karst geohazards in the uk: the use of digital data...

Karst geohazards in the UK: the use of digital data forhazard management

A.R. Farrant & A.H. CooperBritish Geological Survey, Keyworth, Nottingham NG12 5GG, UK (e-mail: [email protected])

An essential prerequisite for any engineering orhydrogeological investigation of soluble rocksis the identification and description of theircharacteristic, observable and detectable dis-

solution features, such as stream sinks, springs, sink-holes and caves. The British Geological Survey (BGS) iscreating a National Karst Database (NKD) that recordssuch features across the country. The database currentlycovers much of the region underlain by CarboniferousLimestone, the Chalk, and particularly the Permo-Triassicgypsum and halite where rapid, active dissolution hascaused significant subsidence and building damage. Inaddition to, and separate from, the identification ofspecific karst features, the BGS has created a NationalKarst Geohazard geographical information system (GIS).This has been created to identify areas that may poten-tially contain karst geohazards. Initially, all the solublerock units identified from the BGS 1:50 000 scale digitalgeological map are extracted. Each soluble unit has beengiven an objective score, interpreted, as based on factorsincluding lithology, topography, geomorphological pos-ition and characteristic superficial cover deposits. Thisnational zonation of these soluble rocks can then be usedto identify areas where the occurrence potential forkarstic features is significant, and where dissolution fea-tures might affect the stability of buildings and infrastruc-ture, or where karstic groundwater flow might occur. Bothdatasets are seen as invaluable scientific tools that havealready been widely used to support site investigation,groundwater investigations, planning, construction andthe insurance underwriting businesses.

Karst features, developed upon and within solublerocks, are a well-known potential geohazard, and cancause significant engineering problems, such as subsid-ence and irregular rockhead (‘top-of-rock’) surfaces.These can pose difficulties for planning and developmentand be very costly for the construction and insuranceindustries. There have been numerous examples ofsubsidence and infrastructure damage resulting fromunanticipated settlement and/or collapse of karst fea-tures (Waltham et al. 2005); in extreme cases they cancause properties to collapse and put lives at risk. Morecommonly, karstic rocks can make ground conditionsmore difficult, increasing design and construction costs.Underground cavities can also act as pathways alongwhich hazardous liquid and gaseous contaminants can

travel, commonly some distance from their source, thusposing another form of unanticipated environmentalrisk.

Databases and maps of karst hazards are importantfor understanding the severity of the problem, and theyconstitute useful tools for public governance choices inhazard avoidance that have relevance to planning, engi-neering, development and the insurance underwritingindustry. Developers, planners and local governmentcan be expected to operate effectively only if theyhave advance scientific warning about the hazards thatmight be present and have access to relevant geologicalinformation.

In the UK, karst is most typically associated with thelocally varied limestone successions of Early Carbonif-erous age, referred to informally as the CarboniferousLimestone, but karst features are also found in a host ofother carbonate and evaporite lithologies throughoutthe geological column (Fig. 1). The vast majority ofkarst features present in the UK are thought to be ofQuaternary age, although many have had a complexevolution over one or more glacial–interglacial cycles,

Fig. 1. Simplified map of the karstic rocks in the UK.

Quarterly Journal of Engineering Geology and Hydrogeology, 41, 339–356 1470-9236/08 $15.00 � 2008 Geological Society of London

and some caves are potentially much older. Triassicpalaeokarst is present in some areas (Simms 1990), but isgenerally of local extent. An overview of the UK karsthas been given by Waltham et al. (1997).

Palaeozoic and Neoproterozoic limestones

The oldest rocks that exhibit karstic features in the UKare the thin metacarbonate beds preserved within partsof the Dalradian Supergroup of the Scottish Highlands.The features here are perhaps analogous to theScandinavian stripe karst (Lauritzen 2001). Numeroussmall caves, sinkholes, stream sinks and springs havebeen recorded in the Appin and Schiehallion regions ofthe Scottish Highlands (Oldham 1975). However, manyof these are in remote upland areas and generally poselittle risk to infrastructure. Elsewhere in NW Scotland,Cambrian to Ordovician limestones and dolomitesbelonging to the Durness Group crop out in a longnarrow belt along the line of the Moine Thrust and onthe Isle of Skye. Several extensive well-developed karsticcave systems, for example the Allt nan Uamh system,and those in the Traligill valley have been described (e.g.Lawson 1988; Waltham et al. 1997), although againmany of these are in remote upland areas.

