watershed hydrology of the - soil and water lab: home...
TRANSCRIPT
1
Watershed Hydrology of the (Semi) Humid Ethiopian Highlands 1
2
Tegenu A. Engda1,2, Haimanote K. Bayabil1,2, Elias S. Legesse1,3, Essayas K. Ayana3,4, Seifu A. 3
Tilahun3,4, Amy S. Collick1,3,4, Zachary M. Easton4, Alon Rimmer5, Seleshi B. Awulachew6 and4
Tammo S. Steenhuis1,3,4 5
6
1 Cornell Master’s program in Integrated Watershed Management and Hydrology, Bahir Dar, Ethiopia 7
2 Amhara Regional Agricultural Research Institute, Debra Tabor, Ethiopia 8
3 Department of Water Resources and Environmental Engineering, Bahir Dar University, Bahir Dar, 9
Ethiopia10
4 Department of Biological and Environmental Engineering, Cornell University, Ithaca, NY USA 11
5 The Kinneret Limnological Laboratory, Migdal, Israel 12
6 International Water Management Institute, East Africa Office, Addis Ababa, Ethiopia 13
* Corresponding author: Tammo S. Steenhuis 14
Phone: 607-255-2489 15
Fax: 607-255-4449 16
Email: [email protected] 17
Keywords: Variable Source Area, perched water table, saturation excess, infiltration excess, hillslope 18hydrology19
202122
Email23Tegenu Ashagrie Engda <[email protected]>, Haimanote K. Bayabil <[email protected]>,24Elias Leggesse <[email protected]>, Essayas Kaba Ayana <[email protected]>, 25Seifu A. Tilahun <[email protected]>, Amy S Collick [email protected],26Zach Easton<[email protected]>, Alon Rimmer <[email protected]> 27Seleshi B. Awulachew <[email protected]>, Tammo S. Steenhuis <[email protected]>28
2
29
Table of Contents30Abstract ..................................................................................................................................................... 3311. Introduction .......................................................................................................................................... 4322. Watershed Descriptions ........................................................................................................................ 6333. Rainfall runoff characteristics for monsoonal climates. ....................................................................... 834
3.1 Analysis of rainfall discharge data .................................................................................................. 8353.2 Infiltration and precipitation intensity measurements ................................................................... 10363.3 Piezometers and ground water table measurements ..................................................................... 12373.4 Conceptual model for predicting watershed discharge ................................................................. 1438
4. Simulation results ............................................................................................................................... 16395. Conclusions ........................................................................................................................................ 1940References ............................................................................................................................................... 2041Abbreviations .......................................................................................................................................... 2342List of Tables and Figures ....................................................................................................................... 2443
Table 1: Location, description, and data used in the model from the three SCRP research sites ....... 2544Table 2: Average soil infiltration rate at different soil type, slope range and land use types .............. 2545Table 3: Model input values for surface flow components for the three SCRP watersheds and the Nile 46upstream of the border with Sudan. The watershed is divided into regions with different 47characteristics: exposed bedrock and saturated areas that contribute surface runoff when the soil is 48saturated or hillsides that produce recharge when the soil is above field capacity. Maximum storage 49of water is the amount of water needed above wilting point to become either saturated (runoff 50contributing areas) or to reach field capacity (hill slopes). Model input values for the baseflow and 51interflow reservoirs are the maximum storage of the linear base flow reservoir; the time in days to 52reduce the volume of the baseflow reservoir to half under no recharge conditions, t* is the duration 53of the period after a single rainstorm until interflow ceases, Nash-Sutcliffe efficiency for simulated 54daily averaged discharge for the three SCRP watersheds and ten day average values for the Blue Nile 55Basin were calculated for the periods of calibration and validation. .................................................. 2656
5758
3
59
Abstract60
Understanding the basic relationships between rainfall, runoff and soil loss is vital for effective 61
management and utilization of water resources and soil conservation planning. A study was conducted 62
in three small watersheds in or near the Blue Nile basin in Ethiopia, with long-term records of rainfall 63
and discharge. To better understand the water movement within the watershed, piezometers were 64
installed and infiltration rates were measured in the 2008 rainy season. We also reanalyzed the 65
discharge from small plots within the watersheds. Infiltration rates were generally in excess of the 66
rainfall rates. Based on this and plot discharge measurements, we concluded that most rainfall 67
infiltrated into the soil, especially in the upper, steep and well-drained portions of the watershed. Direct 68
runoff is generated either from saturated areas at the lower and less steep portions of the hill slopes or 69
from areas of exposed bedrock. Using these principles, a simple distributed watershed hydrology model 70
was developed. The models reproduce the daily discharge pattern reasonably well for the small 71
watershed and the ten-day discharge values for the whole Blue Nile Basin in Ethiopia. The simplicity 72
and scalability of the model hold promise for use in un-gauged catchments. 73
4
74
1. Introduction 75
A better understanding of the hydrological characteristics of different watersheds in the headwaters of 76
the Ethiopian highlands is of considerable importance because only 5% of Ethiopia’s surface water 77
(0.6% of the Nile Basin’s water resource) is being currently utilized by Ethiopia while cyclical droughts 78
cause food shortages and intermittent famine (Arseno and Tamrat 2005). At the same time the 79
