR and CWI ina spatially distributed manner (see also the section onModelling regional patterns of CWI).

Quantifying CWI

The quantification of CWI by a forest canopy hasimproved considerably in recent years: from the simplis-tic equating of (1) the catch by some type of passivefog gauge to that by an actual live forest canopy (e.g.Ataroff, 1998) or (2) assigning all excess TF over rainfallto CWI (e.g. Holder, 2003), via (3) comparisons of TFversus P relationships for times with and without fog (e.g.Harr, 1982; Hafkenscheid et al., 2002) or for times withand without strong winds (Clark et al., 1998; cf. Tanakaet al., 2010), to (4) the more explicit process-based wet-canopy water budget method (WCWB; Juvik and Nullet,1995a; Hutley et al., 1997; Holwerda et al., 2006; McJan-net et al., 2007b,d; Holwerda et al., 2010a,b; Giambel-luca et al., 2010a,b; Schmid et al., 2010; Takahashi et al.,2010) and (5) direct measurement using sophisticatededdy-covariance and spectrometer equipment (Eugsteret al., 2006; Holwerda et al., 2006; Beiderwieden et al.,2008; Schmid et al., 2010). Arguably, results obtainedwith the WCWB method may be considered the mostrealistic, provided wind- and topographical effects on pre-cipitation are incorporated in the analysis, but the confi-dence limits of the estimates are generally wide due to theaccumulation of error associated with the measurement ofthe respective components (see discussion in Holwerdaet al., 2006, 2010b; McJannet et al., 2007b; Giambellucaet al., 2010b). In a related method, the relative contribu-tions by precipitation and fog to TF can be evaluated byanalysing P, fog and TF for their stable isotope contentand using a mass-budget approach (Schmid et al., 2010;cf. Scholl et al., 2010). However, contrasts in stable iso-tope levels in P and fog water need to be sufficiently largefor this approach to give meaningful results (Dawson,1998; Scholl et al., 2010). Similarly, rates of CWI deter-mined with the eddy-covariance method in sloping terrainmay be (heavily) underestimated due to the condensationof water vapour as the air moves upslope. The result-ing difference in fog water deposition at the height ofmeasurement (typically several metres above the canopy)and that at the top of the canopy needs to be taken intoaccount to obtain a realistic value of CWI [see Holwerdaet al. (2006) and references therein].

Reported values of apparent CWI as obtained with theWCWB method for various TMCF locations (n D 15, ofwhich nine studies applied a correction to incident precip-itation to account for wind- and topographical effects; seeCWI column in Table IV) range from ¾22 mm year�1

in the low-rainfall zone above the TWI in Hawai‘i (esti-mated from data in the study of Juvik and Nullet, 1995a),and 54 mm year�1 under conditions of very low windspeeds near the top of the cloud belt in eastern Mex-ico (Holwerda et al., 2010c), to more than 1000 mm

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Cross-Out
brul
Inserted Text
c
Page 14: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

year�1 (range 1045–1990 mm year�1; cf. Figure 8) atwind-exposed sites within the zone of maximum cloudoccurrence in Puerto Rico (Holwerda et al., 2006), theHawaiian archipelago (Giambelluca et al., 2010b; Taka-hashi et al., 2010) and northern Queensland (McJannetet al., 2007b). Expressed on a per day basis, apparentCWI varies almost a 100-fold, from 0Ð06 to 5Ð45 mmday�1 (mean 1Ð93 š 1Ð74 (S.D.) mm day�1; n D 11, withP corrected for wind and topographical effects whereaverage wind speed exceeds 1Ð5 m s�1). Giambellucaand Gerold (2011) regrouped the results for many of theCWI-measurement sites listed in Table IV plus new workfrom the Hawaiian archipelago (including Juvik et al.,2010), according to site exposure. Leeward sites had amean daily CWI of 0Ð3 š 0Ð2 mm (19 š 8% of P), versus1Ð6 š 1Ð4 mm (16 š 15% of P) for windward sites, and4Ð0 š 2Ð6 mm (154 š 163% of P) for ridge-top sites andisolated trees. Although the data compiled by Giambel-luca and Gerold (2011) represent a mixture of WCWBestimates, directly measured rates of fog deposition, andamounts of fog drip (i.e. net CWI after wet-canopy evap-oration) and do not distinguish between cloud foresttypes, the results clearly illustrate the importance of siteexposure.

Modelling regional patterns of CWI

For both hydrological and conservation reasons it isof practical interest to identify ‘hot spots’ of CWI atvarious levels of scale (from local to continental). Fogclimates may be characterized and compared betweenlocations using some kind of passive fog gauge, butthe typically high spatial variability in fog frequencyand duration—and thus fog water inputs (cf. Lawtonet al., 2010)—and the many different gauges designs thathave been used in different studies (each with their spe-cific collection efficiency (Schell et al., 1992; Frumauet al., 2010a) and drawbacks [Juvik and Nullett, 1995b;Schemenauer and Cereceda, 1994, 1995; see Bruijnzeelet al. (2005) for fuller discussion] preclude direct com-parisons between regions. At the same time, the limiteddata set for measured CWI (n D 15, Table IV) does notpermit the identification of regional hot spots of CWIeither. The data for those studies that have corrected pre-cipitation inputs for wind and topographic effects (n D 8,excluding the extreme total of 1990 mm derived forBellenden Ker, northern Queensland) suggest a modestpositive correlation (r D 0Ð58) between annual appar-ent CWI (mm) and average wind speed (u, in m s�1).However, there is reason to believe that this correlationpartly reflects an increasing degree of ‘contamination’ ofapparent CWI by WDR, despite the use of wind correc-tion functions (see also the section on Modelled CWIacross the tropics). For example, on the basis of observa-tions of CWI on precipitation-free days, Holwerda et al.(2006) estimated an annual total of ¾770 mm for a wind-exposed ECF in Puerto Rico. Use of the WCWB method(including wind and slope corrections) at the same siteyielded an estimated annual ‘CWI’ of 1045 mm, sug-gesting as much as 275 mm of total occult contributions

may have consisted of WDR that was not accounted forby the wind-correction equations (cf. discussion by Fru-mau et al., 2010b). Corresponding values derived withthe same methods at a nearby but more sheltered ECFsite were 400 mm year�1 (fog-only) and 510 mm year�1

(all occult contributions; Holwerda et al., 2010b). Addi-tional information on fog liquid water content (LWC)might improve the relationship between CWI and windspeed but LWC has been measured at only a handful ofTMCF locations (e.g. Eugster et al., 2006; Schmid et al.,2010). In addition, LWC is typically derived for fog par-ticles <50 µm only [Eugster et al. (2006) and referencestherein], thereby missing the larger drop sizes that are stillamenable to near-horizontal transport by wind, includingdrizzle (cf. Blanchard, 1953; Frumau et al., 2010b).

An alternative approach is to model variations in CWIacross the landscape. Several models of various degreesof complexity are available for this [e.g. Yin and Arp,1994; Walmsley et al., 1996, 1999; Mulligan and Burke,2005a; see also discussion by Bruijnzeel et al. (2005)].Below, the first attempt at modelling CWI across theentire tropics is presented using the FIESTA fog deliverymodel (FIESTAFD) developed by Mulligan and Burke(2005a).

The FIESTAFD model

FIESTAFD is a spatially explicit model for the compu-tation of the water budget of tropical mountain regions,with particular emphasis on a physically based repre-sentation of inclined rainfall, CWI and ET. These pro-cesses are modelled on a raster pixel basis at 1 ha or1 km spatial resolution and at regional to continentalextents. The model operates with four time steps eachmonth designed to characterize the diurnal cycle withinthat month, producing 48 time steps in total to rep-resent a full year. Mean 50-year values of basic cli-mate variables for each month are combined with arecent land-cover data set to compute the water bal-ance in its most basic form: precipitation (includingeffects of inclined precipitation where appropriate) CCWI � ET. The model has no subsurface componentssince there are no data to parameterize soils and base-flow at these spatial scales. Thus, FIESTAFD is essen-tially a water budget model and not an explicit runoffmodel. FIESTAFD has been applied and tested widelythroughout Latin America and parts of tropical Africaand Asia. It is now embedded within the AguAAndespolicy support system and can be applied readily for anypart of the world through a web-based model interface(http://www.policysupport.org/links/waterworld). Themodel’s set of equations describing and linking therespective processes, its initial application to predictCWI and the water budget—as well as changes thereinafter prescribed changes in regional land cover andclimate—across Costa Rica (at 90 m resolution) andCentral America as a whole (at 1 km resolution) are doc-umented in detail by Mulligan and Burke (2005a). Thereis also a downloadable version of the model running in

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 15: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

the PCRaster GIS (Wesseling et al., 1996) with data setsfor Costa Rica at http://www.ambiotek.com/fiesta.

FIESTAFD is driven by global data sets for terrain(SRTM; Farr and Kobrick, 2000), climate (WorldClim;Hijmans et al., 2004) and land cover (MODIS VegetationContinuous Fields VCF2000; Hansen et al., 2006) andrequires 127 individual digital maps to run a simulation.These maps are used to spatially model inputs via inclinedrainfall (along the lines of Arazi et al., 1997) and CWI,with particular emphasis on the influence of wind speed,topographical exposure and vegetation structure (distin-guishing between tree- and herb-functional types) on thecapture of both rainfall and fog. To model CWI, infor-mation is required on (1) the spatial extent of ground-level cloud (i.e. fog), (2) the type of vegetation present(trees versus shorter vegetation), (3) the magnitude andvariability of sedimentation (settling) and impaction (byturbulent diffusion) of fog droplets and (4) the fog inter-ception efficiency of the vegetation present [estimatedas a function of the LAI in relation to the angle ofthe drops; see Mulligan and Burke (2005a) for details].The pan-tropical monthly and diurnal cloud climatologyderived by Mulligan (2006a,b) from the MODIS MOD35product is used to obtain atmospheric cloud frequency ateach time step. This is then combined with a lifting con-densation level (i.e. the cloud base) as computed fromtemperature lapse rate calculations over the climate gridto estimate the frequency of ground-level cloud. CloudLWC is not usually measured and in the model is there-fore assumed to scale linearly with the measured relativehumidity whilst total fog inputs are calculated as the sumof sedimentation and turbulent fluxes to the vegetativeleaf area (Mulligan and Burke, 2005a).

Fog inputs are calculated to occur predominantlythrough sedimentation in locations of minimal wind(e.g. topographical lees and hollows), and predominantlythrough turbulent diffusion (impaction) where vegetationis exposed (e.g. on windward slopes, ridges and aroundsummits). Whilst inputs via impaction can be veryhigh per unit area (cf. Giambelluca and Gerold, 2011),areas where the vegetation is highly exposed to strongfog-bearing airflows tend to be spatially restricted incomparison with the more extensive lee areas that may besubject to lower rates of fog deposition. Thus, althoughhigh local rates of impaction may create very wetindividual ‘hot spots’ (Juvik et al., 2010), overall foginputs via sedimentation may be higher when occurringover sufficiently large spatial scales. For a given pixel, itstopographical exposure, fog and wind speed conditions,as well as its vegetation structure (both height andpercentage cover of tree- and herb-functional types,with characteristic LAI values assigned to each type)determine the balance between the two processes of fogdelivery. Uniquely, FIESTAFD accounts for the impactof atmospheric and ground-level cloud frequency usingthe cloud climatology of Mulligan (2006a,b) on solarradiation inputs, and therefore on potential and actualET. Model outputs include the degree of precipitationinclination (and its interaction with topography), fog

impaction and sedimentation, ET and the water balanceas monthly and annual integrals.

At the regional to continental scales at whichFIESTAFD is typically applied, there are significantuncertainties in some of the model inputs, particularlyfor rainfall, wind speed and direction, land cover andeven topography. However, FIESTAFD has been usedpredominantly to indicate the likely effect of scenariosfor land-cover change on the water balance in compari-son with a baseline simulation, rather than to predict themagnitude of the water balance per se. In this way, themodel is less sensitive to error in input data but moresensitive to the representation of the physical processesunderlying the computations of CWI, inclined and wind-driven rain and ET and their spatial distribution (Mulliganand Burke, 2005a). In addition, when the model is used tocompare relative spatial patterns of CWI over large scales(as in the pan-tropical application below), it is also lesssensitive to absolute errors in the input data sets since thefocus is primarily on geographical patterns. Nevertheless,applying these models at finer temporal and spatial reso-lutions should be considered a goal of future research.

Modelled CWI inputs across the tropics

FIESTAFD was applied to assess the geographicalvariation in both absolute and relative (i.e. expressedas a percentage of total inputs, Pt D P C CWI) amountsof CWI between 23Ð5 °N and 35 °S (Figures 5–7). Theexercise indicates that for areas with high rainfall suchas Costa Rica, annual CWI generally makes up lessthan 5% of the corresponding Pt, with maximum valuesreaching ¾10% on exposed ridges (Figure 5b). Expressedin absolute terms, annual CWI for Costa Rica amountsto 50–250 mm over most upland areas, whereas on moreexposed ridges this approximates 300 mm (Figure 5a).By contrast, in areas with lower rainfall, for examplemuch of Mexico, the absolute values of annual fog inputscan be similar in magnitude to those found over CostaRica (100–200 mm, Figure 5a) but these now representa greater proportion of Pt (15–20%, Figure 5b), withvalues up to 30% in some areas during the driest months[not shown here, see the study of Mulligan and Burke(2005a)].