Farther south, minor dissolution features andenlarged fractures have been identified from limestonesof Wenlock (Silurian) age in the West Midlands and theWelsh Borders, but no significant cave systems are yetknown. Limestones of Devonian age crop out in parts ofDevon, particularly around Plymouth, Buckfastleighand Torbay. Several well-developed cave systems areknown in these areas (Oldham et al. 1986), along withstream sinks, ‘losing’ streams, karstic springs, sinkholes(dolines) and areas of irregular rockhead. Infrastructureproblems with karstic cavities and buried sinkholes havebeen reported in the Plymouth area.

It is the Carboniferous Limestone that hosts thebest-developed karst landscapes and the longest andmost-developed cave systems in the country. It occurswidely throughout western and northern England and inWales, and karst features are present on and within themajority of the outcrop (Waltham et al. 1997). Particu-larly well-developed karst occurs in the Mendip Hills,around the northern crop of the South Wales coalfield(Ford 1989), in the Derbyshire Peak District and in theYorkshire Dales and adjacent areas, running up into thenorthern Pennines. Less well-known karst areas includethe Forest of Dean, the southern crop of the SouthWales coalfield from Glamorgan through the Gower toPembrokeshire, in North Wales around the Vale ofClwyd, and around the fringes of the Lake District. Inall these areas, well-developed karstic drainage systems,sinkholes and extensive cave systems are common,many of which are associated with significant allogenicdrainage catchments as a source of the dissolutionwaters.

The major challenges associated with these karst areasare water supply protection (both quality and quantity),geological conservation and practical engineering prob-lems related to mitigating and accommodating theresulting subsurface voids. Subsidence associated withsinkhole formation is commonly encountered in remoteand rural areas with little impact on property andinfrastructure, but damage to houses, tracks and roads islocally a problem. Commonly of greater significance,many of these subsidence hollows are sites for illegaltipping of farm and other refuse or waste, which cancause rapid contamination of the groundwater and localdrinking supplies (Fig. 2).

Permian dolomites and limestones

The Permian carbonates in the UK are dominated bymagnesium-rich limestones and dolomites. The solu-bility of these rocks is characteristically lower than thatof the pure limestones, and so karstic features are lesswell developed in this terrain. However, the dolomitesare closely associated with bodies of gypsum, whichtypically are heavily karstified. Numerous small (gener-ally less than a few tens of metres in length) cave systemsare present along the outcrop of the magnesian lime-stones from near Mansfield in the south to Sunderlandin the north. Some sinking streams are present as arenumerous springs, but very few sinkholes are to befound (possibly because of infilling by agricultural prac-tices over the centuries) and the rock is generally notproblematical for engineering purposes. However,numerous open joints, incipient conduit systems on bed-ding planes, palaeokarst, and sediment-infilled fissurescan be identified in road cuttings and quarries. Hydro-geologically, these rocks support a minor to moderate-sized aquifer, potentially susceptible to resource damagefrom pollution and land development, especially fromlandfill contamination from former infilled quarries.

In the Mendips, South Wales and parts ofDevon, limestone-rich Permo-Triassic conglomerates

Fig. 2. A sinkhole formed after severe flooding in 1968 nearCheddar, Somerset [ST 4765 5620]. Superficial loessic coversands have slumped into the underlying GB Cavern. Anattempt was made to fill in the hole with old cars, which canstill be seen in the cave below. Photo A. Farrant, copyrightNERC.

A.R. FARRANT & A.H. COOPER340

(predominantly, but not exclusively derived fromCarboniferous limestone uplands) can be expected tohost karstic cave systems, perhaps the most famousexample being Wookey Hole in Somerset.

Jurassic limestones

Jurassic limestones extend across much of central,southern and eastern England in a belt from Dorset toNorth Yorkshire, and in parts of western Scotland.Although the karst in these rocks is not as well devel-oped as in the more massive Palaeozoic limestoneterrain, karst features do occur, particularly in some ofthe Portlandian and Purbeck limestones in Dorset andWiltshire, in the Cornbrash, and in the Corallian lime-stones around Oxford and on the southern flank of theNorth York Moors, where a cave system >1 km longhas recently been discovered. Karst is also well devel-oped in the Lincolnshire Limestone south of Grantham(Hindley 1965). Here, water supply protection is themain issue, although localized subsidence and areas ofirregular rock-head are to be expected.