Ethiopian highlands are the origin, or source, of much of the river flow reaching the Nile River, 80
contributing greater than 60% of Nile flow (Ibrahim 1984; Conway and Hulme 1993) possibly 81
increasing to 95% during the rainy season (Ibrahim 1984). In addition, there is a growing anxiety about 82
climate and human-induced changes of the river discharge (Sutcliffe and Parks 1999), especially 83
because there have been limited studies on basin characteristics, climate conditions, and hydrology of 84
the Upper Nile Basin in Ethiopia (Arseno and Tamrat 2005). 85
86
To predict future water availability generally three types of models have been used in the Blue Nile 87
basin. Simple engineering approaches such as the Rational Method (Desta 2003), pure water balance 88
models, and semi distributed models. Pure water balance approaches have been made by Ayenew and 89
Gebreegziabher (2006) for Lake Awassa, Conway (1997) and Kim and Kaluarachchi (2008) at the 90
upper Blue Nile, and Kebede et al. (2006) at Lake Tana. These models provide only information on 91
water quantity at the watershed outlet and perform best at a monthly time scale. Semi distributed 92
models that have been applied in the Nile basin are SWAT (Setegn et al. 2008), Water Erosion 93
Prediction Project (WEPP) (Zeleke 2000), the Agricultural Non-Point Source model (AGNPS) 94
(Haregeweyn and Yohannes 2003; Mohammed et al. 2004), and for South Central Ethiopia PRMS 95
(Legesse et al. 2003).96
97
5
Most semi distributed models use the SCS runoff equation for determining surface runoff. The SCS 98
curve number method is based on a statistical analysis of plot runoff data from the midwest USA with a 99
temperate climate. When applied to Ethiopia with a monsoon climate, it has been a bit problematic. 100
While most of these models were run on a daily time step, the results are presented on a monthly time 101
step, which indicates these models do not work well at the daily scale. By integrating the result over a 102
monthly time step, errors in daily surface runoff are compensated for by opposite errors in interflow 103
predictions. Thus, the monthly validation indicates that the monthly water balances are met but not 104
necessarily that the daily rainfall-runoff relationship for landscape units in the watershed are correct. In 105
order to understand why a statistically derived and widely used SCS curve number fails to predict the 106
daily direct runoff in monsoonal climates, the effect of climate on the hydrology during the dormant 107
and growing seasons must be considered. The similarity between temperate and monsoonal climate 108
types is that both have a dormant period and a growing period. However, the similarity between the two 109
climates stops there. In the temperate climates, the growth in the dormant season is limited by the 110
temperature. There is usually sufficient precipitation and there is little evaporation with the result that 111
the soils wet up and watershed outflow increases. In monsoonal climates, the limiting factor is 112
insufficient rainfall with the consequence that the soils dry out and therefore during the dormant 113
season, discharge out of the watershed decreases. Understanding the effect of climate on the hydrology 114
during the growing season is more complicated. We can simply make the observation that the 115
landscape dries out in temperate climates and the opposite is true for monsoonal climate. These 116
differences in how climate interacts with the hydrology, indicates that only physically sound models 117
can be applied in both climates and statistical techniques will be less successfully transferred.118
119
Thus, in order to refine the estimates of watershed outflow in the Ethiopian highlands with a 120
monsoonal climate, a better understanding is needed of rainfall and runoff relationships of the various 121
6
landscape units. In this chapter, we will use recently collected information by Engda (2009), Leggesse 122
(2009) and Bayabil (2009) in three Soil Conservation Research Program (SCRP) watersheds in the 123
Ethiopian highlands - Andit Tid, Anjeni, and Maybar - to derive how interflow, surface runoff, and 124
baseflow are generated spatially. This information then will be used to develop a physically sound 125
model that can be used in the Ethiopian highlands,126
127
2. Watershed Descriptions 128
Soil Conservation Research Program (SCRP) watersheds have the longest and most accurate record of 129
both rainfall and runoff data available for small watersheds in Ethiopia. Three of the watersheds are 130
located in the Amhara region either in or close to the Nile Basin: Andit Tid, Anjeni, and Maybar (Fig. 131
1). All three sites are dominated by agricultural with soil erosion control structures built to assist the 132
rain-fed subsistence farming; but the watersheds differ in size, topographic relief, and climate (Table 1). 133
134
The Andit Tid watershed unit covers a total area of 481 ha, with a hydrological surface area of 477 ha. 135
It is situated 180 km northeast of Addis Ababa at 39°43’ east and 9°48’ north in the Blue Nile Basin. 136
The topography of Andit Tid ranges from 3000 m near the research station in the western reach of the 137
research unit to 3500 m in the southeast. Andit Tid, the largest watershed, is also the highest and least 138
populated. It receives more than 1500 mm/year and has only one rainy season, typically May through 139
October. Hill slopes are very steep and degraded, resulting in 54% of the long-term precipitation 140
becoming runoff. Despite its larger size, stream flow quickly returns to nearly zero during the typically 141
dry months of November through March. Some of the contour buns constructed at the start of the 142
project were destroyed after installation because, according to farmers, they increased erosion (Engda 143
2009). Terraces and small contour drainage ditches were installed by the farmers to carry off excess 144
rainfall. From 1987 to 2004, rainfall was measured and discharge was recorded at the outlet and from 145
7
four runoff plots. Evaporation was measured using the standard pan. During the 2008 rainy season, soil 146
infiltration rates were determined at 10 different locations throughout the watershed using a 30-cm 147
diameter single-ring infiltrometer. In addition, piezometers were installed in transects to measure the 148
water table depths. Finally, soil depth estimations were taken by field technicians throughout the 149