According to FIESTAFD, half of the Latin Ameri-can land area receives annual fog inputs of less than25 mm, whereas only 12% of the area receives morethan 100 mm. The latter areas include the Andes (withparticularly high inputs projected for Ecuador and north-ern Peru), parts of Central America and Mexico, as wellas the Guyana shield and small parts of central Amazo-nia (Figure 5a). Whilst the latter finding may come asa surprise given the lack of montane topography, valleyfogs are known to be a frequent phenomenon in the morn-ings both in central Amazonia (Araujo, 2009) and FrenchGuiana (Gradstein et al., 2010). However, expressed ona per day basis, the projected amounts of fog depositionin the latter areas are of the same order of magnitude asthe canopy storage capacity of lowland rain forest (Lloyd

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 16: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

(a)

(b)

Figure 5. Mean annual fog water inputs for Latin America according to the FIESTAFD model (scale truncated at high end): (a) absolute values inmm, (b) expressed as a percentage of total precipitation (rainfall plus fog)

and Marques, 1988; Jetten, 1996) and they are thereforeunlikely to generate fog drip, unless the fog is accompa-nied by rainfall (cf. Garcıa-Santos and Bruijnzeel, 2010).For 91% of Latin America the projected fog inputs repre-sent less than 7% of annual Pt. For most of the remaining9% of the land the contribution by CWI is less than 25%but for much of coastal Chile and coastal Peru the con-tribution by fog exceeds 75% of annual Pt (Figure 5b)which confirms reports on the ground (e.g. Aravena et al.,1989; Cereceda and Schemenauer, 1991; Calamini et al.,1998; Manrique et al., 2010).

Model results also indicate that 71% of Africa appar-ently receives less than 40 mm of fog input annu-ally, with the remaining 29% of the continent receivingbetween 40 and 350 mm year�1. Projected fog inputsare concentrated in tropical Central and West Africa

(Congo, Democratic Republic of Congo (DRC), Cen-tral African Republic (CAR), Gabon, Cameroon andEquatorial Guinea) with significant inputs also in partsof Ethiopia, Madagascar, Cote d’Ivoire and Liberia(Figure 6a). As a percentage of total inputs, the highestfog inputs are found in parts of Gabon, Angola, DRC,Cameroon, CAR and coastal Namibia (Figure 6b). For85% of the African land area, fog is estimated to con-tribute less than 7Ð5% of total inputs, with the remain-ing 15% of the land receiving up to 18% of Pt viaCWI (Figure 6b). Field measurements of CWI in trop-ical Africa appear to be lacking thus far (cf. Kerfoot,1968; Hemp, 2005a).

Fog inputs in tropical Asia are apparently widespread,with concentrations indicated by the model in easternNepal, Bhutan, eastern India, northern Thailand, Laosand Myanmar, central Borneo, West Java and Sumatra,

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 17: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

(a)

(b)

Figure 6. Mean annual fog water inputs across Africa according to the FIESTAFD model (scale truncated at high end): (a) absolute values in mm,(b) expressed as a percentage of total precipitation (rainfall plus fog)

as well as parts of Vietnam, the Philippines, Irian Jaya,Papua New Guinea and Sulawesi. Some 63% of the landarea shows annual inputs of less than 36 mm with mostof the remaining 36% of the land apparently receivingbetween 75 and 250 mm and a few small areas receiv-ing much more (Figure 7a). Expressed as a percentageof Pt the greatest contributions are indicated for southernChina, northern Thailand, eastern Myanmar and north-ern Laos (Figure 7b). For 53% of the tropical Asian landsurface, CWI apparently contributes less than 1% of Pt.For most of the remaining 47% of the land fog is esti-mated to contribute between 1% and 15% of Pt, whereasin small areas (totalling 0Ð2% of the land surface) CWIis projected to contribute up to 30% of Pt (Figure 7b).Modelled results for northern Queensland, Australia (notincluded in Figure 7 for reasons of scale) suggest large

areas with no significant fog input, a few areas with upto 25 mm year�1 and some isolated exposed ridges with100 mm year�1. These amounts are very small relativeto Pt, contributing from close to zero to 7Ð5% on exposedridges. They are also very much lower than reportedannual totals of occult precipitation measured in the fieldat the plot scale (McJannet et al., 2007b; Table IV). Forthe Hawaiian archipelago, projected cloud-water inputs(again, not shown for reasons of scale) vary from 0 to300 mm year�1 in parts of Maui and the Big Island(Waimea area), representing up to 25% of Pt in iso-lated parts of the drier western half of the Big Island.As in Queensland, plot-scale field measured annual totalsof apparent CWI in the wetter parts of Maui (Giambel-luca et al., 2010b) and the Big Island (Takahashi et al.,

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 18: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

(a)

(b)

Figure 7. Mean annual fog inputs for Asia according to the FIESTAFD model (scale truncated at high end): (a) absolute values in mm, (b) expressedas a percentage of total precipitation (rainfall plus fog)

2010) are much higher than these modelled totals. Unfor-tunately, no field measurements of occult precipitation orCWI are available for many of the areas in Figure 7a thatare modelled to receive significant cloud-water inputs.However, MacQuarrie et al. (2001) reported that verylarge volumes of ‘cloud water’ are occasionally capturedby large fog collection screens in eastern Nepal, whereasLiu et al. (2004) stressed the ecohydrological importanceof frequent (radiation) fog-induced drip (determined at90 mm year�1) during the dry season in southern Yun-nan, China.

The results presented in Figures 5–7 are based on sim-ulations of CWI at a spatial resolution of 1 ð 1 km usinglong-term climatic means for rainfall, cloud frequency,temperature, humidity and wind speed, and coarse-scalerepresentations of vegetation structure based on globaldata sets. As such, modelled values for individual sites

located in complex topography and variable vegetationcover are not directly comparable with point measure-ments made at specific sites, with specific vegetationcharacteristics and for a specific period. Nevertheless, afirst-level assessment was made by comparing the annualCWI totals as projected by FIESTAFD with amounts ofoccult inputs (HP D CWI C possible contributions byWDR) as derived with the WCWB method (itself anestimate rather than a direct measurement) for the 15CWI-measurement sites listed in Table IV.

Because of locational uncertainty in the model’s inputdata sets and for some of the plot locations, a comparisonwas made between the field-measured point HP total andthe mean, minimum and maximum modelled fog inputsover the corresponding 1 km pixel, as well as minimum,maximum and mean for the two neighbouring pixels oneither side [i.e. a window of 2 km (24 pixels) around

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 19: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Figure 8. Comparison of field-measured (WCWB method) annual occult precipitation (HP D CWI C WDR, in mm) at specific cloud forest locations(n D 15) against corresponding values of cloud-water interception (CWI, mm) projected by the FIESTAFD model. Mean modelled values of CWI (�)represent the average values for the 1-km pixel under consideration plus 24 neighbouring pixels (two on all sides). Average measured wind speedsfor the field measurement sites (+) are added for comparison. Where possible, a distinction has been made between measured CWI values with ( )

(n D 9) and without (ž) (n D 15) corrections of precipitation inputs for the interactive effects between falling rain drops, wind and topography

the pixel under consideration]. The results (Figure 8)show that measured values of apparent CWI (i.e. HP)are generally significantly higher than modelled valuesof CWI, except for sites with low wind speeds. Tosome extent this is to be expected because the modelledvalues include fog only, whereas the measured values arelikely to include some WDR captured more efficientlyby the canopies than by a rain gauge, particularly athigh wind speeds. For sites with low average windspeeds and presumably negligible WDR (u < 1Ð5 ms�1) the measured apparent CWI values are close tothe corresponding modelled mean values (Figure 8).However, close to and above wind speeds of 1Ð5 m s�1

measured values of apparent CWI increase sharply incomparison with modelled values of CWI, possiblybecause inputs via WDR increase with wind speedand may thus increasingly ‘contaminate’ the ‘CWI’measurement. Although the interactive effects of windand topography were included in a number of field studies(Holwerda et al., 2006, 2010b; McJannet et al., 2007b;Giambelluca et al., 2010b; Figure 8), the correctionsmay not always have been representative of prevailingprecipitation drop sizes. For example, the findings ofFrumau et al. (2010b) in a cloud forest in northern CostaRica indicate smaller median drop sizes for given rainfallintensities compared to the values predicted by suchwidely used relationships as that of Laws and Parsons(1943). Needless to say, smaller drops are more amenableto be deflected by wind and the conventional correctionstherefore tend to underestimate precipitation inclinations(see discussion by Frumau et al., 2010b). Further work isneeded on the characterization of precipitation drop sizes

in montane cloud forest settings (cf. Blanchard, 1953;Frumau et al., 2010b)

It is also likely that the FIESTAFD model under-predicts the impact of wind speed on CWI becausetopographical smoothing of rugged terrain represented at1 km spatial resolution reduces topographical exposureand thus fog inputs. Indeed, a previous application ofFIESTAFD at a resolution of 90 m in northern CostaRica (Mulligan and Burke, 2005a) did yield fog waterinputs to exposed slopes and ridges that were of thesame order of magnitude as those derived by independentmeans (eddy-covariance data) by Schmid et al. (2010).Clearly, applying fog deposition models at sufficientlyfine spatial resolutions to transects that include windward,ridge-top and leeward sites and for which independentmeasurements of CWI that are unaffected by WDR areavailable should be considered a prime goal of future fogdeposition research. Without such a comparison of high-resolution simulations with directly-measured fog inputs,it is difficult to ascertain whether the FIESTAFD modelunderestimates the magnitudes of fog inputs.

A further check of the degree of ‘contamination’ ofmeasured apparent CWI by ungauged amounts of WDR ispossible by examining the relationship of the discrepancybetween field-measured and modelled CWI on the onehand, and the enhancement of conventionally measuredP by concentration on slopes that are topographicallyexposed to inclined and wind-driven rain as modelled byFIESTAFD (Figure 9). Clearly, a significant topograph-ical potential for enhanced inputs via WDR may leadto a significant underestimation of modelled CWI com-pared with the field measurement (Figure 9). Indeed, in

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 20: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Figure 9. Modelled minus measured apparent CWI versus modelledenhancement of precipitation inputs by WDR (both in mm year�1) atthe 15 cloud forest locations for which wet-canopy water budget-based

estimates of occult precipitation inputs are available

areas where modelled WDR inputs are low, and in shel-tered leeward areas (i.e. negative values on the x-axisin Figure 9) field- and modelled CWI values are simi-lar. This holds true for all but two cases (Volcano NP1,Hawai‘i and Upper Barron, Queensland) where modelledWDR is relatively low but field-measured CWI values arestill significantly higher than modelled values. Althoughthe topography and aspects of the two deviating sites sug-gest they are not highly exposed to rain-bearing winds (atleast according to the FIESTAFD data sets), they do havehigh wind speeds (both measured and modelled). UpperBarron has a mean measured wind speed of 2Ð9 m s�1

(McJannet et al., 2007d) versus a modelled wind speed of

3Ð3 m s�1, whilst the corresponding values for the Volca-noes NP1 site are 4Ð4 m s�1 (measured; Takahashi et al.,2010) and 4Ð4 m s�1 (modelled). Therefore, whilst thetwo sites may not be particularly exposed to topograph-ical redistribution of rainfall because of their low slopegradients, their canopies are likely to be subject to sig-nificant WDR impacts.

Because of the difficulties of obtaining WCWB basedCWI measurements that are not influenced by WDRand of comparing point-scale field measurements withmodelled values at much coarser spatial scales, othermethods of model verification are worth exploring. Oneof these is to use cloud forests themselves as a ‘biosensor’by assuming that their geographical distribution reflectsareas of high fog inputs. For this, the FIESTAFD-modelled annual fog inputs were determined for eachof the 472 sites in the UNEP-WCMC world cloudforest data base with the most reliable coordinates(Aldrich et al., 1997; cf. Figure 1). Next, 472 pointswere taken randomly from areas defined as mountainous(>500 m.a.s.l. according to SRTM; Farr and Kobrick,2000) and forested (>40% tree cover according toVCF2000; Hansen et al., 2006) within the same tropicalregion as covered by the cloud forest data set. Becausethe two point data sets were selected independently ofthe model results, the TMCF points should exhibit higherannual CWI values on average than the randomly chosenmontane forest points, that is, if the model correctlydefines the distribution of CWI and if fog inputs areimportant to creating and maintaining cloud forests.

As shown in Figure 10, there is some overlap betweenthe two CWI frequency distributions. This is to beexpected because (1) not all cloud forests will have

Figure 10. Frequency distributions of modelled annual cloud-water interception (mm) for 472 sites with confirmed presence of tropical montanecloud forest (Aldrich et al., 1997) and for 472 randomly chosen tropical montane forest sites

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 21: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

high fog inputs—some may be maintained by very highrainfall (Mulligan and Burke, 2005b) and (2) becausesome of the randomly chosen ‘mountain forests’ mayin fact be cloud-affected forest in areas for whichthere are no cloud forest records in the UNEP-WCMCdatabase. Nevertheless, the TMCF data points show asignificant tendency for higher modelled annual CWItotals with a median value of 105 mm (mean 112 š 70(S.D.) compared with 79 mm (mean 96 š 70 S.D.) forthe randomly selected mountain forests. A two-sampleKolmogorov–Smirnov test also indicated that the twodistributions are drawn from different populations (p <0Ð01). Because the TMCF site data set and the foginterception model are independent, this result indicatesthat the geographical distribution of modelled fog inputslikely reflects TMCF distribution in reality.

Cloud forest water use

Being comprised of only 15 studies in total, the currentdata set on cloud forest water use [both for transpiration(Et) and total evapotranspiration (ET), i.e. including wet-canopy evaporation (Ei)] is much smaller than that forTF and stemflow. Nevertheless, considerable progresshas been made in the last five years. Whilst Bruijnzeeland Proctor (1995) were unable to report any annualvalues for Et other than indirect (and therefore highlyuncertain) estimates derived using the catchment waterbudget method, the number of studies using sap-flowequipment—and to a lesser extent micrometeorologicalmethods—has increased steadily in recent years. Apreliminary analysis of the available data (summarizedin Table V) is offered below, although the conclusionsneed to be viewed with caution given the small numberof replications for UMCF and SACF–ECF in particular.