Cretaceous Chalk

The Chalk is the most widespread carbonate rock in thecountry and of immense importance for its water supplyresource. It forms the UK’s most important aquifer.Karst features occur in many parts of the Chalk out-crop, particularly along the margin of the Palaeogenecover (Farrant 2001). In these areas, the development ofdissolutionally enlarged fissures and conduits can poten-tially cause problems for groundwater supply by creat-ing rapid-transport contaminant pathways though theaquifer (Maurice et al. 2006). This is particularly import-ant as parts of the Chalk outcrop underlie major trans-port corridors and urban areas; both of which arenotorious suppliers of surface-gathered contaminantsvia runoff. Chalk dissolution also generates subsidencehazards and presents difficult ground conditions associ-ated with the development of clay-filled pipes and fis-sures. These problems include irregular rockhead,localized subsidence, increased mass compressibility anddiminished rock mass quality. Well-developed karsticgroundwater flow systems occur in Dorset, near Salis-bury, around Newbury and Hungerford, and in manyparts of the Chilterns, particularly along the Palaeogenemargin between Beaconsfield and Hertford (Walthamet al. 1997).

Triassic and Permian salt

In the UK, halite (rock salt) occurs mainly within theTriassic strata of the Cheshire Basin, and to a lesserextent in parts of Lancashire, Worcestershire andStaffordshire. It is also present in the Permian rocks of

NE England and Northern Ireland (Notholt & Highley1973). Where the saliferous Triassic rocks come tooutcrop, most of the halite has dissolved and the over-lying and interbedded strata can be expected to havecollapsed or foundered, producing a buried salt karsthazard. This represents a more difficult site classificationproblem than perhaps elsewhere. These areas commonlyhave saline springs, indicative of continuing salt dis-solution and the active nature of the karst processes. Thesalt deposits were exploited using these springs frombefore early Roman times to Victorian times, when moreintense and organized brine and mined salt extractionwas undertaken. Halite dissolves rapidly and subsidence,both natural and mining-induced, has affected the mainTriassic salt fields including Cheshire, Staffordshire(Stafford), Worcestershire (Droitwich), coastalLancashire (Preesall) and parts of Northern Ireland.Much of the salt mining has been by shallow brineextraction, the results of which also mimic the effects ofnatural karstification of the salt deposits.

Permian salt is present at depth beneath coastalYorkshire and Teesside. Here the salt deposits and thekarstification processes are much deeper than in theTriassic salt. The salt deposits are bounded up-dip by adissolution front and collapse monocline (Cooper 2002).The depth of the salt dissolution means that very fewkarstic features are present, either as formed at or asmigrated upward to the present ground surface,although some salt mining (brining) subsidence hasoccurred.

Triassic and Permian gypsum

Compared with the British limestone karst, gypsumkarst occurs only in relatively small areas. It is presentmainly in a belt 3 km wide and about 100 km long in thePermian rocks of eastern and northeastern England(Cooper 1989). Significant thicknesses of gypsum alsooccur along the eastern flank of the Vale of Eden (Ryder& Cooper 1993). It also locally occurs in the Triassicstrata (Cooper & Saunders 1999), but the effects ofkarstification are much less severe than in the Permianrocks. The difference is mainly caused by the thicknessof gypsum in the Permian sequence and the fact thatit has interbedded dolomite aquifers. In contrast, theTriassic gypsum is present mainly in weakly permeablemudstone sequences. The gypsum karst has formedphreatic cave systems dissolved by water flow below thepiezometric surface to depths of up to 70–80 m. Therapid solubility rate of the gypsum (Klimchouk et al.1997) means that the karst is evolving on a humantime scale. Active subsidence occurs in many places,especially around the town of Ripon (Cooper 1986,1989, 1998) and to a lesser extent in the eastern part ofurban Darlington (Lamont-Black et al. 2002); it alsooccurs in several other locations along the outcropforming a belt about 100 km long and 3–5 km wide. The

KARST GEOHAZARDS IN THE UK 341

active nature of the dissolution and these continuingsubsidence features cause difficult ground conditions forplanning and development (Cooper & Saunders 1999;Paukstys et al. 1999; Cooper & Saunders 2002). Gypsumpalaeokarst features also occur, especially along thecoast of NE England and in the Firth of Forth offeastern Scotland (Cooper 1997).