watershed and registered using Global Positioning System (GPS) points. Further information can be 150
found in Hurni (1984), Bosshart (1997a), and Engda (2009). 151
152
The Anjeni watershed is located in the Blue Nile Basin of Amhara Region in one of the country’s more 153
productive agricultural areas and is dominated by highlands. The watershed is oriented north-south and 154
is flanked on three sides by plateau ridges. It is located at 37o31’E and 10o40’N and lies 370 km NW of 155
Addis Ababa to the south of the Choke Mountains. Minchet, a perennial river, starts in the watershed 156
and flows towards the Blue Nile Gorge. The Anjeni watershed covers a total area of 113 ha. It is the 157
most densely populated among the three watersheds. The topography of Anjeni ranges from 3000 m 158
near the research station in the western reach of the research unit to 3500 in the southeast. This site 159
receives more rain than the other two watersheds and has only one rainy season, typically May through 160
October. This watershed has extensive soil and water conservation measures, mainly terraces and small 161
contour drainage ditches, installed each year by the farmers to carry off excess rainfall. From 1987 to 162
2004, rainfall was measured at five different locations, discharge was recorded at the outlet and from 163
four runoff plots. Forty-five percent of the rainfall becomes runoff. During the 2008 rainy season, soil 164
infiltration rate was measured at 10 different locations throughout the watershed using a 30-cm 165
diameter single-ring infiltrometer. In addition, piezometers were installed in transects to measure the 166
water table depths. Finally, soil depth estimations were taken by field technicians throughout the 167
watershed and registered using GPS points.168
169
8
Finally, Maybar is located in the northeastern part of the central Ethiopian highlands situated in 170
Southern Wollo Administrative Zone near Dessie Town. It is the first of the SCRP research sites. The 171
gauging station lies at 39o39’E and 10o51’ N. The area is characterized by highly rugged topography 172
with steep slopes ranging between 2530 and 2860 m, a 330 m altitude difference within a 112.8 ha)173
catchment area. From 1988 to 2004, rainfall data was available using automatic rain gauges and two 174
manual rain gauges at two different locations, one in the upper part of the catchment and the other near 175
the office (climatic station). Discharge was measured with a flume installed in the Kori River using two 176
methods: float-actuated recorder and manual recording. During the main rainy season in 2008, 34 % of 177
the long-term precipitation in Maybar became discharge at the outlet. The ground water table levels 178
were measured with 29 piezometers. The saturated area in the watershed was delineated and mapped 179
using combined information collected using GPS instrument, field observation, and ground water level 180
data (piezometer head readings). 181
182
3. Rainfall runoff characteristics for monsoonal climates. 183
3.1 Analysis of rainfall discharge data 184185
In order to understand the rainfall/runoff relations in monsoonal climates, Liu et al. (2008) examined 186
how the watershed outflow changed as a function of precipitation for the three SCRP sites. To find the 187
most appropriate representation of watershed behavior, daily rainfall, evaporation, and discharge data 188
were summed over biweekly periods; only with longer time periods could the total stream responses to 189
a rainfall event be determined because interflow lasts several days.190
191
To investigate runoff response patterns, the biweekly sums of discharge were plotted as a function of 192
effective rainfall (i.e., precipitation-evapotranspiration) during the rainy season and dry season, 193
respectively. In Fig. 2a, an example is given for the Anjeni watershed. As is clear from Fig. 2a, the 194
9
watershed behavior changes as the wet season progresses, with precipitation later in the season 195
generally producing a greater percentage of runoff. As rainfall continues to accumulate during the rainy 196
season, the watershed eventually reaches a threshold point where runoff response can be predicted by a 197
linear relationship with effective precipitation, indicating that the proportion of the rainfall that became 198
runoff was constant during the remainder of the rainy season. For the purpose of this study, an 199
approximate threshold of 500 mm of effective cumulative rainfall, P-E, was selected after iteratively 200
examining rainfall vs. runoff plots for each watershed. The proportion Q/(P-E) varies within a relative 201
small range for the three SCRP watersheds despite their differing characteristics. In Anjeni, 202
approximately 48% of late season effective rainfall, P-E, became runoff, while ratios for Andit Tid and 203
Maybar were 56% and 50%, respectively (Liu et al. 2008). There was no correlation between biweekly 204
rainfall and discharge during the dry seasons at any of the sites.205
206
Since each of the SCRP watersheds showed a similar linear response after the threshold cumulative 207
rainfall was satisfied, the latter parts of the wet seasons were all plotted in the same graph (Fig. 2b). 208
Despite the great distances between the watersheds and the different characteristics, the response was 209
surprisingly similar. The Anjeni and Maybar watersheds had almost the same runoff characteristics, 210
while Andit Tid had more variation in the runoff amounts but, on average, the same linear response was 211
noted with a higher intercept (Fig. 2b). Linear regressions were generated for both the combined results 212
of all three watersheds and for the Anjeni and Maybar watersheds in combination (Fig. 2b).The 213
regression slope does not change significantly, but this is due to the more similar Anjeni and Maybar 214
values dominating the fit (note that these regressions are only valid for the end of rainy seasons when 215
the watersheds are wet). 216
217
Why these watersheds behave so similarly after the threshold rainfall has fallen is an interesting 218
10
question to explore. It is imperative to look at various time scales, since focusing on just one type of 219
visual analysis can lead to erroneous conclusions. For example, looking only at storm hydrographs of 220
the rapid runoff responses prevalent in Ethiopian storms, one could conclude that infiltration excess is 221
the primary runoff generating mechanism. However, looking at longer time scales in Fig. 2a, it can be 222