Figure 11 summarizes the average annual totals ofET and its two main components (Et and Ei) for thethree types of cloud forests distinguished in this paper.No comparative data are available on transpiration forLMRF (cf. Bruijnzeel, 2005) but average total ET forthis type of forest is 1280 š 72 mm year�1 (n D 7). With1188 š 239 mm year�1 (n D 9) the average value forLMCF does not differ significantly from that for LMRF,but ET totals for UMCF (783 š 112 mm year�1, n D 3)and ECF (547 š 25 mm year�1, n D 2) are distinctlylower. And although the small number of observationsfor the latter forest types preclude statistical testing andrequire additional work, they tentatively confirm the needto distinguish between lower montane (cloud) forests onthe one hand, and upper montane (and elfin) forestson the other (Table V). Combining averaged data perforest type for precipitation inputs (Figure 3) with thecorresponding average ET values (Figure 11) gives someindication of the precipitation-surplus values that maybe expected for the respective forest types—and there-fore the probability of persistently high soil water levels.Although such estimates are obviously only approximate,the derived average annual precipitation-surplus valuesshow a clear increase from ¾1200 and ¾1500 mm in

LMRF and LMCF, respectively, to ¾2100 mm in UMCFand ¾2900 mm in ECF (cf. Kitayama, 1995; Schaweet al., 2010). The consequences of this increasing wetnesswith elevation for forest productivity, nutrient availabilityand, ultimately, forest stature have recently been dis-cussed elsewhere (Bruijnzeel et al., 2010b; Giambellucaand Gerold, 2011).

As discussed earlier for ET, average annual amountsof Et also decrease gradually with elevation: from¾645 mm in LMCF to ¾385 mm in UMCF and¾355 mm in ECF (Figure 11). Interestingly, both Et andET (not shown) are significantly related toforest LAI, also after normalization for net radiation input(Rn; Figure 12). In addition, the fraction of availableenergy used by the respective forest types for transpira-tion (Et/Rn) shows a marked reduction from 0Ð61 š 0Ð08in lowland evergreen rain forest (LERF) to 0Ð49 š 0Ð14 inLMCF, 0Ð37 š 0Ð06 in UMCF and 0Ð265 in ECF. In otherwords, one or more factors other than radiation and (byimplication) temperature, must be considered responsiblefor the observed decrease in Et with altitude. In viewof the demonstrated strong relationship between Et/Rn

and montane forest LAI (Figure 12), and the fact thatLAI under given climatic conditions tends to decreasesharply as the degree of waterlogging increases (Santiagoet al., 2000; Vernimmen et al., 2007), it is likely that theabove trend for Et at least partly reflects the unfavourableedaphic conditions associated with increasingly persis-tent waterlogged conditions [see Bruijnzeel et al. (2010b)for fuller discussion]. In addition, with the increasinglyhigh values of precipitation surplus found with increas-ing elevation (Schawe et al., 2010), the canopy is likelyto remain wet for longer periods of time (cf. Holw-erda et al., 2010a), thereby further reducing transpiration(Santiago et al., 2000; Burgess and Dawson, 2004; Holw-erda, 2005; Garcıa-Santos, 2007; McJannet et al., 2007c)and photosynthetic activity (Letts and Mulligan, 2005;Mildenberger et al., 2009; Letts et al., 2010).

Across all sites listed in Table V, mean annual precip-itation (MAP) is negatively correlated with net radiantinputs, forest water uptake (Et), and LAI. Up to 42%of the variance in Et between sites can be explained byMAP, mean annual temperature, and Rn (controlled inturn by the frequency and degree of cloud cover). LAIalone explains as much as 73% of the variance in Et

(Figure 12) and 65% of that in ET (not shown).By and large, although varying considerably, relative

contributions by Et to total ET are roughly as importantas those by wet-canopy evaporation, Ei (¾50% each)in both LMCF and UMCF. However, there are insuf-ficient data for meaningful statements in this regard forSACF–ECF (see Table V for details). Occasionally, veryhigh values have been derived for Ei (e.g. for severalnear-coastal forests in northern Queensland; McJannetet al., 2007d). It is not clear to what extent this findingreflects differences in methodology or a physical reality(e.g. advected energy from the ocean or from deforestedlowlands sustaining high evaporation rates during rain-fall) given that various other ‘maritime’ sites (Puerto

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 22: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Table V. Annual transpiration (Et), wet canopy (interception) evaporation (Ei), and total evapotranspiration (ET) as measured indifferent types of tropical rain forest (mm)

Location and forest type Elevation(m.a.s.l.)

MAPŁ

(mm)LAI

-Et Ei

(mm year�1)ET Ei/Et

(%)Et/Rn

(�)Setting/method for Et or ET

Lowland Evergreen Rain ForestCentral Amazonia, Duckea 80 2500 6Ð0 980 330 1310 25 0Ð59 Continental (C),

micrometeorologyPasoh, Peninsular Malaysiab 100 1800 6Ð5 1206 242 1448 17 0Ð69 Maritime (M),

micrometeorologyLambir Hills, Sarawakc 50 2740 6Ð2 946 357 1303 27 0Ð54 M, micrometeorologyLower montane rain forest little affected by fogMadagascar, Perinetd 1010 2080 1295 M, watershed budget for

ETColombia, Sierra Nevadae 1150 1985 1265 IdemEcuadorf 1950 2050 7Ð3 626 655 1281 105 C, catchment water

budget for ET and EtKenya, Kerichog 2200 2130 1337 C, catchment water

budget for ETIdem, sub-basing 2350 2015 1240 IdemKenya, Kimakiah 2440 2305 1156 IdemTanzania, Mbeyai 2500 1925 1381 IdemLower montane cloud forestQueensland, Gambubalj 1000 1350 5Ð5 845 414 1259 33 — C/M, sapflow, soil water

budgetIdem, Upper Barronk 1050 2985 4Ð1 591 854 1445 59 0Ð37 M, sapflowIdem, Mt. Lewisk 1100 3040 4Ð5 579 880 1459 60 0Ð42 M, sapflowHawai’i, Volcano NPl,m 1219 2500 4Ð8 645 588 1233 48 0Ð30 M, micrometeorologyIndonesia, Javan 1750 3300 — 510 660 1170 56 — M, catchment water

budget for ET and Et

Ecuadoro,p 2030 2140 5Ð7 555 364 919 40 0Ð69 C, sapflowMexico, Veracruzq,r 2100 3100 6Ð6 876 542 1418 38 0Ð63 C/M, sapflowThailand, Kog-Mas 1268 1768 4Ð0 626 530 812 46 0Ð46 C, sapflow and

micrometeorologyVenezuela, San Eusebiot 2300 1465 — 675 305 980 31 0Ð44 C/M, energy and site

water budgetsVenezuela, La Mucuyu 2350 3125 2Ð9 558 C, porometry for Et

Upper montane cloud forestCosta Rica, Monteverdev,w 1450 6000 3Ð0 361 415 776 53 0Ð36 M, micrometeorologyQueensland, Bellenden Kerk 1560 7470 3Ð3 349 549 898 61 0Ð31 M, sapflowPuerto Rico, Luquillo Mnts.u 900 4450 4Ð6 484 190 674 28 0Ð44 M, micrometeorologyElfin cloud forestPuerto Rico, Luquillo Mnts.x 1010 4500 2Ð1 296 268 564 47 0Ð27 M, micrometeorologyLa Gomera, Canary Islandsy,z 1270 900 4Ð2 412 117 529 22 0Ð26 M, sapflow

a Shuttleworth (1988).b Tani et al. (2003).c Kumagai et al. (2005).d Bailly et al. (1974).e Herrmann (1971).f Fleischbein et al. (2006).g Blackie (1979a).h Blackie (1979b).i Edwards (1979).j Hutley et al. (1997).k McJannet et al. (2007d).l Giambelluca et al. (2009).m T.W. Giambelluca (personal communication).n Gonggrijp (1941).o Motzer et al. (2010).p Oesker et al. (2010).q Gomez-Cardenas (2009).r Holwerda et al. (2010c).s Tanaka et al. (2003, 2010).t Steinhardt (1979).u Ataroff and Rada (2000).v Bruijnzeel (2006).w K.F.A. Frumau (personal communication).x Holwerda (2005).y Garcıa-Santos and Bruijnzeel (2010).z Garcıa-Santos (2007).

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 23: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Figure 11. Mean and median annual (a) rainfall interception, (b) transpiration and (c) total evapotranspiration for LMRF (ET only), LMCF, UMCFand SACF–ECF. Box plots denote 25th and 75th percentiles. Data are listed in Table V

Figure 12. Cloud forest transpiration (Et) versus LAI: (a) daily Et values(mm), (b) normalized for the evaporation equivalent of net radiant energy(Rn). LE D lowland evergreen rain forest, L D lower montane cloud

forest, U D upper montane cloud forest and EC D elfin cloud forest

Rico, Hawai‘i, Costa Rica) do not exhibit similarly highvalues (see Tables IV and V). Recently, Holwerda et al.(2010d) presented evidence in support of the idea thatthe high rates of wet-canopy evaporation observed insome tropical forests, but not in others reflect a topo-graphically enhanced aerodynamic roughness rather thanunmeasured additions of advected energy. There is a needfor additional observations of evaporation in UMCF andSACF–ECF, in particular.

CLOUD FOREST CONVERSION AND WATERYIELD

On the basis of the very large contrast in streamflowproduced by windward and leeward forests in northernCosta Rica (Figure 2), Zadroga (1981) postulated thatconversion of Atlantic cloud forest might well resultin significantly reduced streamflows. However, recentresults of detailed hydrological field and modelling stud-ies in cloud forest and pasture in the same area raisequestions about the validity of some of the standardassumptions as to why cloud forests produce such highamounts of streamflow. Conversion of cloud forest topasture in windward northern Costa Rica did not pro-duce the expected decreases in annual or even seasonalwater yield; rather the effect on streamflow was moreor less neutral. This surprising finding was attributed tothe compensatory effects of: (1) comparatively low addi-tional inputs by occult precipitation under forest com-pared to rough pasture that contained low scrub and ferns,

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 24: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

under conditions of very high rainfall (including sub-stantial amounts of WDR) and (2) the lower water useof pasture compared to cloud forest (Bruijnzeel, 2006;Schellekens, 2006). As such, it now seems likely that thestudy by Zadroga (1981) indeed underestimated areal pre-cipitation inputs, although not so much because of a lackof gauges in the uppermost and rainiest headwater areas(as postulated by Zadroga) but rather because of underes-timated amounts of inclined rainfall and ungauged WDR(Bruijnzeel, 2006; cf. Frumau et al., 2010b). Whilstrunoff response to rainfall (‘quickflow’) at the smallcatchment scale (<10 ha) was roughly doubled after con-version to grazed pasture—presumably due to dimin-ished infiltration opportunity and increased overlandflow on compacted cow trails (Tobon et al., 2010b;cf. Zimmermann and Elsenbeer, 2008)—the effect wasnot detectable at the operational catchment scale scale(100 km2;Bruijnzeel, 2006; Schellekens, 2006).

Whilst CWI, as opposed to WDR, constituted only amodest contribution to overall water inputs in the CostaRican case, the situation may be different in areas witha more seasonal rainfall regime where CWI may becomeimportant during the dry season—as for instance in partsof Pacific Central America (Holder, 2003; Mulligan andBurke, 2005a; cf. Figure 5b). Recent work on CWI andforest and pasture water use in central Veracruz, east-ern Mexico—an area with strongly seasonal rainfall andreportedly receiving substantial amounts of occult pre-cipitation (Vogelmann, 1973)—demonstrated that cloudforest conversion would lead to a major local increasein flows (Munoz-Villers, 2008; Gomez-Cardenas, 2009)amongst other reasons because of low CWI by the localLMCF (Holwerda et al., 2010c; cf. Ponette-Gonzalezet al., 2009) and a much higher water consumption bythe forest than by pasture (Holwerda et al., 2007; Gomez-Cardenas, 2009). As such, conversion of LMCF likelyleads to similar changes in water yield as described forLMRF below the main cloud belt (Blackie, 1979a,b;Edwards, 1979). Finally, the partial clearing of Douglasfir forest in the Pacific Northwest of the USA—an areasubject to heavy fog incidence (Harr, 1982)—caused atemporary decrease in baseflows during the dry summermonths. Interestingly, the effect gradually disappearedafter 5–6 years, as the regenerating trees were apparentlyeffective at capturing sufficient amounts of cloud wateragain (Ingwersen, 1985).

Reliable information on the effect of cloud forest con-version to agricultural cropping (with or without asso-ciated soil degradation and ensuing negative effects ondry-season flows; cf. Bruijnzeel, 2004) or conversion toshade-coffee plantations is not available, although theeffect on water yield could be modelled from infor-mation on coffee plantation water use (Holwerda et al.,2007) and corresponding amounts of rainfall- and CWI(Ponette-Gonzalez et al., 2009).

Summarizing, the few available data suggest that con-version of (very) wet, exposed MCF to aerodynamicallyrough pasture does not produce significant changes in

water yield but major increases in flows may be expectedafter converting LMCF receiving small occult inputs.It should also be noted that areas where fog consti-tutes a significant hydrological input, tend to be spatiallyrestricted (Figures 5–7). As such, inputs by CWI aresoon ‘diluted’ by rainfall inputs as one moves down-stream, and fog water contributions to the flows carriedby most rivers in the lowlands are mostly insignificant(cf. Brown et al., 1996; Mulligan and Burke, 2005a).

HYDROLOGICAL PROCESSES IN CLOUDFORESTS AND CLIMATE CHANGE

Observed and projected climate change in cloud forestareas

During the ‘anthropocene’ (Crutzen and Stoermer,2000) most of Earth’s ecosystems—cloud forests in-cluded—have become subject to sustained human impact.This raises a number of important questions, both withregard to the impacts of climate change on cloud for-est biodiversity (discussed recently in some detail byBruijnzeel et al., 2010b) and on the hydrological servicesthat these forests provide.