The National Karst Database (NKD)

Geologists and engineers have recognized for some timethat the availability of geoscience baseline data is essen-tial for the public safety and engineering assessment ofgeological hazards. Understanding the severity of theproblem is a prerequisite for hazard avoidance. Landdevelopers, planners and local government can planaccordingly and responsibly if they can cite advancescientific–technical warning about the hazards thatmight be present and have access to relevant geologicalinformation. General guidance for the development ofunstable land is written into British Government plan-ning policy in the Planning Policy Guidance Note 14:Development on Unstable Land (Department of theEnvironment 1990), and the supplementary Annex 2(Department of Transport, Local Government & theRegions 2002).

When this policy was implemented, rudimentary base-line data were collected in an initial database of naturalcavities commissioned by the Department of theEnvironment and produced by a private geologicalconsultancy, Applied Geology Ltd (1993). This is nowheld, and maintained as a legacy dataset, by the BritishGeological Survey (BGS), and is offered as a technicalservice by Peter Brett Associates, a private geologicalconsultancy. This study demonstrated the nationalcharacter and distribution of karst and other naturalcavities, based on documentary evidence from a widevariety of sources. However, in many cases the spatialrecording was not very accurate, commonly only to thenearest kilometre, and in many cases was based upononly the barest minimum of supporting data. Further-more, the database was over-complex and many of thedatabase fields remained empty. At the time the data-base fields were recognized as essential to the bestunderstanding of mitigative prediction, and the decisionwas made to present the categorical need for such data,even if the bare minimum was not then available. Sincethat time, recent mapping, particularly in Chalk terrains,has shown that the previously available data signifi-cantly under-represents the actual density of karst fea-tures now generally known to be present. Moreover,many more karst features have been identified since1993, both by more recent geological mapping, but alsothrough the advent of more sophisticated remote sensingtechniques such as LiDAR (light detection and ranging).Cavers have also opened up many newly discovered cave

systems, including one of Britain’s longest cave systems,Ogof Draenen near Blaenavon (Farrant 2004), nowknown to be over 70 km in length, and also havediscovered the country’s deepest natural vertical cavern,the 145 m deep Titan shaft in Derbyshire’s Peak Cavern.Documented speleological studies of cave systems canprovide useful insights into the potential for karst geo-hazards to develop in the surrounding area (Klimchouk& Lowe 2002).

Much of the information in the relict AppliedGeology dataset was extracted directly from publishedBGS maps, memoirs and reports. However, additionalinformation now is available for systematic integration,as represented by historical unpublished geological fieldslips and notebooks held in the National GeologicalData Centre in Keyworth. Karst features have beenrecorded routinely by field geologists on their field slipssince the Survey’s inception in 1835. However, they haveonly rarely been depicted on published map and reports,and the data contained on them can be difficult to access.

In 2000, as part of a drive to make its datasets moreaccessible for the common good, the BGS embarked onconstructing a more comprehensive geographic infor-mation system (GIS) and database of karst information(Cooper et al. 2001), from which the National KarstDatabase was set up. The aim was to retrieve the karstdata contained in the field slips and on paper maps heldin the archives, and make them accessible in a GISformat. The BGS field slips and maps have now beenscanned and many of them geo-rectified so that they canbe viewed via an in-house customized GIS system. Anexample of a historical field slip from Lincolnshire isshown in Figure 3.

Additional fine-scale karst information is now beingroutinely gathered both in the field and from existing

Fig. 3. Field geologist’s field slip [OLL 131 NW (E)], compiledby F. B. A. Welch, surveyed in 1941, describing a stream sinkin the Upper Lincolnshire Limestone Member (Jurassic), nearBurton-le-Coggles [SK 9683 2595], and its response to a flood.The railway line is the East Coast Main Line.


documentary data sources. The information gatheredduring fieldwork is recorded either digitally on portabletablet computers or on pro forma field data sheetsorganized with the same datafields as the GIS andits underlying database. Documentary information isgathered from existing datasets such as scanned andgeo-rectified copies of the geologists’ field maps, histori-cal and modern geo-rectified Ordnance Survey maps,cave surveys, academic papers and historical documents.The data are systematically entered into a customizedGIS system, initially achieved using ArcView3 (Cooperet al. 2001), but now using ArcGIS9. The point-specificinformation and database tables are then copied tocentralized Oracle databases. Other datasets such asthe Applied Geology database and the Mendip CaveRegistry data are being used to identify other sources ofinformation where relevant.