seen that the ratio of Q/(P-E) is increasing with cumulative precipitation and consequently the 223
watersheds behave differently depending on how much moisture is stored in the watershed. This 224
suggests that saturation excess processes play an important role in the watershed runoff response. If 225
infiltration excess was controlling runoff responses, discharge would only depend on the rate of 226
rainfall, and there would be no clear relationship with antecedent cumulative precipitation, as is clearly 227
the case in Fig. 2a. 228
229
3.2 Infiltration and precipitation intensity measurements 230231
To further investigate the likelihood of infiltration excess, the infiltration rates are compared with 232
rainfall intensities in the Andit Tid watershed where infiltration rates were measured in 2008 by Engda 233
(2009) and rainfall intensity records were available from the SCRP project for 1986 to 2004 on the 234
pluviometric charts. These were transcribed to digital form by Engda (2009). The exceedance 235
probability of the average intensities of 23,764 storm events is plotted in Fig. 3 (blue line). These 236
intensities were calculated by dividing the rainfall amount on each day by the duration of the storm. In 237
addition, the exceedance probability for actual intensities of short periods ranging from 5 to 10 minutes 238
are plotted in Fig 3 (red line). Since there were bursts of high intensity rainfall within each storm, the 239
rainfall intensities for short periods exceeded that of the storm averaged intensities in Fig. 3. 240
241
The infiltration rates for 10 locations measured with the 30-cm diameter single-ring infiltrometer 242
(shown in Table 2) varied between a maximum of 87 cm/hr on a terraced eutric cambisol in the bottom 243
11
of the watershed to a low of 2.5 cm/hr on a shallow soil near the top of the hill slope. This low 244
infiltration rate was mainly caused by the compaction of free roaming grazing animals. Bush lands, 245
which are dominant on the upper watershed, have significantly higher infiltration rates. Terraced and 246
cultivated lands in general have also higher infiltration rates. The average infiltration rate of all 10 247
measurements indicates that the storm intensities were 20.3 cm/h and the medium 4.8 cm/h. The 248
median infiltration rate of 4.8 cm/h is the most meaningful number to compare with the rainfall 249
intensity since it represents a spatial average. This median intensity has an exceedance probability of 250
0.03 for the actual storm intensities and 0.006 for the storm averaged intensities. Thus, the medium 251
intensity was exceeded only 3% of the time and for less than 1% of the storms. Storms with greater 252
intensities were all of short duration with amounts of less than 1 cm of total precipitation, except once, 253
in which almost 4 cm of rain fell over a 40 minute period. The runoff generated during short duration 254
intense rainfall can infiltrate into the soil in the subsequent period down slope when the rainfall 255
intensity is less or the rain has stopped. In the Maybar watershed, Derib (2005) performed 16 256
infiltration tests and observed even greater infiltration rates than in Andit Tid. The final steady state 257
infiltration rates ranged from 1.9 cm/hr to 60 cm/hr with a median of 17.5 cm/hr. 258
259
Thus, the infiltration measurements confirm that infiltration excess runoff is not a common feature in 260
these watersheds. Consequently, most runoff that occurs in these watersheds is from degraded soils 261
where the top soil is removed and by saturation excess in valley bottoms where the interflow 262
accumulates. Since the degraded soils have little storage, the runoff can be classified as either 263
infiltration excess or saturation excess. 264
265
The finding that saturation excess is occurring in watersheds with a monsoonal climate is not unique. 266
For example, Hu et al. (2005), Lange et al. (2003), and Merz et al. (2006) found that saturation excess 267
12
could describe the flow in a monsoonal climate in China, Spain, and Nepal. There are no previous 268
observations published for Ethiopia on the suitability of these saturation excess models to predict 269
runoff even though attempts to fit regular models based on infiltration excess principles were not 270
always satisfactorily (Haregeweyn et al. 2003; Zeleke 2000).271
272
3.3 Piezometers and ground water table measurements 273274
Ground water table height measurements allow us to determine how the rain that falls on the upslope 275
areas reaches the river. In all three watersheds, transects of piezometers were installed and ground 276
water tables were observed in 2008 during the rainy season until the beginning of the dry season. 277
278
Both Andit Tid and Maybar have hill slopes with shallow to medium depth soils (0.5-2.0 m depth) 279
above a sloping slowly permeable layer (either a hardpan or bedrock). Consequently, the water table 280
height above the slowly permeable horizon (as indicated by the piezometers) behaved similarly for both 281
watersheds. An example is given for the Maybar watershed where ground water table levels were 282
measured with 29 piezometers across eight transects. The whole watershed was divided into three slope 283
ranges: upper steep slope [25.1° – 53.0°], mid slope [14.0°– 25.0°], and relatively low-lying areas [0°- 284
14.0°]. For each slope class, the daily perched ground water depths were averaged (i.e., the height of 285
the saturated layer above the restricting layer, Figure 4a). The depth of the perched ground water above 286
the restricting layer in the steep and upper parts of the watershed is very small and disappears if there is 287
no rain for a few days. The depth of the perched water table on the mid slopes is greater than upslope 288
areas. The perched ground water depths are, as expected, the greatest in relatively low-lying areas. 289
Springs occur at the locations where the depth from the surface to the impermeable layer is the same as 290
the depth of the perched water table and are the areas that the surface runoff is generated. 291
292
13
The water table behavior is consistent with what one would expect if interflow is the dominate 293
conveyance mechanism. All else being equal, the greater the driving force (i.e., the slope of the 294
impermeable layer), the smaller the perched ground water depth required to transport a given quantity 295
of water downslope. Moreover, the drainage area and the discharge increase with down slope position. 296