On the basis of preliminary modelling for selectedcloud forest sites, Still et al. (1999) and Foster (2001)concluded that climate change resulting from anthro-pogenic greenhouse gas emissions will significantly affectthe temperature and humidity of cloud forest environ-ments world-wide. Similarly, increasing observational(e.g. Pounds et al., 1999; Barradas et al., 2010) andexperimental evidence suggests cloud forests may warmup locally as a result of deforestation in adjacent low-lands (Lawton et al., 2001; Ray et al., 2006; Nair et al.,2010; Van der Molen et al., 2010). Scatena (1998) inter-preted the presence of isolated stands of large and veryold (>600 years) Colorado trees (Cyrilla racemiflora) inthe Luquillo Mountains of eastern Puerto Rico at ele-vations well below the current cloud base as evidenceof a gradual upward shift in vegetation zonation overthe past several centuries. Cyrilla is currently a domi-nant tree in areas above the cloud base (>600 m.a.s.l.)and is most common where mean annual rainfall exceeds4000 mm. Similarly, Brown et al. (1996) reported theoccurrence of pockets of mossy cloud forest below thecurrent average cloud base in Honduras. According tothe temperature lapse rates prevailing on most tropicalmountains (0Ð5–0Ð65 °C per 100 m rise in elevation), aprojected rise in temperature of 3 °C would imply a rise incloud-base level, the magnitude of which would dependon concurrent changes in atmospheric humidity (cf. Stillet al., 1999).

Since cloud forest environments are defined by thepresence of ground-level cloud (i.e. fog), the rises incloud-base levels implied by regional warming and dry-ing (Lawton et al., 2001; Ray et al., 2006) can beexpected to have profound impacts upon many aspectsof the cloud forest ecosystem, ranging from changesin biological diversity (Pounds et al., 1999; Nadkarni

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Sticky Note
Page 24, end of first para: merge Bruijnzeel tc. with previous sentence (i.e. delete Return).
Page 25: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

and Solano, 2002; Williams et al., 2003; Colwell et al.,2008) to changes in productivity (Raich et al., 1997;Aiba et al., 2010) and, quite possibly, hydrological func-tioning (Hemp, 2005a; Ponette-Gonzalez et al., 2009).Generally speaking, because of the fragmented nature ofTMCF distributions and their occurrence along narrowelevational bands in heterogeneous mountain landscapes(Vazquez-Garcıa, 1995; cf. Figure 1), cloud forests mustbe considered highly vulnerable to local and regional cli-mate changes (Mulligan, 2010). This is true both in termsof the very real possibility of ‘mountain-top extinctions’(where summit ecotones are driven out of existence uponlifting of the cloud base), and in terms of spatial disconti-nuities between current and projected elevational rangesfor groups of species (Williams et al., 2003; Colwellet al., 2008). In addition, due to their often exposed loca-tions (Jarvis and Mulligan, 2010), cloud forests in areasaffected by hurricane passage (notably the Caribbean,Central America, east Asia, northern Queensland, andthe south-western Pacific; cf. Scatena et al., 2005) maybecome subject to increasingly frequent or intense dis-turbance by tropical cyclones due to the warming of theoceans, although the topic is still a matter of much debate(e.g. Emanuel, 2005; Webster et al., 2005; McBride et al.,2006; cf. Mann et al., 2007).

Because of the lack of precise information on thelocation of different types of cloud forests and the scale ofavailable models, there has been little research to date thathas looked at the potential impacts of climate change oncloud forests as a whole and which areas are more likelyto be impacted than others (cf. Still et al., 1999; Foster,2001; Williams et al., 2007). Mulligan and Burke (2005b)used the modelled distributions of cloud-affected forest(CAF) shown in Figure 2 to analyze both historicallyobserved and future projected climate change in areaswith CAF (cf. Mulligan, 2010) and their hydrologicalimpacts. Mulligan and Burke first used the climatic dataset of New et al. (2000) to examine whether a changewas detectable between the mean temperatures in areaswith CAF for the periods 1901–1946 and 1947–1993.The observed change was a cooling of around 1 °C overall CAF except those in the southern Andes, in contrast tothe warming observed for tropical lowlands over the sameperiod. This cooling may be insignificant or could be theresult of reduced nocturnal cloud cover leading to lowernight-time temperatures. The pattern for historical rainfallchange in areas with CAF as derived by Mulligan andBurke (2005b) was less clear, with some areas indicatingstationarity, others drying and yet others wetting. Inalmost all cases, the observed changes were modest andwithin 20% of the 1901–1993 mean rainfall.

Cloud frequency is perhaps the most significant climatevariable for CAF. To assess changes in cloud cover,Mulligan and Burke (2005b) used the HIRS cloudclimatology data set of the University of Wisconsin (cf.Jin et al., 1996)—which at that time spanned a periodof 22 years (1979–2001). They extracted annual cloudfrequency data for all areas with known cloud forestaccording to the UNEP–WCMC database (Aldrich et al.,

1997) and calculated the change in the frequency ofclear-sky (i.e. no cloud) observations between the periods1979–1990 and 1991–2001. The results indicated adecrease in the number of clear days (i.e. an increasein cloudiness) for cloud forests near the equator, and thereverse (decreased cloudiness) for the 10–20° latitudinalbelts to the north and south. Observed cloud frequencydeclined particularly over cloud-forest areas in south-eastAsia, whereas in Latin America there was a decline in thenorth and an increase in cloud frequency in the south.The maximum observed changes in cloud-free days werearound 4% per decade.

Having found mixed signals of climate change inthe limited historical data set (i.e. a modest tem-perature decline, and small-scale positive or negativechanges in rainfall and cloud frequency), Mulligan andBurke (2005b) examined the results of two differ-ent general circulation models (GCMs)—ECHAM5 andHADCM3—using the SRES A2 CO2 scenarios preparedfor the Intergovernmental Panel on Climate Change ThirdAssessment Report published in 2001. The monthly grid-ded GCM data were split into two periods, 1950–2000and 2000–2050, and mean values for temperature andrainfall were calculated for the two periods and thencompared (cf. Mulligan, 2010). HADCM3 predicted amean global temperature increase of 3Ð3 °C between thetwo periods, with CAF seeing changes between C2 andC5 °C. The greatest temperature changes were appar-ent for the most ‘continental’ cloud forests (located inthe African interior), and the smallest changes were forcloud forests on small islands (Figure 13a). A similarspatial pattern was predicted by ECHAM5 although themagnitude of the changes was somewhat less than thoseobtained with HADCM3 (Figure 13b). Given the typi-cal elevational range of ¾500 m (equivalent to a rise intemperature of 3 °C) observed for many tropical mon-tane tree species (Kappelle et al., 1995; Kitayama, 1995;Bach, 2004) and the often narrower ranges found for ahost of animals (cf. Williams et al., 2003; Colwell et al.,2008; Rovito et al., 2009), these changes in temperatureare bound to have major ecological consequences (seeBruijnzeel et al. (2010b) for a fuller discussion). Theresults for changes in rainfall as projected by the twoGCMs were both more extreme and more uncertain thanthose for temperature, with large differences between thetwo GCMs at the regional and sub-regional scale (seeMulligan and Burke (2005b)). A subsequent (prelimi-nary) ensemble run of 21 GCMs (Mulligan and Rubiano,2009) suggested the results obtained with ECHAM5(Figure 14a) to be somewhat more representative thanthose obtained with HADCM3 (not shown) although fur-ther work remains necessary.

Impacts of climate change on TMCF water balanceand hydrological processes

Mulligan and Burke (2005b) also developed a simplemodel to illustrate spatial and temporal patterns in waterbalance across the tropics (23Ð5 °N–35 °S). The modelhas a spatial resolution of 1 km and operates at a

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Cross-Out
brul
Inserted Text
1
Page 26: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Figure 13. Temperature changes (1950–2000 vs 2000–2050) in°C in the vicinity of cloud forests as projected by (a) the HADCM3- and (b) theECHAM5 GCMs

Figure 14. Modelled impact of projected climate change (2000–2050) on (a) annual rainfall (mm) and (b) the water balance (mm year�1) ofcloud-affected forest areas across the tropics as projected by the ECHAM5 GCM

monthly time step. Grid-cell water balance is calculatedas precipitation (including modelled contributions bywind-driven rain but not cloud water) minus potentialevapotranspiration (PET). Using grids of current climatedata, baseline conditions were simulated [see map in thestudy of Mulligan and Burke (2005b), p. 59] to allowpreliminary evaluation of the spatial changes in monthlywater budget as predicted by the model after applying thechanges in temperature (Figure 13b), solar radiation and

rainfall (Figure 14a) projected by the ECHAM5 GCMfor the period 2000–2050 (Figure 14b). The predictedimpact of projected climate change on the water balanceof cloud-affected forests is variable and dominated bythe projected changes in rainfall. A projected increasedprecipitation surplus occurs in some areas (e.g. the Andesand Central Africa) and significant drying in other areas(West Africa, Madagascar, south-east Asia, the BrazilianAtlantic cloud forests and the CAFs of the Guyana shield;

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 27: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Figure 14b). Naturally, these results should be viewedas highly preliminary given the scale of analysis andthe simplicity of the model used. Furthermore, becauserainfall dominates the water balance signal, the simulatedimpacts are highly dependent on the rainfall projectionsof the GCM used. As stated earlier, these tend to differbetween GCMs, both in terms of direction and magnitude.Moreover, the simulations did not take into account theeffect of seasonal changes in rates of glacial meltingat higher elevations (cf. Kaser et al., 2005; Juen et al.,2007) nor the complex relationships between changesin lowland energy budget and upland cloud-base height(cf. Ray et al., 2006; Van der Molen et al., 2006, 2010;Nair et al., 2010). Like CWI, all of these effects wereconsidered to have a small impact at the pan-tropicalscale and Mulligan and Burke (2005b) showed the GCM-based rainfall change scenario to have a greater impacton the overall water budget than that simulated for allcumulative cloud forest conversion until 2001.

Apart from any changes in water budget and vegeta-tion water use, climate change is also bound to affecthydrological processes that are specific to cloud forestssuch as CWI. Both CWI and rainfall interception canbe expected to be affected, not only in a direct manner(through changes in fog frequency or density, and in rain-fall amounts and intensity) but also indirectly (throughchanges in the biomass and composition of canopy epi-phytes and bryophytes). Nadkarni and Solano (2002)have documented the major changes in epiphyte andbryophyte diversity that may occur upon climatic dryingor opening up of cloud forest canopies (cf. Foster, 2001).Indeed, given their exposed position and dependence onrainfall and fog inputs and the nutrients contained therein(Hietz, 2010), epiphytes and bryophytes in the upper partsof the canopy must be regarded as being particularlyvulnerable to climatic drying (Benzing, 1998; Zotz andBader, 2009; cf. Lugo and Scatena, 1992). As such, theirdemise or disappearance is likely to change the canopy’scloud- and rainfall intercepting capacity in ways that areas yet not fully understood (cf. Chang et al., 2002; Mul-ligan et al., 2010; Tobon et al., 2010a).

CONCLUDING REMARKS

With human populations and pressure on the land on theincrease in many tropical montane areas, the wet, windyand steep environments characterizing mossy forestsoften no longer provide the de facto protection theydid historically (Mosandl et al., 2008). Climate changeaside, cloud forests continue to be seriously at risk andmay even disappear completely during this century ifno major conservation action is taken in the comingdecades (Bubb et al., 2004; Mulligan, 2010). The manyunderlying issues such as poverty, insecurity of landownership, failed forest policies and the lack of effectivelaw enforcement continue to challenge our globalizedsociety. As a result, montane forest clearing, acceleratedsoil erosion and landslides, as well as disruptions of the

hydrological regime and, ultimately, species extinctionwill continue to take place.

Fortunately, knowledge of the geographical occur-rence, and awareness of the importance and valueof cloud forests has increased considerably in recentdecades, both amongst people living in and near cloudforests and at national and global levels. There is now abetter understanding of the importance of cloud forestsin terms of the various environmental goods and servicesthey provide, such as biodiversity, recreation and tourism,protection against soil erosion and perhaps most impor-tantly, the stable streamflow regimes and high-qualitywater that are indispensable for agricultural crop irriga-tion, hydro-electric power generation and drinking waterfor downstream populations.

In addition, during the last 50 years an increasing num-ber of tropical montane cloud forest sites have receivedsome kind of protected area status, ranging from multiple-use protected landscapes and forest reserves, to restricted-use national parks and absolute reserves. However, ifhumanity is to preserve a large part of cloud forest diver-sity, a conservation strategy will need to be elaboratedthat goes beyond the current networks of protected coreareas, buffer zones and corridors. This will require a con-certed, global effort with focused local action, in whichgovernment agencies, local authorities, scientists, NGOs,private enterprises and civil society all work together, todevelop and implement strategies that halt the threats ofinvasive species, deforestation and fragmentation, and soensure that cloud forests can continue to serve both natureand mankind in the long run. An essential element of suchan effort is the quantification and valuation of the overallset of environmental goods and services offered by cloudforests to local, regional and even global communitieswhich will make their preservation and wise managementan economic necessity, socio-politically feasible, cultur-ally rewarding, and ecologically balanced. The costs ofcloud forest protection can then be considered relative tothe benefits received from them.

Hydrological investigations in cloud forest environ-ments have proliferated in recent years and much moreis currently known about their hydrological functioningcompared to even a decade ago (cf. Bruijnzeel and Proc-tor, 1995; Cavelier, 1996; Bruijnzeel, 2001). However,whilst the current collection of papers shows significantprogress is being made with the quantification of rain-fall and occult inputs in cloud forest environments, morework is needed to better understand the respective con-tributions by wind-driven rain and CWI, the amounts ofwater lost to evaporation from UMCF and ECF, as wellas the impacts of land-use and climatic change on variouskey hydrological processes and downstream water yields.

ACKNOWLEDGEMENT

The authors gratefully acknowledge permission fromCambridge University Press to use material from thesummary chapter by Bruijnzeel et al. (2010b) of which

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 28: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

this paper is an extension. Thanks are also due toall colleagues and friends who have provided data,insights and encouragement, notably Arnoud Frumau,Tom Giambelluca and Friso Holwerda. Jim Juvik, SteveHowell and Nobuaki Takanaka provided unpublishedwind data for the Pu’u La’Au, Gambubal and Kog Masites, respectively, that have been included in Figure 8.Maarten Waterloo provided critical assistance with theproduction of Figures 8 and 9. The paper benefitedgreatly from the perceptive comments of Friso Holwerdaand Tom Giambelluca.

REFERENCES

Aiba S, Takyu M, Kitayama K. 2010. Biennial variation in tree diametergrowth during eight years in montane cloud forests on Mount Kinabalu,Sabah, Malaysia. In Tropical Montane Cloud Forests. Sciencefor Conservation and Management , Bruijnzeel LA, Scatena FN,Hamilton LS (eds). Cambridge University Press: Cambridge, UK;579–583.

Aldrich M, Billington C, Edwards M, Laidlaw R. 1997. A Global Direc-tory of Tropical Montane Cloud Forests , UNEP-World ConservationMonitoring Centre: Cambridge, UK.

Araujo AC. 2009. Spatial variation of CO2 fluxes and lateral transportin an area of terra firme forest in central Amazonia. PhD Dissertation,VU University Amsterdam, Amsterdam, The Netherlands.

Aravena R, Suzuki O, Pollastri A. 1989. Coastal fog and its relation toground-water in the IV region of northern Chile. Chemical Geology79: 83–91.

Arazi A, Sharon D, Khain A, Huss A, Mahrer Y. 1997. The windfieldand rainfall distribution induced within a small valley: field observa-tions and 2-D numerical modelling. Boundary-Layer Meteorology 83:349–374.

Asbjornsen H, Garnica-Sanchez Z. 2010. Fire dynamics and communitymanagement of fire in montane cloud forests in south-easternMexico. In Tropical Montane Cloud Forests. Science for Conservationand Management , Bruijnzeel LA, Scatena FN, Hamilton LS (eds).Cambridge University Press: Cambridge, UK; 659–670.

Ashton PS. 2003. Floristic zonation of tree communities on wet tropicalmountains revisited. Perspectives in Plant Ecology, Evolution andSystematics 6: 87–104.

Asquith N, Wunder S (eds). 2008. Payments for Watershed Services:The Bellagio Conversations . Fundacion Natura Bolivia: Santa Cruz,Bolivia.

Ataroff M. 1998. Importance of cloud water in Venezuelan Andean cloudforest water dynamics. In First International Conference on Fog andFog Collection, Schemenauer RS, Bridgman HA (eds). InternationalDevelopment Research Center: Ottawa, Canada; 25–28.

Ataroff M, Rada F. 2000. Deforestation impact on water dynamics in aVenezuelan Andean cloud forest. Ambio 29: 440–444.

Bach K. 2004. Vegetationskundliche Untersuchungen zur Hohenzon-ierung tropischer Bergregenwalder in den Anden Boliviens. PhDDissertation, University of Gottingen, Gottingen, Germany.

Bailly C, Benoit de Cognac G, Malvos C, Ningre JM, Sarrailh JM. 1974.Etude de l’influence du couvert naturel et de ses modificationsa Madagascar. Experimentations en bassins versants elementaires.Cahiers Scientifiques du Centre Technique Forestier Tropical 4: 1–114.

Barradas VL, Cervantes-Perez J, Ramos-Palacios R, Puchet-Anyul C,Vazquez-Rodriguez P, Granados-Ramirez R. 2010. Meso-scale climatechange in the central mountain region of Veracruz State, Mexico.In Tropical Montane Cloud Forests. Science for Conservationand Management , Bruijnzeel LA, Scatena FN, Hamilton LS (eds).Cambridge University Press: Cambridge, UK; 549–556.

Beiderwieden E, Wolff V, Hsia YJ, Klemm O. 2008. It goes both ways:measurements of simultaneous evapotranspiration and fog dropletdeposition at a montane cloud forest. Hydrological Processes 22:4181–4189.

Bendix J, Rollenbeck R, Richter M, Fabian P, Emck P. 2008. Climate. InGradients in a Tropical Mountain Ecosystem of Ecuador , EcologicalStudies , vol. 198, Beck E, Bendix J, Kottke I, Makeschin F,Mosandl R (eds). Springer Verlag: New York; 63–74.

Benner J, Vitousek PM, Ostertag R. 2010. Nutrient cycling andnutrient limitation in tropical montane cloud forests. In Tropical

Montane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 90–100.

Benzing DH. 1998. Vulnerabilities of tropical forests to climate change:the significance of resident epiphytes. Climatic Change 39: 519–540.

Biggs A, Cereceda P, (eds). 2007. Proceedings of the Fourth InternationalConference on Fog and Fog Collection and Dew. PontificiaUniversidad Catolica de Chile, La Serena (Chile), 22–27 July 2007.

Blackie JR. 1979a. The water balance of the Kericho catchments. EastAfrican Agricultural and Forestry Journal 43: 55–84.

Blackie JR. 1979b. The water balance of the Kimakia catchments. EastAfrican Agricultural and Forestry Journal 43: 155–174.

Blanchard DC. 1953. Raindrop size-distribution in Hawaiian rains.Journal of Meteorology 10: 457–473.

Blocken B, Carmeliet J, Poesen J. 2005. Numerical simulation of thewind-driven rainfall distribution over small-scale topography in spaceand time. Journal of Hydrology 315: 252–273.

Blocken B, Poesen J, Carmeliet J. 2006. Impact of wind on the spatialdistribution of rain over micro-scale topography: numerical modelingand experimental verification. Hydrological Processes 20: 345–368.

Brown MB, de la Roca I, Vallejo A, Ford G, Casey J, Aguilar B,Haacker R. 1996. A Valuation Analysis of the Role of Cloud Forestsin Watershed Protection. Sierra de las Minas Biosphere Reserve,Guatemala and Cusuco N.P. Honduras , RARE Center for TropicalConservation: Philadelphia, USA.

Bruijnzeel LA. 2001. Hydrology of tropical montane cloud forests: areassessment. Land Use and Water Resources Research 1: 1Ð1–1Ð18.

Bruijnzeel LA. 2004. Hydrological functions of tropical forests: notseeing the soil for the trees? Agriculture, Ecosystems and Environment104: 185–228.

Bruijnzeel LA. 2005. Tropical montane cloud forest: a uniquehydrological case. In Forests, Water and People in the HumidTropics , Bonell M, Bruijnzeel LA (eds). Cambridge University Press:Cambridge, UK; 462–483.

Bruijnzeel LA, (ed). 2006. Hydrological impacts of converting tropicalmontane cloud forest to pasture, with initial reference to northernCosta Rica. Final Technical Report for Project R7991, DFID ForestryResearch Programme. VU University Amsterdam: Amsterdam, TheNetherlands.www.ambiotek.com/fiesta. [accessed: September 2010].

Bruijnzeel LA, Hamilton LS. 2000. Decision Time for Cloud Forests ,IHP Humid Tropics Programme Series , vol. 13, UNESCO Divisionof Water Sciences: Paris, France (available at: http://sea.unep-wcmc.org/forest/cloudforest/index.cfm). [accessed: September 2010].

Bruijnzeel LA, Proctor J. 1995. Hydrology and biochemistry of tropicalmontane cloud forests: what do we really know? In Tropical MontaneCloud Forests , Ecological Studies , vol. 110, Hamilton LS, Juvik JO,Scatena FN (eds). Springer Verlag: New York; 38–78.

Bruijnzeel LA, Waterloo MJ, Proctor J, Kuiters AT, Kotterink B. 1993.Hydrological observations in montane rain-forests on Gunung-Silam,Sabah, Malaysia, with special reference to the ‘Massenerhebung effect.Journal of Ecology 81: 145–167.

Bruijnzeel LA, Eugster W, Burkard R. 2005. Fog as an input tothe hydrological cycle. In Encyclopaedia of Hydrological Sciences ,Anderson MG, McDonnell JJ (eds). John Wiley and Sons: Chichester,UK; 559–582.

Bruijnzeel LA, Scatena FN, Hamilton LS (eds). 2010a. Tropical MontaneCloud Forests. Science for Conservation and Management . CambridgeUniversity Press: Cambridge, UK.

Bruijnzeel LA, Kappelle M, Mulligan M, Scatena FN. 2010b. Tropicalmontane cloud forests: state of knowledge and sustainabilityperspectives in a changing world. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 691–740.

Bubb P, May I, Miles L, Sayer J. 2004. Cloud Forest Agenda, UNEP-World Conservation Monitoring Centre: Cambridge, UK (availableat: http://sea.unep-wcmc.org/forest/cloudforest/index.cfm). [accessedin October 2010].

Burgess SSO, Dawson TE. 2004. The contribution of fog to the waterrelations of Sequoia sempervirens D. Don: Foliar uptake and preventionof dehydration. Plant, Cell and Environment 27: 1023–1034.

Caceres G. 1981. Importancia hidrologica de la intercepcion horizontalen un bosque muy humedo premontano en Balalaica, Turrialba, CostaRica. MSc Thesis, University of Costa Rica, Turrialba, Costa Rica.

Calamini G, Giacomin A, Falciai M, Salbitano F, Villasante F. 1998. Foginterception and water budget of Caesalpina spinosa trees in the lomasecosystems of Mejia (Arequipa, Peru). In First International Con-ference on Fog and Fog Collection , Schemenauer RS, Bridgman HA(eds). International Development Research Center: Ottawa; 473–476.

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 29: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Calvo JC. 1986. An evaluation of Thornthwaite’s water balance techniquein predicting stream runoff in Costa Rica. Hydrological SciencesJournal 31: 51–60.

Campanella R. 1995. The role of GIS in evaluating contour-basedlimits of cloud forest reserves in Honduras. In Tropical MontaneCloud Forests , Ecological Studies , vol. 110, Hamilton LS, Juvik JO,Scatena FN (eds). Springer Verlag: New York; 116–124.

Cao G, Giambelluca TW, Stevens D, Schroeder T. 2007. Inversionvariability in the Hawaiian trade wind regime. Journal of Climate 20:1145–1160.

Cavelier J. 1996. Environmental factors and ecophysiological processesalong altitudinal gradients in wet tropical mountains. In TropicalForest Plant Ecophysiology , Mulkey SS, Chazdon RL, Smith AP(eds). Chapman and Hall: New York; 399–439.

Cavelier J, Jaramillo M, Solis D, De Leon D. 1997. Water balance andnutrient inputs in bulk precipitation in tropical montane cloud forest inPanama. Journal of Hydrology 193: 83–96.

Cereceda P, Schemenauer RS. 1991. The occurrence of fog in Chile.Journal of Applied Meteorology 30: 1097–1105.

Chang M, Flannery LA. 2001. Spherical gauges for improving theaccuracy of precipitation measurements. Hydrological Processes 15:643–654.

Chang M, Harrison L. 2005. Field assessments on the accuracy ofspherical gauges in rainfall measurements. Hydrological Processes 19:403–412.

Chang S-C, Lai IL, Wu J. 2002. Estimation of fog on epiphyticbryophytes in a subtropical montane forest ecosystem in north-easternTaiwan. Atmospheric Research 64: 159–167.

Clark KL, Nadkarni NM, Schaefer D, Gholz HL. 1998. Atmosphericdeposition and net retention of ions by the canopy in a tropicalmontane forest, Monteverde, Costa Rica. Journal of Tropical Ecology14: 27–45.

Climatology Working Group. 2010. Fifth International Conference on Fogand Fog Collection and Dew, 25–30 July 2010, Muenster, Germany(available at: http://meetings.copernicus.org/fog2010/fogconference2010 conference book.pdf).

Colwell RK, Brehm G, Cardelus CL, Gilman AC, Longino JT. 2008.Global warming, elevational range shifts, and lowland biotic attritionin the wet Tropics. Science 322: 258–261.

Crutzen PJ, Stoermer EF. 2000. “The Anthropocene”. IGBP Newsletter41: 17–18.

Dawson TE. 1998. Fog in the California redwood forest: ecosystem inputsand use by plants. Oecologia 117: 476–485.

DeLay JK, Giambelluca TW. 2010. History of fog and cloud waterinterception research in Hawai‘i. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 332–341.

Dietz J, Hoelscher D, Leuschner Ch, Hendrayanto . 2006. Rainfallpartitioning in relation to forest structure in differently managedmontane forest stands in Central Sulawesi, Indonesia. Forest Ecologyand Management 237: 170–178.

Doumenge C, Gilmour DA, Ruiz-Perez M, Blockhus J. 1995. Tropicalmontane cloud forests: conservation status and management issues.In Tropical Montane Cloud Forests , Ecological Studies , vol. 110,Hamilton LS, Juvik JO, Scatena FN (eds). Springer Verlag: New York;24–37.

Edwards KA. 1979. The water balance of the Mbeya experimentalcatchments. East African Agricultural and Forestry Journal 43:231–247.

Edwards PJ. 1982. Studies of mineral cycling in a montane rain forest inNew Guinea. V. Rates of cycling in throughfall and litter fall. Journalof Ecology 70: 807–827.

Emanuel KA. 2005. Increasing destructiveness of tropical cyclones overthe past 30 years. Nature 436: 686–688.

Eugster W, Burkard R, Holwerda F, Scatena FN, Bruijnzeel LA. 2006.Characteristics of fog and fog-water fluxes in a Puerto Rican elfincloud forest. Agricultural and Forest Meteorology 139: 288–306.

Fallas J. 2002. Net precipitation patterns in undisturbed and fragmentedCosta Rican cloud forests. In Proceedings of the second internationalcolloquium on hydrology and water management , ed. Gladwell JS.CATHALAC: Panama City, Panama; 389–398.

Farr TG, Kobrick M. 2000. Shuttle Radar Topography Mission producesa wealth of data. Eos, Transactions of the American Geophysical Union81: 583–585.

Fleischbein K, Wilcke W, Goller R, Boy J, Valarezo C, Zech W,Knoblich K. 2005. Rainfall interception in a lower montane forestin Ecuador: effects of canopy properties. Hydrological Processes 19:1355–1371.

Fleischbein K, Wilcke W, Valarezo C, Zech W, Knoblich K. 2006. Waterbudgets of three small catchments under montane forest in Ecuador:experimental and modelling approach. Hydrological Processes 20:2491–2507.

Førland EJ, Allerup P, Dahlstrom B, Elomaa E, Jonsson T, Madsen H,Perala J, Rissanen P, Vedin H, Vejen F. 1996. Manual for operationalcorrection of Nordic precipitation data, Norwegian MeteorologicalInstitute: Oslo, Norway.

Foster P. 2001. The potential negative impacts of global climate changeon tropical montane cloud forests. Earth-Science Reviews 55: 73–106.