Consequently, one expects that the perched groundwater table depth increases with down slope position 297
as both slope decreases and drainage area increases. 298
299
These findings are different than those generally believed to be the case, i.e., that the vegetation 300
determines the amount of runoff in the watershed. We therefore plotted the average daily depth of the 301
perched water table under the different crop types (Figure 4b). Unexpectedly, there was a strong 302
correlation of perched water depth with crop type as well. The grassland had the greatest perched water 303
table depth, followed by cropland and bush land with the lowest ground water level. However, some 304
local knowledge was needed to interpret these data, as the grasslands are mainly located in the often 305
saturated lower lying areas (too wet to grow a crop), the croplands are in the mid-slope (with a 306
consistent water supply but not saturated) and the bush lands are in the upper steep slope areas (too 307
droughty for good yield). Since land use is related to slope class, we expect the same relationship 308
between crop type and soil water table height as slope class and water table height. Thus, there is an 309
indirect relationship between land use and hydrology determined by the landscape features. The 310
landscape determines the water availability and thus the land use. 311
312
In the Anjeni watershed, which had relatively deep soils and no flat bottom land, the only water table 313
that was found was near the stream. The water table level was above the stream level indicating that the 314
rainfall infiltrates the landscape first and then flows laterally to the stream. Although more 315
measurements are needed, we speculate that there was a portion of the watershed that had a hard pan at 316
14
a relative shallow depth causing saturation excess overland flow. 317
318
These findings are consistent with the measurements taken by McHugh (2006) in the Lenche Dima 319
watershed near Woldea where the surface runoff of the flat bottom lands was much greater than the 320
runoff (and erosion) from the hillsides. Thus, in summary, the generally held opinion that the hill slopes 321
are the high surface runoff producing areas is not true, at least at a minimum, for the season of 322
observation in the three watersheds. The only areas that are expected to produce surface runoff are the 323
severely eroded areas where the bedrock or subsoil is exposed and other areas that saturate during the 324
storms. 325
326
3.4 Conceptual model for predicting watershed discharge 327328
In order to develop a realistic hydrological model, the interflow and saturation excess flow phenomena 329
must be included. In our conceptual watershed model, the watershed is divided into three areas based 330
on slope steepness, soil depth, and infiltration capacity of the soil. Surface runoff source areas include 331
areas near the river and the degraded hillsides with little or no soil cover. The well-drained hillsides 332
transmit water as interflow to the stream and are modeled as the third component. Both the degraded 333
and the bottom lands produce surface runoff after they are saturated. For a better understanding of the 334
processes, the three areas are schematically superimposed on a photograph of the upper part of the 335
Andit Tid watershed (Figure 5a). In addition to the three surface areas modeled, we included a 336
subsurface reservoir that generates baseflow. 337
338
For each of the three source areas, a water balance is kept in the model 339
340
(1) 341
15
342
where P is precipitation, (mm d-1); AET is the actual evapotranspiration, (mm d-1), St- t, previous time 343
step storage, (mm), R saturation excess runoff (mm d-1), Perc is percolation to the subsoil (mm d-1) and344
t is the time step. 345
346
Based on a linearly decreasing evaporation rate from the maximum moisture content at saturation or 347
field capacity to wilting point, the surface soil layer moisture storage can be written as: 348
349
(2) 350
351
where Smax (mm) is the maximum available soil storage capacity and is defined as the difference 352
between the amount of water stored in the topsoil layer at wilting point and the maximum moisture 353
content, equal to either the field capacity for the hill slope soils or saturation (e.g., soil porosity) in 354
runoff contributing areas. Smax varies according to soil characteristics (e.g., porosity, bulk density) and 355
soil layer depth. 356
357
By assigning a maximum storage to each of the three source areas, a water balance is maintained for 358
each of the watershed source areas. Surface runoff occurs from the degraded hillsides and the valley 359
bottom when the water balance indicates that the soil is saturated. The flatter areas remain wet even 360
during dry months of the year; only the top most soil layer will dry due to small amounts of water 361
percolating downward from the hills. Hence, these areas need only a small amount of rainfall to start 362
generating surface runoff. There is even less rainfall needed at the beginning of the rainy season for the 363
degraded hillsides to produce surface runoff 364
365
16
Interflow is generated by the excess rainfall when the hillside is at field capacity. Because these 366
hillsides drain fully and are at or near the wilting point before the rainy season, a significant amount 367
(on the order of 300 - 400 mm) of effective rainfall (defined as precipitation minus potential 368
evapotranspiration) is needed before the interflow starts to contribute to the lower slope areas. The 369
interflow can be modeled as a zero reservoir. This means that the same quantity of water drains from 370
the hillside each day for each storm. The result is in a linear recession curve. The outflow from 371
different storms is superimposed. Finally, the baseflow is modeled as a linear reservoir with a 372
maximum storage. This base flow reservoir adds water to the stream during the dry period. 373
374
This model is similar to that developed by Steenhuis et al. (2009) for the whole Ethiopian Blue Nile 375
Basin but slightly different from Collick et al. (2009) who tested a similar water balance model for the 376
same three SCRP watersheds. Collick et al. (2009), using the semi distributed model, did not have 377
particular landscape units in mind but fitted four areas that produced runoff and interflow based on 378
specific ratios. Our current semi distributed model for the Ethiopian Highlands is more physical based 379