Foster P. 2010. Changes in mist immersion. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 57–66.

Frahm JP, Gradstein SR. 1991. An altitudinal zonation of tropical rainforests using bryophytes. Journal of Biogeography 18: 669–676.

Frumau KFA, Burkard R, Schmid S, Bruijnzeel LA, Tobon C, Calvo-Alvarado JC. 2010a. A comparison of the performance of three typesof passive fog gauges under conditions of wind-driven fog andprecipitation. Hydrological Processes . DOI:10.1002/hyp.7884.

Gabriel G, Jauze L. 2008. Fog water interception by Sophora denudatatrees in a Reunion upper montane forest, Indian Ocean. AtmosphericResearch 87: 338–351.

Frumau KFA, Bruijnzeel LA, Tobon C. 2010b. Precipitation measure-ment and derivation of precipitation inclination in a windy moun-tainous area in northern Costa Rica. Hydrological Processes . DOI:10.1002/hyp.7860.

Garcıa-Santos G. 2007. An ecohydrological and soils study in a mon-tane cloud forest in the National Park of Garajonay, La Gomera(Canary Islands, Spain). PhD Dissertation, VU University Amsterdam,Amsterdam, The Netherlands. http://www.falw.vu.nl/nl/onderzoek/earth-sciences/geo-environmental-science-and-hydrology/hydrology-dissertations/index.asp. [accessed: July 2010].

Garcıa-Santos G, Bruijnzeel LA. 2010. Rainfall, fog and throughfalldynamics in a sub-tropical ridge-top cloud forest, National Parkof Garajonay (La Gomera, Canary Islands, Spain). HydrologicalProcesses . DOI: 10.1002/hyp.7760.

Garriguata MR, Balvanera P. 2009. Tropical forest service flows. ForestEcology and Management . 258: 1825–1829.

Gerold G, Schawe M, Bach K. 2008. Hydrometeorologic, pedologic andvegetation patterns along an elevational transect in the montane forestof the Bolivian Yungas. Die Erde 139: 141–168.

Giambelluca TW, Gerold G. 2011. Hydrology and biogeochemistry oftropical montane cloud forests. In Hydrology and Biogeochemistryof Forest Ecosystems , Levia DF, Carlyle-Moses D, Tanaka T (eds).Springer Verlag: New York; in press.

Giambelluca TW, Martin RE, Asner GP, Huang M, Mudd RG, Nul-let DA, DeLay DK, Fote D. 2009. Evapotranspiration and energy bal-ance of native wet montane cloud forest in Hawai’i. Agricultural andForest Meteorology 149: 230–243.

Giambelluca TW, DeLay JK, Nullet MA, Scholl MA, Gingerich SB.2010a. Interpreting canopy water balance and fog screen observations:Separating cloud water from wind-blown rainfall at two contrastingforest sites in Hawai‘i. In Tropical Montane Cloud Forests. Sciencefor Conservation and Management , Bruijnzeel LA, Scatena FN,Hamilton LS (eds). Cambridge University Press: Cambridge, UK;342–351.

Giambelluca TW, DeLay JK, Nullet MA, Scholl MA, Gingerich SB.2010b. Canopy water balance of windward and leeward Hawaiiancloud forests on Haleakala, Maui, Hawai‘i. Hydrological Processes .DOI: 10.1002/hyp.7738.

Gomez-Cardenas M. 2009. Transpiration by contrasting vegetation covertypes in the montane cloud forest belt of eastern Mexico. PhDDissertation, Iowa State University, Ames, Iowa, USA.

Gomez-Peralta D, Oberbauer SF, McClain ME, Philippi TE. 2008.Rainfall and cloud-water interception in tropical montane forests inthe eastern Andes of Central Peru. Forest Ecology and Management255: 1315–1325.

Gonggrijp L. 1941. Evaporation from montane forest in West Java at analtitude of 1750–2000 m a.s.l. Tectona 34: 437–447 (in Dutch, witha summary in English).

Gradstein SR, Obregon A, Gehrig C, Bendix J. 2010. Tropical lowlandcloud forest: a neglected forest type. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 130–133.

Grubb PJ. 1977. Control of forest growth and distribution on wet tropicalmountains: with special reference to mineral nutrition. Annual Reviewof Ecology and Systematics 8: 83–107.

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Inserted Text
o
Page 30: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Grubb PJ, Whitmore TC. 1966. A comparison of montane and lowlandrain forest in Ecuador. II. The climate and its effects on the distributionand physiognomy of the forests. Journal of Ecology 54: 303–333.

Hafkenscheid RLLJ, Bruijnzeel LA, de Jeu RAM, Bink NJ. 2002. Waterbudgets of two upper montane rain forests of contrasting stature in theBlue Mountains, Jamaica. In Proceedings of the Second InternationalColloquium on Hydrology and Water Management in the HumidTropics , Technical Documents in Hydrology , vol. 52, Gladwell JS (ed).IHP-UNESCO: Paris; 399–424.

Hager A, Dohrenbusch A. 2010. Hydrometeorology and structure oftropical montane cloud forests in north-western Costa Rica undercontrasting biophysical conditions. Hydrological Processes . DOI:10.1002/hyp.7726.

Hamilton LS, Juvik JO, Scatena FN (eds). 1995. In Tropical MontaneCloud Forests , Ecological Studies , vol. 110. Springer Verlag: NewYork.

Hansen M, DeFries R, Townshend JR, Carroll M, Dimiceli C,Sohberg R. 2006. Vegetation continuous fields MOD44B, 2001 per-cent tree cover, collection 4., University of Maryland: College Park,Maryland.

Harr RD. 1982. Fog-drip in the Bull Run municipal watershed, Oregon.Water Resources Bulletin 18: 785–789.

Hemp A. 2005a. Climate change driven forest fires marginalize theimpact of ice cap wasting on Kilimanjaro. Global Change Biology11: 1013–1023.

Hemp A. 2005b. Continuum or zonation? Altitudinal gradients in theforests on Mt. Kilimanjaro. Plant Ecology 184: 27–42.

Hemp A. 2010. Altitudinal zonation and diversity patterns in theforests of Mount Kilimanjaro, Tanzania. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 134–141.

Herrmann R. 1971. Die zeitliche Anderung der Wasserbindung im Bodenunter verschiedenen Vegetationsformationen der Hohenstufen einestropischen Hochgebirges (Sierra Nevada de Sta. Marta, Kolumbien).Erdkunde 25: 90–102.

Hietz P. 2010. Ecology and ecophysiology of epiphytes in tropicalmontane cloud forests. In Bruijnzeel LA, Scatena FN, Hamilton LS(eds). Tropical Montane Cloud Forests.. Science for Conservationand Management , Bruijnzeel LA, Scatena FN, Hamilton LS. (eds).Cambridge University Press: Cambridge, UK; 67–76.

Herwitz S, Slye RE. 1992. Spatial variation in the interception of inclinedrainfall by a tropical rainforest canopy. Selbyana 13: 62–71.

Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2004. TheWorldClim interpolated global terrestrial climate surfaces. Version 1.3.(available at http://biogeo.berkeley.edu/).

Holder CD. 2003. Fog precipitation in the Sierra de las Minas BiosphereReserve, Guatemala. Hydrological Processes 17: 2001–2010.

Holder CD. 2004. Rainfall interception and fog precipitation in a tropicalmontane cloud forest of Guatemala. Forest Ecology and Management190: 373–384.

Holscher D, Kohler L, Van Dijk AIJM, Bruijnzeel LA. 2004. Theimportance of epiphytes to total rainfall interception by a tropicalmontane rain forest in Costa Rica. Journal of Hydrology 292: 308–322.

Holwerda F. 2005. Water and energy budgets of rain forests alongan elevation gradient under maritime tropical conditions. PhD Dis-sertation, VU University Amsterdam, Amsterdam, The Netherlands.http://www.falw.vu.nl/nl/onderzoek/earth-sciences/geo-environmental-science-and-hydrology/hydrology-dissertations/index.asp. [accessed inJune 2010].

Holwerda F, Burkard R, Eugster W, Scatena FN, Meesters ACGA,Bruijnzeel LA. 2006. Estimating fog deposition at a Puerto Rican elfincloud forest site: comparison of the water budget and eddy covariancemethods. Hydrological Processes 20: 2669–2692.

Holwerda F, Barradas VL, Cervantes J, Bruijnzeel LA. 2007. Balanceshıdricos y de energıa de un cafetal de sombra en el centro de Veracruz,Mexico. Reporte tecnico final del proyecto INE/A1-064/2007 , Institutode Ecologıa: Xalapa, Veracruz, Mexico; VU University Amsterdam:Amsterdam, The Netherlands.

Holwerda F, Bruijnzeel LA, Scatena FN. 2010a. Comparison of passivefog gauges for determining fog duration and fog interception bya Puerto Rican elfin cloud forest. Hydrological Processes . DOI:10.1002/hyp.7641.

Holwerda F, Bruijnzeel LA, Oord AL, Scatena FN. 2010b. Foginterception in a Puerto Rican elfin cloud forest: a wet-canopywater budget approach. In Tropical Montane Cloud Forests. Sciencefor Conservation and Management , Bruijnzeel LA, Scatena FN,Hamilton LS (eds). Cambridge University Press: Cambridge, UK;282–292.

Holwerda F, Bruijnzeel LA, Munoz-Villers LE, Equihua M, Asb-jornsen H. 2010c. Rainfall and cloud water interception in mature andsecondary lower montane cloud forests of central Veracruz, Mexico.Journal of Hydrology 384: 84–96.

Holwerda F, Bruijnzeel LA, Scatena FN, Vugts HF, Meesters AGCA.2010d. Wet canopy evaporation from a Puerto Rican lower montanerain forest: the importance of erodynamic conductance. Poster paperpresented at the 2010 Fall Meeting of the American GeophysicalUnion, 13–17 December 2010, San Francisco.

Hutley LB, Doley D, Yates D, Boonsaner A. 1997. Water balance ofan Australian subtropical rainforest at altitude: the ecological andphysiological significance of intercepted cloud and fog. AustralianJournal of Botany 45: 311–329.

Ingwersen JB. 1985. Fog drip, water yield and timber harvesting in theBull Run Municipal Watershed, Oregon. Water Resources Bulletin 21:469–473.

Iremonger S, Ravilious C, Quinton T (eds). 1997. A global overview offorest conservation. CD-ROM. World Conservation Monitoring Centre:Cambridge, UK. http://www.unep-wcmc.org/forest/global map.htm.[accessed: June 2010].

Jarvis A, Mulligan M. 2010. The climate of tropical montane cloudforests. Hydrological Processes . DOI: 10.1002/hyp.7847.

Jetten VG. 1996. Interception of tropical rain forest: Performance of acanopy water balance model. Hydrological Processes 10: 671–685.

Jin Y, Rossow WB, Wylie DP. 1996. Comparison of the climatologiesof high-level clouds from HIRS and ISCCP. Journal of Climate 9:2850–2879.

Juen I, Georges C, Kaser G. 2007. Modelling observed and future runofffrom a glacierized tropical catchment (Cordillera Blanca, Peru). Globaland Planetary Change 59: 37–48.

Juvik JO, Ekern PC. 1978. A Climatology of Mountain Fog on MaunaLoa, Hawai‘i Island , Technical Report , vol. 118. Water ResourcesResearch Center, University of Hawai‘i: Honolulu, Hawai‘i.

Juvik JO, Nullet D. 1995a. Relationships between rainfall, cloud-waterinterception and canopy throughfall in a Hawaiian montane forest.In Tropical Montane Cloud Forests , Ecological Studies , vol 110,Hamilton LS, Juvik JO, Scatena FN (eds). Springer Verlag: New York;165–182.

Juvik JO, Nullet D. 1995b. Comments on ‘A proposed standard fogcollector for use in high elevation regions’. Journal of AppliedMeteorology 34: 2108–2110.

Juvik JO, DeLay JK, Kinney KM, Hansen E. 2010. A 50th Anniversaryreassessment of the seminal “Lana‘i fog drip study” in Hawai‘i.Hydrological Processes . DOI: 10.1002/hyp.7803.

Kapos V, Rhind J, Edwards M, Price MF. 2000. Developing a mapof the world’s mountain forests. In Forests in Sustainable Moun-tain Development: A State-of-Knowledge Report for 2000,, Price MF,Butt N (eds). CAB International: Wallingford, UK; 4–9. http://www.unep-wcmc.org/habitats/mountains/homepage.htm. [accessed: June2010].

Kappelle M. 1995. Ecology of mature and recovering Talamancan mon-tane Quercus forests, Costa Rica. PhD Dissertation, University ofAmsterdam, Amsterdam, The Netherlands.

Kappelle M, Brown AD (eds). 2001. Bosques Nublados del Neotropico.Editorial INBio: Santo Domingo de Heredia, Costa Rica.

Kappelle M, Van Uffelen JG, Cleef AM. 1995. Altitudinal zonation ofmontane Quercus forests along two transects in the Chirripo NationalPark, Costa Rica. Vegetatio (Plant Ecology) 119: 119–153.

Kaser G, Georges C, Juen I, Moelg T. 2005. Low latitude glaciers:unique global climate indicators and essential contributors to regionalfresh water supply. A conceptual approach. In Global Changeand Mountain Regions: A State of Knowledge Overview , Advancesin Global Change Research, vol 23, Huber U, Bugmann HKM,Reasoner MA (eds). Kluwer Publishers: New York; 185–196.

Kerfoot O. 1968. Mist precipitation on vegetation. Forestry Abstracts 29:8–20.

Kitayama K. 1995. Biophysical conditions of the montane cloud forests ofMount Kinabalu, Sabah, Malaysia. In Tropical Montane Cloud Forests ,Ecological Studies , vol. 110, Hamilton LS, Juvik JO, Scatena FN(eds). Springer Verlag: New York; 183–197.

Kitayama K, Muller-Dombois D. 1994a. An altitudinal transect analysisof the windward vegetation on Haleakala, a Hawaiian island mountain:(1) climate and soils. Phytocoenologia 24: 111–133.

Kitayama K, Muller-Dombois D. 1994b. An altitudinal transect analysisof the windward vegetation on Haleakala, a Hawaiian island mountain:(2) vegetation zonation. Phytocoenologia 24: 135–154.

Kohler L, Tobon C, Frumau KFA, Bruijnzeel LA. 2007. Biomass andwater storage dynamics of epiphytes in old-growth and secondary

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 31: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

montane cloud forest stands in Costa Rica. Plant Ecology 193:171–184.

Kohler L, Holscher D, Bruijnzeel LA, Leuschner C. 2010. Epiphytebiomass in Costa Rican old-growth and secondary montane rainforests and its hydrological significance. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 268–274.

Kumagai T, Saitoh TM, Sato Y, Takahashi H, Manfroi OJ, Morooka T,Kuraji K, Suzuki M, Yasunari T, Komatsu H. 2005. Annual waterbalance and seasonality of evapotranspiration in a Bornean tropicalrainforest. Agricultural and Forest Meteorology 128: 81–92.

Kumaran S. 2008. Hydrometeorology of tropical montane rain forestsof Gunung Brinchang, Pahang Darul Makmur, Malaysia. PhDDissertation, Universiti Putra Malaysia, Serdang, Malaysia.

LaBastille A, Pool DJ. 1978. On the need for a system of cloud-forest parks in Middle America and the Caribbean. EnvironmentalConservation 5: 183–190.

Laws JO, Parsons DA. 1943. The relation of raindrop size to intensity.Transactions of the American Geophysical Union 24: 452–460.

Lawton RO, Nair US, Pielke RA, Sr. Welch RM. 2001. Climatic impactof tropical lowland deforestation on nearby montane cloud forests.Science 294: 584–587.

Lawton RO, Nair US, Ray DK, Regmi A, Pounds A, Welch RM. 2010.Quantitative measures of immersion in cloud and the biogeographyof cloud forests. In Tropical Montane Cloud Forests. Sciencefor Conservation and Management , Bruijnzeel LA, Scatena FN,Hamilton LS (eds). Cambridge University Press: Cambridge, UK;217–227.

Letts MG, Mulligan M. 2005. The impact of light quality and leafwetness on photosynthesis in north-west Andean tropical montanecloud forest. Journal of Tropical Ecology 21: 549–557.

Letts MG, Mulligan M, Rincon-Romero ME, Bruijnzeel LA. 2010.Environmental controls on photosynthetic rates of lower montanecloud forest vegetation in south-western Colombia. In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 465–478.

Liu W, Fox JED, Xu Z. 2002. Nutrient fluxes in bulk precipitation,throughfall and stemflow in montane subtropical moist forest on AilaoMountain in Yunnan, south-west China. Journal of Tropical Ecology18: 527–541.

Liu WJ, Meng FR, Zhang YP, Li HM. 2004. Water input from fog dripin the tropical rain forest of Xishuangbanna, SW China. Journal ofTropical Ecology 20: 517–524.

Liu WJ, Liu WY, Li PJ, Gao L, Shen YX, Wang PY, Zhang YP, Li HM.2007. Using stable isotopes to determine sources of fog drip in atropical seasonal rain forest of Xishuangbanna, SW China. Agriculturaland Forest Meteorology 143: 80–91.

Lloyd CR, Marques A. de O. 1988. Spatial variability of throughfalland stemflow measurements in amazonian rainforest. Agricultural andForest Meteorology 42: 63–73.

Loope LL, Giambelluca TW. 1998. Vulnerability of island tropicalmontane cloud forests to climate change, with special reference toEast Maui, Hawaii. Climatic Change 39: 503–517.

Lugo AE, Scatena FN. 1992. Epiphytes and climate change in theCaribbean: a proposal. Selbyana 13: 123–130.

Lundgren L, Lundgren B. 1979. Rainfall, interception and evaporation inthe Mazumbai forest reserve, West Usambara Mts., Tanzania and theirimportance in the assessment of land potential. Geografiska Annaler61: 157–178.

MacQuarrie KIA, Pokhrel A, Shrestha Y, Osses P, Schemenauer RS.2001. Results from a high elevation fog water supply project inNepal. In Second International Conference on Fog and Fog Collection,Schemenauer RS, Puxbaum HA (eds). International DevelopmentResearch Center: Ottawa, Canada; 227–229.

Malkus JS. 1955. The effects of a large island upon the trade-wind airstream. Quarterly Journal of the Royal Meteorological Society 81:538–550.

Mamanteo BP, Veracion VP. 1985. Measurements of fog drip, throughfalland stemflow in the mossy and Benguet pine (Pinus kesiya Royle exGordon) forests in the upper Agno river basin. Sylvatrop (PhilippinesForestry Research Journal) 10: 271–282.

Mann ME, Emanuel KA, Holland GL, Webster PJ. 2007. Atlantictropical cyclones revisited. Eos, Transactions of the AmericanGeophysical Union 88(36): 349–350.

Manrique R, Ferrari C, Pezzi G. 2010. The influence of El NinoSouthern Oscillation (ENSO) on fog oases along the Peruvian and

Chilean coastal deserts. Fifth International Conference on Fog andFog Collection and Dew , Climatology Working Group, University ofMuenster: Muenster, Germany; 148–151.

Martin PH, Sherman RE, Fahey TJ. 2007. Tropical montane ecotones:climate gradients, natural disturbance, and vegetation zonation in theCordillera Central, Dominican Republic. Journal of Biogeography 34:1792–1806.

Marzol-Jaen MV, Sanchez-Megia J, Garcıa-Santos G. 2010. Effects offog on climatic conditions at a sub-tropical montane cloud forestsite in northern Tenerife (Canary Islands, Spain). In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 359–364.

McBride J, Kepert J, Chan J, Heming J, Holland GJ, Emanuel K,Knutson T, Willoughby H, Landsea C. 2006. Statement on tropicalcyclones and climate change. WMO/CAS Tropical MeteorologicalResearch Program, Steering Committee for Project TC-2, 4.

McJannet DL, Wallace JS, Reddell P. 2007a. Precipitation interception inAustralian tropical rainforests: I. Measurement of stemflow, throughfalland cloud interception. Hydrological Processes 21: 1692–1702.

McJannet DL, Wallace JS, Reddell P. 2007b. Precipitation interceptionin Australian tropical rainforests: II. Altitudinal gradients of cloudinterception, stemflow, throughfall and interception. HydrologicalProcesses 21: 1703–1718.

McJannet DL, Fitch PG, Disher MG, Wallace JS. 2007c. Measurementsof transpiration in four tropical rainforest types of north Queensland,Australia. Hydrological Processes 21: 3549–3564.

McJannet DL, Wallace JS, Fitch PG, Disher MG, Reddell P. 2007d.Water budgets of tropical rainforests in northern Queensland, Australia.Hydrological Processes 21: 3473–3483.

Meyer J-Y. 2010. Montane cloud forest on remote tropical islandsof Oceania: the example of French Polynesia. In TropicalMontane Cloud Forests. Science for Conservation and Management,,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 121–129.

Mildenberger K, Beiderwieden E, Hsia YJ, Klemm O. 2009. CO2 andwater vapor fluxes above a subrtropical mountain cloud forest—theeffect of light conditions and fog. Agricultural and Forest Meteorology .149: 1730–1736.

Mosandl R, Gunter S, Stimm B, Weber M. 2008. Ecuador suffers thehighest deforestation rate in South America. In Gradients in a TropicalMountain Ecosystem of Ecuador , Ecological Studies , vol. 198, Beck E,Bendix J, Kottke I, Makeschin F, Mosandl R (eds). Springer Verlag:New York; 37–40.

Motzer T. 2005. Micrometeorological aspects of a tropical mountainforest. Agricultural and Forest Meteorology 135: 230–240.

Motzer T, Munz N, Anhuf D, Kuppers M. 2010. Transpiration andmicro-climate of a tropical montane rain forest, southern Ecuador.In Tropical Montane Cloud Forests. Science for Conservationand Management , Bruijnzeel LA, Scatena FN, Hamilton LS (eds).Cambridge University Press: Cambridge, UK; 447–455.

Mulligan M. 2006a. Global gridded 1 km TRMM rainfall climatologyand derivatives. Version 1.0. http://www.ambiotek.com/1kmrainfall.[accessed: November 2010].

Mulligan M. 2006b. MODIS MOD35 pan-tropical cloud climatol-ogy. Version 1. September 2006. http://www.ambiotek.com/clouds.[accessed: November 2010].

Mulligan M. 2010. Modeling the tropics-wide extent and distribution ofcloud forest and cloud forest loss, with implications for conservationpriority. In Tropical Montane Cloud Forests. Science for Conservationand Management,, Bruijnzeel LA, Scatena FN, Hamilton LS (eds).Cambridge University Press: Cambridge, UK; 14–38.

Mulligan M, Burke SM. 2005a. FIESTA: Fog Interception for theEnhancement of Streamflow in Tropical Areas, Appendix 4ato Final Technical Report of DFID-FRP Project no. R7991.http://www.ambiotek.com/fiesta. [accessed: November 2010].

Mulligan M, Burke SM. 2005b. Global cloud forests and environmentalchange in a hydrological context. Final Report of DFID FRP ProjectZF0216. http://www.ambiotek.com/cloud forests. [accessed: November2010].

Mulligan M, Rubiano J. 2009. Climate change and its impact onagriculture. Paper presented at the 16th International Oil PalmConference, Cartagena, Colombia.

Mulligan M, Jarvis A, Gonzalez J, Bruijnzeel LA. 2010. Using ‘biosen-sors’ to elucidate rates and mechanisms of cloud water interception byepiphytes, leaves, and branches in a sheltered Colombian cloud forest.In Tropical Montane Cloud Forests. Science for Conservation and Man-agement , Bruijnzeel LA, Scatena FN, Hamilton LS (eds). CambridgeUniversity Press: Cambridge, UK; 249–260.

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

brul
Inserted Text
, Vitez F, Kowalchuk K, Taylor R.
Page 32: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

L. A. BRUIJNZEEL, M. MULLIGAN AND F. N. SCATENA

Munoz-Pina C, Guevara A, Torres JM, Brana J. 2008. Paying for thehydrological services of Mexico’s forests: Analysis, negotiations andresults. Ecological Economics 65: 725–736.

Munoz-Villers LE. 2008. Efecto del cambio en el uso de suelo sobrela dinamica hidrologica y calidad de agua en el tropico humedo delcentro de Veracruz, Mexico. PhD Dissertation, Autonomous Universityof Mexico, Mexico City, Mexico.

Nadkarni NM. 1984. Epiphyte biomass and nutrient capital of aneotropical elfin forest. Biotropica 16: 249–256.

Nadkarni NM, Solano R. 2002. Potential effects of climate changeon canopy communities in a tropical cloud forest: an experimentalapproach. Oecologia 131: 580–584.

Nair US, Ray DK, Lawton RO, Welch RM, Pielke RA Sr, Calvo-Alvarado J. 2010. The impact of deforestation on orographiccloud formation in a complex tropical environment. In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 538–548.

Nespor V, Sevruk B. 1999. Estimation of wind-induced error of rainfallgauge measurements using a numerical simulation. Journal ofAtmospheric and Oceanic Technology 16: 450–464.

New M, Hulme M, Jones PD. 2000. Global Monthly Climatology forthe Twentieth century , Oak Ridge National Laboratory DistributedActive Archive Center: Oak Ridge, Tennessee, USA. http://www.daac.ornl.gov. [accessed: June 2010].

Nullet DA, Juvik JO. 1994. Generalised mountain evaporation profilesfor tropical and subtropical latitudes. Singapore Journal of TropicalGeography 15: 17–24.

Oesker M, Homeier J, Dalitz H, Bruijnzeel LA. 2010. Spatial hetero-geneity of throughfall quantity and quality in tropical montane forestsin southern Ecuador. In Tropical Montane Cloud Forests. Science forConservation and Management,, Bruijnzeel LA, Scatena FN, Hamil-ton LS (eds). Cambridge University Press: Cambridge, UK; 393–401.

Pagiola S. 2002. Paying for water services in Central America: learningfrom Costa Rica. In Selling Forest Environmental Services: Market-based Mechanisms for Conservation and Development , Pagiola S,Bishop J, Landell-Mills N (eds). Earthscan: London, UK: London, UK;36–60.

Penafiel SR. 1995. The biological and hydrological values of the mossyforests in the Central Cordillera Mountains, Philippines. In TropicalMontane Cloud Forests , Ecological Studies , vol 110, Hamilton LS,Juvik JO, Scatena FN (eds). Springer Verlag: New York; 266–273.

Pocs T. 1980. The epiphytic biomass and its effect on the water balance oftwo rain forest types in the Uluguru Mountains (Tanzania, East Africa).Acta Botanica Academiae Scientiarum Hungariae 26: 143–167.

Ponette-Gonzalez AG, Weathers KC, Curran LM. 2009. Water inputsacross a tropical montane landscape in Veracruz, Mexico: synergisticeffects of land cover, rain and fog seasonality, and interannualprecipitation variability. Global Change Biology 16: 946–963.

Porras IT, Grieg-Gran M, Neves N. 2008. All That Glitters. A Reviewof Payments for Watershed Services in Developing Countries ,International Institute for Environment and Development: London, UK.

Pounds JA, Fogden MPL, Campbell JH. 1999. Biological response toclimate change on a tropical mountain. Nature 398: 611–615.

Pounds JA, Bustamante MR, Coloma LA, Consuegra JA, Fogden MPL,Foster PN, La Marca E, Masters KL, Merino-Viteri A, Puschendorf R,Ron SR, Sanchez-Azofeifa JA, Still CJ, Young BE. 2006. Widespreadamphibian extinctions from epidemic disease driven by globalwarming. Nature 439: 161–167.

Prada S, Menezes de Sequeiro M, Figueira C, Oliveira da Silva M. 2009.Fog precipitation and rainfall interception in the natural forests ofMadeira Island (Portugal). Agricultural and Forest Meteorology 149:1179–1187.

Pruppacher HR, Klett JD. 1978. Microphysics of Clouds and Precipita-tion, D. Reidel Publishing Company: Dordrecht, The Netherlands.