and we do not need to specify a ratio between surface runoff and percolation of water. 380
4. Simulation results381
382
The mathematics for this conceptual model is presented in Steenhuis et al. (2009). The mathematical 383
model was calibrated for each of the three watersheds by Legesse (2009) for Anjeni, by Bayabil (2009) 384
for Maybar, by Engda (2009) for Andit Tid, and by Steenhuis et al (2009) for the whole Blue Nile 385
Basin. We will present in this chapter the results for two of the years between 1992 and 1995 depending 386
on what quality data was available for each of the three SCRP watersheds. Parameters were slightly 387
adjusted to represent the period after which the conservation practices were in place. 388
389
17
Parameters needed to simulate discharge include potential evaporation (PET), which varies little 390
between years and between watersheds. On average, PET was 5 mm d-1 during the dry season and 3 391
mm d-1 during the rainy season. The precipitation values used were measured for the small SCRP 392
watershed with rain gauges in the watershed itself and for the whole Blue Nile Basin, the average of 10 393
stations were used.394
395
For calibration, the maximum storage values, Smax, for the contributing areas and hill slopes were based 396
initially on the values of Steenhuis et al. (2009) for the whole Blue Nile Basin. Smax values for the three 397
source areas were then varied around these values to obtain the best fit. Subsurface parameters were 398
adjusted at the same time to fit the recession flows. The parameter set with the highest Nash Sutcliff 399
efficiency was selected. The validation used the most optimum parameter set. 400
401
The optimum parameter set for each of the three basins and for the Ethiopian highlands are shown in 402
Table 3 and the comparison of observed versus predicted values for the SCRP watersheds are shown in403
Figs. 6a, b, and c. The model fit the observed data well for the SCRP watersheds with Nash Sutcliffe 404
efficiencies for daily values in the range of 0.78 to 0.88 (Table 3) for the small watersheds. Due to 405
difficulties with obtaining a true average rainfall for the whole Blue Nile Basin, Nash Sutcliffe 406
efficiency for the Blue Nile Basin was lower. 407
408
The parameter values for all three SCRP watersheds fall in the same range: 0-23% for the saturated 409
bottom lands with a maximum storage of 110 mm; 0-20% for the degraded lands with storage values 410
less than 150 mm; and 50-75% for the hillsides with storages up to 250 mm. The degraded land in the 411
Anjeni watersheds consists of agricultural soils on terraced land, which soil has a restricting layer at 412
shallow depths. In Maybar and Andit Tid, the degraded soils consist of both exposed hardpan areas and 413
18
slipping hillside areas. Maybar has less of the soils than Andit Tid. This is the reason that the 414
maximum storage for Anjeni is greater than for the other two watersheds.415
416
Good fits were obtained between simulated and observed values for both the daily calibration and 417
validation periods. However, the model underestimates most peak flow periods for Anjeni and Andit 418
Tid (Figures 6b and 6c). Water balance type models have difficulty handling intense convective storms 419
or events of very short duration but high intensity rainfall. In addition, it is likely that larger parts of the 420
hillside contributed to the flow than initialized in the model structure in which the watershed is divided 421
in three regions. Seasonal variations in rainfall amount and distribution may affect the extent of 422
saturated areas and thereby stream flow generation. Several researchers point out that the dynamic 423
variation of stream zone saturated areas due to accumulation of lateral water flow from upslope and the 424
ground water system is responsible for a highly non-linear catchment response during storm events 425
(Todini 1995; Bari and Smettem 2004). Adding a fourth region that can saturate during large storm 426
events may produce better estimates for these high runoff events. 427
428
The variation among the parameters was in agreement with the landscape characteristics. In Maybar, 429
the model fitting indicated that there were almost no degraded hill slopes, as was the case. For the 430
Anjeni watershed that had a deep gully and ground water tables generally at least 2 m below the 431
surface, the model fitted (fit?) the data best if there was no saturated area included. Note also that for 432
the two smallest watersheds, Maybar and Anjeni, the total area did not add up to one (one what?)433
because of regional flows where water drained under the gage. But, for the whole Blue Nile Basin and 434
Andit Tid, the water balance closed and the total area summed to one (one what?). Finally, the 435
magnitude of the subsurface parameters increased with the size of the watershed as expected because as 436
watershed size increases, more deep flow paths become activated in transport. 437
19
438
5. Conclusions 439
440
Direct runoff is generated either from saturated areas at the lower portions of the hill slopes or from 441
areas of exposed bedrock while the upper hill slopes are infiltration zones. As a result, the watersheds 442
were divided into variable saturated areas, exposed rock and hill slopes. This was verified by high 443
measured infiltration rates on hill slope areas and shallow ground water depth at the bottom flat lands. 444
Other findings showed that the lower slope areas produced high runoff compared to high slopes for a 445
given rainfall event. The discharge in each of three watersheds was modeled by separately using a 446
simple water balance type model for degraded hillsides, saturated areas, and non-degraded hillsides. 447
The main input data for the model are rainfall, evaporation, the relative magnitude of the three areas, 448
and soil water holding capacity for the three areas. In addition, interflow and baseflow constants are 449
needed. The model results were encouraging, not only for the three small watersheds, but for the Blue 450
Nile watershed at the Ethiopian-Sudan border as well. The model has the potential to predict runoff in 451
ungauged basins using a small amount of field data.452
20
453
References454
455
Arseno Y, and Tamrat I. 2005. Ethiopia and the Eastern Nile Basin. Aquatic Sciences 16: 15- 27. 456
Ayenew, T., Gebreegziabher, Y. 2006. Application of a spreadsheet hydrological model for computing 457