Raich JW, Russell AE, Vitousek PM. 1997. Primary productivity andecosystem development along an elevational gradient on Mauna Loa,Hawai’i. Ecology 78: 707–721.

Rautenbach H, Oliver J (eds). 2004. Third International Conference onFog and Fog Collection, International Development Research Center:Ottawa, Canada.

Ray DK, Nair US, Lawton RO, Welch RM, Pielke RA, Sr. 2006. Impactof land use on Costa Rican tropical montane cloud forests: sensitivityof orographic cloud formation to deforestation in the plains. Journalof Geophysical Research 111: D02108. DOI: 10.1029/2005JD006096.

Riehl H. 1979. Climate and Weather in the Tropics , Academic Press:London, UK.

Rodriguez-Zuniga JM. 2003. Paying for forest environmental services:the Costa Rican experience. Unasylva 212(54): 31–33.

Roman L, Scatena FN, Bruijnzeel LA. 2010. Global and local variationsin tropical montane cloud forest soils. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 77–89.

Rovito SM, Parra-Olea G, Vasquez-Alamazan CR, Papenfuss TJ, WakeDB. 2009. Dramatic declines in Neotropical salamander populationsare an important part of the global amphibian crisis. Proceedings ofthe National Academy of Sciences 106: 3231–3236.

Santiago LS, Goldstein G, Meinzer FC, Fownes J, Mueller-Dombois D.2000. Transpiration and forest structure in relation to soil waterloggingin a Hawaiian montane cloud forest. Tree Physiology 20: 673–681.

Scatena FN. 1998. An assessment of climatic change in the LuquilloMountains of Puerto Rico. American Water Resources Association TPS98-2: 193–198.

Scatena FN, Planos-Gutierrez EO, Schellekens J. 2005. Natural distur-bances and the hydrology of humid tropical forests. In Forests, Waterand People in the Humid Tropics , Bonell M, Bruijnzeel LA (eds).Cambridge University Press: Cambridge, UK; 489–512.

Scatena FN, Bruijnzeel LA, Bubb P, Das S. 2010. Setting the stage.In Tropical Montane Cloud Forests. Science for Conservationand Management , Bruijnzeel LA, Scatena FN, Hamilton LS (eds).Cambridge University Press: Cambridge, UK; 3–13.

Schawe M, Gerold G, Bach K, Gradstein SR. 2010. Hydrometeorolog-ical patterns in relation to montane forest types along an eleva-tional gradient in the Yungas of Bolivia. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press: Cam-bridge, UK; 199–207.

Schell D, Georgii H-W, Maser R, Jaeschke W, Arends BG, Kos GPA,Winkler D, Schneider T, Berner A, Kruisz C. 1992. Intercomparisonof fog water samplers. Tellus 44B: 612–631.

Schellekens J. 2006. CQ-FLOW: A distributed hydrological modelfor the prediction of impacts of land-cover change, withspecial reference to the Rio Chiquito catchment, northwestCosta Rica. Annex 3 to Final Technical Report DFID-FRPProject no. R7991. VU University Amsterdam, The Netherlands.http://www.geo.vu.nl/¾fiesta/. [accessed: June 2010].

Schemenauer RS, Bridgman HA (eds). 1998. First InternationalConference on Fog and Fog Collection, International DevelopmentResearch Center: Ottawa, Canada.

Schemenauer RS, Cereceda P. 1994. A proposed standard fog collectorfor use in high-elevation regions. Journal of Applied Meteorology 33:1313–1322.

Schemenauer RS, Cereceda P. 1995. Reply to comments by Juvik andNullet (1995). Journal of Applied Meteorology 34: 2111–2112.

Schemenauer RS, Puxbaum HA (eds). 2001. Second InternationalConference on Fog and Fog Collection, International DevelopmentResearch Center: Ottawa, Canada.

Schmid S, Burkard R, Frumau KFA, Tobon C, Bruijnzeel LA, Sieg-wolf R, Eugster W. 2010. Using eddy covariance and stable isotopemass balance techniques to estimate fog water contributions to a CostaRican cloud forest during the dry season. Hydrological Processes . DOI:10.1002/hyp.7739.

Scholl MA, Eugster W, Burkard R. 2010. Understanding the role of fogin forest hydrology: stable isotopes as tools for determining input andpartitioning of cloud water in montane forests. Hydrological Processes .DOI: 10.1002/hyp.7762.

Schroter Ch. 1926. Das Pflanzenleben der Alpen, Albert RausteinPublishers: Zurich, Switzerland.

Schrumpf M, Axmacher J, Zech W, Lyaruu HVM. 2010. Net precipita-tion and soil water dynamics in clearings, old secondary and old-growthforests in the montane rain forest belt of Mount Kilimanjaro, Tanzania.Hydrological Processes . DOI: 10.1002/hyp.7798.

Sharon D. 1980. Distribution of hydrologically effective rainfall incidenton sloping ground. Journal of Hydrology 46: 165–188.

Shuttleworth WJ. 1988. Evaporation from Amazonian rain forest.Philosophical Transactions of the Royal Society (London) Series B 233:321–346.

Sidle RC, Ziegler AD, Negishi JN, Nik AR, Siew R, Turkelboom F.2006. Erosion processes in steep terrain: truths, myths, anduncertainties related to forest management in Southeast Asia. ForestEcology and Management 224: 199–225.

Silver WL, Lugo AE, Keller M. 1999. Soil oxygen availability andbiogeochemistry along rainfall and topographic gradients in uplandwet tropical forest soils. Biogeochemistry 44: 301–328.

Stadtmuller T. 1987. Cloud Forests in the Humid Tropics: A Biblio-graphic Review. United Nations University (UNU): Tokyo: CATIE:

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

Page 33: Hydrometeorology of tropical montane cloud forests ...Hydrometeorology of tropical montane cloud forests: ... scale at which the model was applied (1 ð 1 km) and the scale of the

HYDROMETEOROLOGY OF TROPICAL MONTANE CLOUD FORESTS

Turrialba, Costa Rica. http://www.unu.edu/unupress/unupbooks/80670e/80670E00.htm. [accessed: 2009].

Stadtmuller T, Agudelo N. 1990. Amount and variability of cloudmoisture input in a tropical cloud forest. International Association ofHydrological Sciences Publication 193: 25–32.

Steinhardt U. 1979. Untersuchungen uber den Wasser- und Nahrstoff-haushalt eines andinen Wolkenwaldes in Venezuela. GottingerBodenkundliche Berichte 56: 1–185.

Still CJ, Foster PN, Schneider H. 1999. Simulating the effects of climatechange on tropical montane cloud forests. Nature 398: 608–610.

Takahashi M, Giambelluca TW, Mudd RG, DeLay DK, Nullet MA.2010. Rainfallpartitioning and cloud water interception in native forestan invaded forest in Hawai‘I Volcanoes National Park. HydrologicalProcesses . DOI: 10.1002/hyp.7797.

Tanaka K, Takizawa H, Tanaka N, Kosaka I, Yoshifuji N, Tantasirin C,Piman S, Suzuki M, Tangtham N. 2003. Transpiration peak over ahill evergreen forest in northern Thailand in the late dry season:assessing the seasonal changes in evapotranspiration using a multilayermodel. Journal of Geophysical Research 108(D17): 4533. DOI:10.1029/2002JD003028.

Tanaka N, Tantasirin C, Kuraji K, Suzuki M, Tangtham N. 2005. Inter-annual variation in rainfall interception at a hill evergreen forest innorthern Thailand. Bulletin of the Tokyo University Forest 113: 11–44.

Tanaka N, Kuraji K, Tantasirin C, Takizawa H, Tangtham N, Suzuki M.2010. Relationships between rainfall, fog and throughfall at a hillevergreen forest site in northern Thailand. Hydrological Processes .DOI: 10.1002/hyp.7729.

Tani M, Rahim Nik A, Yasuda Y, Noguchi S, Shamsuddin SA,Mohamed S, Takanashi S. 2003. Long-term estimation of evapotran-spiration from a tropical rain forest in Peninsular Malaysia. Interna-tional Association of Hydrological Sciences Publication 280: 267–274.

Tanner EVJ. 1980. Studies on the biomass and productivity in a series ofmontane› rain forests in Jamaica. Journal of Ecology 68: 573–588.

Tobon C. 2009. Los bosques andinos y el agua. Serie Investigaciony Sistematizacion 4. Programa Regional ECOBONA—INTERCO-OPERATION—CONDESAN: Quito, Ecuador.

Tobon C, Kohler L, Frumau KFA, Bruijnzeel LA, Burkard R, Schmid S.2010a. Water dynamics of epiphytic vegetation in a lower montanecloud forest: fog interception, storage and evaporation. In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 261–267.

Tobon C, Bruijnzeel LA, Frumau KFA, Calvo-Alvarado JC. 2010b.Changes in soil physical properties alter conversion of tropicalmontane cloud forest to pasture in northern Costa Rica. In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 502–515.

Tognetti S, Aylward B, Bruijnzeel LA. 2010. Assessment needs tosupport the development of arrangements for Payments forEcosystem Services from tropical montane cloud forests. In TropicalMontane Cloud Forests. Science for Conservation and Management ,Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge UniversityPress: Cambridge, UK; 671–685.

Van der Molen MK, Dolman AJ, Waterloo MJ, Bruijnzeel LA. 2006.Climate is affected more by maritime than by continental land usechange: a multiple-scale analysis. Global and Planetary Change 54:128–149.

Van der Molen MK, Vugts HF, Bruijnzeel LA, Scatena FN, Pielle RASr., Kroon LJM. 2010. Meso-scale climate change due to lowlanddeforestation in the maritime tropics. In Tropical Montane CloudForests. Science for Conservation and Management , Bruijnzeel LA,Scatena FN, Hamilton LS (eds). Cambridge University Press:Cambridge, UK; 527–537.

Van Steenis CGGJ. 1972. The Mountain Flora of Java, E.J. BrillPublishers: Leiden, The Netherlands.

Vazquez-Garcıa JA. 1995. Cloud forest archipelagos: preservation offragmented montane ecosystems in tropical America. In TropicalMontane Cloud Forests , Ecological Studies , vol. 110, Hamilton LS,Juvik JO, Scatena FN (eds). Springer Verlag: New York; 314–332.

Veneklaas EJ, Van Ek R. 1990. Rainfall interception in two tropicalmontane rain forests, Colombia. Hydrological Processes 4: 311–326.

Veneklaas EJ, Zagt RJ, van Leerdam A, van Ek R, Broekhoven AJ, vanGenderen M. 1990. Hydrological properties of the epiphyte mass of amontane tropical rain forest, Colombia. Vegetatio 89: 183–192.

Vernimmen RRE, Bruijnzeel LA, Romdoni D, Proctor J. 2007. Rainfallinterception in three contrasting rain forest types in Central Kalimantan,Indonesia. Journal of Hydrology 340: 217–232.

Vis M. 1986. Interception, drop size distributions and rainfall kineticenergy in four Colombian forest ecosystems. Earth Surface Processesand Landforms 11: 591–570.

Vogelmann HW. 1973. Fog precipitation in the cloud forests of EasternMexico. BioScience 23: 96–100.

Walmsley JL, Schemenauer RS, Bridgman HA. 1996. A method forestimating the hydrologic input from fog in mountainous terrain.Journal of Applied Meteorology 35: 2237–2249.

Walmsley JL, Burrows WR, Schemenauer RS. 1999. The use of routineobservations to calculate liquid water content in summertime high-elevation fog. Journal of Applied Meteorology 38: 369–384.

Weaver PL. 1972. Cloud moisture interception in the Luquillo mountainsof Puerto Rico. Caribbean Journal of Science 12: 129–144.

Weaver PL, Medina E, Pool D, Dugger K, Gonzalesliboy J, Cuevas E.1986. Ecological observations in the dwarf cloud forest of the LuquilloMountains in Puerto Rico. Biotropica 18: 79–85.

Webster PJ, Holland GJ, Curry JA, Chang HA. 2005. Changes in tropicalcyclone number, duration and intensity in a warming environment.Science 309: 1844–1846.

Wesseling CG, Karssenberg D, Van Deursen WPA, Burrough PA. 1996.Integrating dynamic environmental models in GIS: the development ofa dynamic modelling language. Transactions in GIS 1: 40–48.

Whitmore TC. 1998. Introduction to Tropical Rain Forest . ClarendonPress, Oxford, UK.

Williams S, Bolitho E, Fox S. 2003. Climate change in Australiantropical rainforests: an impending environmental catastrophe.Proceedings of the Royal Society of London Series B 270: 1887–1893.

Williams JW, Jackson ST, Kutzbach JE. 2007. Projected distributionsof novel and disappearing climates by 2100 AD. Proceedings of theNational Academy of Sciences 104: 5738–5742.

Wolf JHD. 1993. Diversity patterns and biomass of epiphytic bryophytesand lichens along an altitudinal gradient in the northern Andes. Annalsof the Missouri Botanical Garden 80: 928–960.

Yang D, Goodison BE, Ishida S, Benson CS. 1998. Adjustment of dailyprecipitation data at 10 climate stations in Alaska: application of WorldMeteorological Organization intercomparison results. Water ResourcesResearch 34: 241–256.

Yin XW, Arp PA. 1994. Fog contributions to the water budget offorested watersheds in the Canadian Maritime Provinces: a generalizedalgorithm for low elevations. Atmosphere-Ocean 32: 553–566.

Zadroga F. 1981. The hydrological importance of a montane cloudforest area of Costa Rica. In Tropical Agricultural Hydrology , Lal R,Russell EW (eds). John Wiley and Sons: New York; 59–73.

Zimmermann B, Elsenbeer H. 2008. Spatial and temporal variabilityof soil saturated hydraulic conductivity in gradients of disturbance.Journal of Hydrology 361: 78–95.

Zotz G, Bader MY. 2009. Epiphytic plants in a changing world—Globalchange effects on vascular and non-vascular epiphytes. Progress inBotany 70: 147–170.

Copyright 2010 John Wiley & Sons, Ltd. Hydrol. Process. (2010)

rmendoza
Inserted Text
insert space. This should read: Rainfall partitioning