the long-term water balance of Lake Awassa, Ethiopia. Hydrological Sciences 51 (3): 418 – 458
431.459
Bayabil. H.K., 2009. Modeling rainfall runoff relationships at Maybar watershed, Wollo, Ethiopia. 460
Master Thesis, Integrated Watershed Management and Hydrology Program, Cornell University, 461
Bahir Dar Ethiopia and Ithaca, NY, USA. 462
Bari, M., and Smettem, K.R.J., 2004. Modeling monthly runoff generation processes following land 463
use changes: Ground water-surface runoff interactions. Hydrology and Earth System Sciences 464
8(5), 903 – 922, 2004. 465
Bosshart, U., 1997. Measurement of River discharge for the SCRP research Catchments: Gauging 466
station profiles. Soil conservation research project, Research reports 31 University of Bern, 467
Bern.468
Collick, A.S., Easton, Z.M., Engda, T.A, Ashagrie, B.B, Adgo, E., Awulachew, S.B., Zeleke, G., and 469
Steenhuis, T.S., 2009. Application of physically based water balance model on four watersheds 470
throughout the upper Nile basin in Ethiopia. Hydrological Processes. 23: 3718–3727.471
Conway D, and Hulme M. 1993. Recent fluctuations in precipitation and runoff over the Nile sub-472
basins and their impact on main Nile discharge. Climatic Change 25: 127-151. 473
Conway D. 1997. A water balance model of the upper Blue Nile in Ethiopia. Hydrological474
Sciences 42: 265-282. 475
Derib S.D., 2005. Rainfall-Runoff Processes at a Hill-slope Watershed: Case of Simple Models 476
Evaluation at Kori-Sheleko Catchments of Wollo, Ethiopia. M.Sc. Thesis. 477
Desta, G. 2003. Estimation of Runoff Coefficient at different growth stages of crops in the highlands of 478
Amhara Region. M.Sc. Thesis, Alemaya University, Ethiopia. 479
Engda, T. A., 2009. Modeling rainfall-runoff-soil loss relationships of North-eastern Highlands of 480
Ethiopia, Andit Tid watershed, Ethiopia. . Master Thesis, Integrated Watershed Management 481
and Hydrology Program, Cornell University, Bahir Dar Ethiopia and Ithaca, NY, USA. 482
21
Haregeweyn, N. and Yohannes, F. 2003. Testing and evaluation of the agricultural non-point source 483
pollution model (AGNPS) on Augucho catchment, western Hararghe, Ethiopia. Agriculture 484
Ecosystems & Environment 99: 201-212.485
Hu, C.H., S.L. Guo, L.H. Xiong, D.Z. Peng. 2005. A modified Xinanjiang model and its application in 486
northern China. Nordic Hydrology 3: 175-192. 487
Hurni, H. 1984. The third progress report. Soil conservation research project, Vol.4. University of Bern 488
and the United Nations University. Ministry of Agriculture, Addis Ababa, Ethiopia. 489
Ibrahim AM. 1984. The Nile – Description, hydrology, control and utilization. Hydrobiologia. 110: 1-490
13.491
Kebede S, Travi Y, Alemayehu T, Marc V. 2006. Water balance of Lake Tana and its sensitivity to 492
fluctuations in rainfall, Blue Nile basin, Ethiopia. Journal of Hydrology 316: 233-247.493
Kim, U. and Kaluarachchi, J. J. 2008. Application of parameter estimation and regionalization 494
methodologies to ungauged basins of the Upper Blue Nile River Basin, Ethiopia. Journal of 495
Hydrology, 362: 39 – 52. 496
Lange J., N. Greenbaum, S. Husary, M. Ghanem, C. Leibundgut, A.P. Schick. 2003. Runoff generation 497
from successive simulated rainfalls on a rocky, semi-arid, Mediterranean hillslope. Hydrological 498
Processes, 17: 279-296. 499
Legesse D, Vallet-Coulomb C, Gasse F. 2003. Hydrological response of a catchment to climate and 500
lane use changes in tropical Africa: Case study south central Ethiopia. Journal of Hydrology, 501
275: 67-85. 502
Legesse, E.S., 2009. Modeling rainfall-runoff relationships for the Anjeni watershed in the Blue Nile 503
basin, Ethiopia. . Master Thesis, Integrated Watershed Management and Hydrology Program, 504
Cornell University, Bahir Dar Ethiopia and Ithaca, NY, USA. 505
Liu B.M., Abebe Y., McHugh O.V., Collick A.S., Gebrekidan B., Steenhuis T.S. 2008. Overcoming 506
limited information through participatory watershed management: Case study in Amhara, 507
Ethiopia. Physics and Chemistry of the Earth 33: 13–21. 508
McHugh O.V. 2006. Integrated water resources assessment and management in a drought-prone 509
watershed in the Ethiopian highlands. PhD dissertation, Department of Biological and 510
Environmental Engineering. Cornell University, Ithaca NY.511
Merz, J., P. M. Dangol, M. P. Dhakal, B. S. Dongol, G. Nakarmi and R. Weingartner. 2006. Rainfall-512
runoff events in a middle mountain catchment of Nepal. Journal of Hydrology, 331 (3-4): 446-513
458.514
22
Mohammed, A., Yohannes, F., Zeleke, G, 2004. Validation of agricultural non-point Source (AGNPS) 515
pollution model in Kori watershed, South Wollo, Ethiopia. International Journal of Applied 516
Earth Observation and Geoinformation 6, 97–109. 517
Setegn, S.G., Srinivasan, R., Dargahi, B. 2008: Hydrological Modeling in the Lake 518
Tana Basin, Ethiopia using SWAT model. The Open Hydrology Journal 2, 49-62. 519
Steenhuis T.S., Collick A.S., Easton Z.M., Leggesse E.S., Bayabil H.K., White E.D., Awulachew S.B., 520
Adgo E, Ahmed A.A. 2009. Predicting Discharge and Erosion for the Abay (Blue Nile) with a 521
Simple Model. Hydrological Processes 23: 3728–3737. 522
Sutcliffe J.V. and Parks Y. P. 1999. The Hydrology of the Nile. IAHS Special Publication 5. 180 pp 523
Todini, E., 1995. The ARNO rainfall-runoff model. Journal of Hydrology. 175, 339-382.524
Zeleke, Gete, 2000. Landscape Dynamics and Soil Erosion Process Modeling in the 525
North-western Ethiopian Highlands. African Studies Series A 16, Geographica 526
Bernensia, Berne. 527
528
23
529
Abbreviations 530
531
GPS Global Positioning System 532
SCRP Soil Conservation Research Program533
SCS Soil Conservation Service534
24
535
List of Tables and Figures 536
537Table 1: Location, description, and data used in the model from the three SCRP research sites. 538
539Table 2: Average soil infiltration rate at different soil types, slope range and land use types. 540
Table 3: Model input values for surface flow components for the three SCRP watersheds and the Nile 541River upstream of the border with Sudan. The watershed is divided into regions with different 542characteristics: (1) exposed bedrock and saturated areas that contribute surface runoff when the soil is 543saturated, or (2) hillsides that produce recharge when the soil is above field capacity. Maximum storage 544of water is the amount of water needed above wilting point to become either saturated (runoff 545contributing areas) or to reach field capacity (hill slopes). Model input values for the baseflow and 546interflow reservoirs are the maximum storage of the linear base flow reservoir; the time in days to 547reduce the volume of the baseflow reservoir to half under no recharge conditions, t*, is the duration of 548the period after a single rainstorm until interflow ceases. Nash-Sutcliffe efficiency for simulated daily 549average discharge for the three SCRP watersheds and 10-day average values for the Blue Nile Basin 550were calculated for the periods of calibration and validation.551
552Figure 1. Locations of the three SCRP watersheds in Ethiopia. 553
554Figure 2. Fourteen day discharge vs. effective precipitation in (1) the Anjeni watershed, and (2) all 555three SCRP watersheds above 500 mm effective precipitation (Liu et al 2008). 556
557Figure 3. Rainfall intensity exceedance probability for the Andit Tid watershed. 558
559Figure 4. Water table height measured in the Maybar watershed during the 2008 rainy season and start 560of dry season. a) Effect of slope on water table heights; b) Land use effects on the water table heights.561
562Figure 5. Conceptual model for predicting watershed discharge superimposed on the upper Andit Tid 563watershed.564
565Figure 6. Measured and modeled streamflow for the a) Maybar watershed, b) Andit Tid watershed and 566c) Anjeni watershed. The light grey line is the simulated discharge values and the tick black curve is 567observed values. The thin black line is the surface runoff.568
25
569
Table 1: Location, description, and data used in the model from the three SCRP research 570sites571
Researchsite
Location(Region)
Area(ha)
Elevation(masl)
Precipitation mm/year Length of data
AnditTid
39o43’E, 9o48’N(Shewa) 477.3 3,040 –
3,548 14671987 -2004
(1993, 1995 – 96 incomplete)
Anjeni 37o31’E, 10o40’N(Gojam) 113.4 2,407 -
2507 1675 1988 - 1997
Maybar 37o31’E, 10o40’N(South Wollo) 112.8 2,530 –
2,858 1417 1988 -2001 (1990 – 93 incomplete)
572573
Table 2: Average soil infiltration rate at different soil types, slope range and land use 574types 575
Testing Sites
Locationin the
Watershed
Average IR
(mm/hr) LocationSlope [°] Soil Type Soil Depth Land Use
3 top 25 15 andosol medium fallow grass 5 top 24 15 andosol medium fallow grass 8 top 594 29 andososl shallow bush Land
1 middle 226 30 regosol Shallow terraced and cultivated
2 middle 26 21 regosol deep terraced and cultivated
4 middle 140 21 andosol medium fallow
7 middle 29 21 humic
andososl shallow fallow 6 bottom 43 10 andosol deep fallow 9 bottom 53 2 eutric regosol medium cultivated
10 bottom 870 18 eutirc
cambisol medium terraced and cultivated
576
26
577
Table 3: Model input values for surface flow components for the three SCRP watersheds and the Nile 578upstream of the border with Sudan. The watershed is divided into regions with different characteristics: 579exposed bedrock and saturated areas that contribute surface runoff when the soil is saturated or 580hillsides that produce recharge when the soil is above field capacity. Maximum storage of water is the 581amount of water needed above wilting point to become either saturated (runoff contributing areas) or to 582reach field capacity (hill slopes). Model input values for the baseflow and interflow reservoirs are the 583maximum storage of the linear base flow reservoir; the time in days to reduce the volume of the 584baseflow reservoir to half under no recharge conditions, t* is the duration of the period after a single 585rainstorm until interflow ceases, Nash-Sutcliffe efficiency for simulated daily averaged discharge for 586the three SCRP watersheds and ten day average values for the Blue Nile Basin were calculated for the 587periods of calibration and validation. 588
Maybar Andit Tid
Anjeni Nile in1
EthiopiaType of source area
Valley Bottom Portion of area 0.23 0.1 0 0.1 Surface runoff Max storage, mm 110 90 - 250 Degraded Hill Portion of area 0.01 0.05 0.2 0.2 Surface runoff Max storage, mm 20 20 150 10 Hillside Portion of area 0.50 0.85 0.6 0.7 Interflow and
baseflow Max storage, mm 150 150 250 500 Max storage linear reservoir, mm 80 90 70 20 Half life linear reservoir, days 60 70 70 140 Drainage time hillsides, days 3 10 20 35 Nash Sutcliff 0.78 0.80 0.88 0.60 1 ten day intervals589
590
Q =
0.4
8P +
15.
7R
² = 0
.83
50100
150
200
rge, Q14-day(mm)
dry
seas
on
cum
. P-E
<=10
0
100<
(cum
. P-E
)<=3
00
300<
(cum
. P-E
)<=5
00
Line
ar (1
00<(
cum
. P-E
)<=3
00)
Line
ar (c
um. P
-E>5
00)
Q =
0.2
0P -
0.01
R2
= 0.
70
0 -100
010
020
030
040
0
Dischar
Effe
ctiv
e Pr
ecip
itatio
n, P
-E14
-day
(mm
)
b
All
3 w
ater
shed
sli
i
250
And
it Ti
dlin
ear r
egre
ssio
nQ
= 0.
52P
+ 18
.1R
2=
0.61
200
(mm)
Anj
eni
May
bar
Anj
eni &
May
bar
linea
r reg
ress
ion
Q=
049
P+
141
100
150
, Q14-day(
Q=
0.49
P+
14.1
R2
= 0.
80
50100
ischarge
050
Di
0-5
00
5010
015
020
025
030
035
0
Effe
ctiv
e Pr
ecip
itatio
n, P
-E14
-day
(mm
)
125
150
100
itymm/hr
5075
nfallinsensi
actual
storm
avaraged
25
rain
g
0 0.00
0.25
0.50
0.75
1.00
Exceed
ance
prob
ability
py
5101520 Water level (cm)
00000
8/11/2008
8/16/2008
8/21/2008
8/26/2008
8/31/2008
Gen
tle S
l
8/31/2008
9/5/2008
9/10/2008
ope
M
9/15/2008
9/20/2008 D
Mid
slo
pe
9/25/2008
9/30/2008
10/5/2008at
e
Ste
ep
10/5/2008
10/10/2008
10/15/2008
p sl
ope
10/20/2008
10/25/2008
Rai
nfal
l
0 1 2 3 4 5 6 7 8 9 1
10/30/2008
11/4/2008
(mm
)
0 10 20 30 40 50 60 70 80 90 100
Rainfall(mm/day)
1
020406080
100120140160180
wat
er le
vel (
cm)
8/11
/200
88/
14/2
008
8/17
/200
88/
20/2
008
G
8/23
/200
88/
26/2
008
8/29
/200
89/
1/20
089/
4/20
08
Grass land
9/4/
2008
9/7/
2008
9/10
/200
89/
13/2
008
9/16
/200
8
Crop lan
9/16
/200
89/
19/2
008
9/22
/200
89/
25/2
008
9/28
/200
8Days
nd Wood
10/1
/200
810
/4/2
008
10/7
/200
810
/10/
2008
10/1
3/20
08
d land R
10/1
3/20
0810
/16/
2008
10/1
9/20
0810
/22/
2008
10/2
5/20
08
Rainfall (mm0102030405060708090
10/2
5/20
0810
/28/
2008
10/3
1/20
08
m)
Rai
nfal
l (m
m/d
ay)
Hillslop
eHillslop
eAreas
infiltration
St
td
interflow
Saturated
Surfaceruno
ff
Expo
sedrock
Surfaceruno
ff
02040608010012014016010
1520253035404550
Rai
nfal
l mm
/day
Dis
char
ge m
m/d
ay N&S = 0.78
0204060801001201401601802000
5101520253035404550
1/1/
94
3/7/
94
5/11
/94
7/15
/94
9/18
/94
11/2
2/94
1/26
/95
4/1/
95
6/5/
95
8/9/
95
10/1
3/95
12/1
7/95
Rai
nfal
l mm
/day
Dis
char
ge m
m/d
ay N&S = 0.78
0 20 40 60 80 100
120
140
160
10152025303540
Rainfall mm/day
Discharge mm/day
N&
S =
0.8
8
0 20 40 60 80 100
120
140
160
180
200
0510152025303540
1/1/93
3/7/93
5/11/93
7/15/93
9/18/93
11/22/93
1/26/94
4/1/94
6/5/94
8/9/94
10/13/94
12/17/94
Rainfall mm/day
Discharge mm/day
N&
S =
0.8
8