Data on five main types of karst feature are collected:sinkholes (or dolines), stream sinks, caves, springs, andincidences of associated damage to buildings, roads,bridges and other engineered works. The details gath-ered are listed in Tables 1–5 (for sinkholes, springs,

stream sinks, caves and building damage, respectively).For each of these datasets, in common with all BGSdatabases, the information added to the systemhas common header data including National Gridcoordinates, date entered, user number and reliability.

For sinkholes, the data can be entered either as apolygon covering the known extent of the feature, wherespatial data exist, or as a point, where the sinkhole is toosmall to be represented or if no spatial data are avail-able. The type description of sinkhole is adapted fromthe classification of Culshaw & Waltham (1987). Thesize of springs and stream sinks is necessarily subjective,depending on the time of year and rainfall, but a proxyfor the average flow can be obtained from the observedchannel width and depth (of both the influent andeffluent channels). For the majority of historical infor-mation gathered from published maps and geologists’field maps and slips, no precise description of spring orstream sink flow is available for direct entry to theNKD. In many cases, some of the dataset entries can beestimated or obtained from other data sources. Infor-mation gathered for caves is also collected as eitherpoint data for cave entrances, or, if they are known, aslinear data for the approximate centre lines of the caves

Table 1. Datafields gathered for sinkholes

Sinkholes, record item Parameters

Sinkhole name Free textSize Size x, Size y, Size z, metresType Compound, collapse,

suffusion, solution, no data,buried

Shape Round, oval, irregular,modified, compound, nodata

Surface profile Pipe, cone, inverted cone,saucer, complex, levelled(filled), no data

Infill deposits British Geological Surveyrock and stratigraphicalcodes with thicknesses

Subsidence type Gradual, episodic,instantaneous, no data

Evidence of quarrying Yes, no, no dataPrimary data source Field mapping, air-photo,

site investigation, database,maps and surveys,literature, Lidar remotesensing, DoE database, nodata

Reliability Good, probable, poor, nodata

Property damage Yes, no, no dataOldest recorded subsidence dd/mm/yyyyIntermediate subsidence events dd/mm/yyyyMost recent subsidence dd/mm/yyyyOther data dd/mm/yyyyReferences Free text

DoE, Department of the Environment.

Table 2. Datafields gathered for springs

Springs, record item Parameters

Spring name Free textElevation MetresSituation Open surface, borehole, concealed,

submerged, submarine, undergroundinlet, no data

Proven dye trace Yes, noFlow Ephemeral, fluctuating, constant,

flood overflow, ebbing and flowing,no data

Water type Normal–fresh, saline, sulphate,tufaceous, other mineral, no data

Size Trickle, small stream, medium stream,large stream, small river, mediumriver, large river, no data

Primary data source Field mapping, air-photo, siteinvestigation, database, maps andsurveys, literature, Lidar remotesensing, DoE database, no data

Artesian Yes, no, no dataThermal Yes, no, no dataKarstic Yes, no, no dataUses None, public, agricultural, industrial,

other, no dataCharacter Single discrete, multiple discrete,

diffuse, no dataReliability Good, probable, poor, no dataEstimated discharge Litres per second (l s�1)Other data dd/mm/yyyyReferences Free text


themselves. It is possible to include full cave surveys,but there are often copyright issues that need to beaddressed before these data entries can be included. Inthese situations, the cave location is documented, and

reference to the survey made in the reference table whereappropriate. Many sites are multiple features; forexample, many stream sinks are also cave entrances,whereas sinkholes might also be sites of building orinfrastructure damage.

Since the database was set up in 2002, much infor-mation has been added to the system. Data coveringmost of the evaporite karst areas of the UK have nowbeen added, along with data covering about 60% of theChalk, and 30% of the Carboniferous Limestone out-crops. To date, over 800 caves, 1300 stream sinks, 5600springs and 10 000 sinkholes have been recorded, andmany of the classic upland karst areas have yet to beincluded.

Other karst datasets

Other karst datasets are available in the UK, most ofwhich have been collated by members of the cavingcommunity. These data generally cover the more well-known Carboniferous and Devonian limestone karstareas, which contain most of the accessible caves.

The British Caving Association (BCA) maintains aregistry of karstic sites in the form of a National CaveRegistry. It holds limited information about eachsite with links to more comprehensive information inregional cave registries maintained by regional cavingorganizations, including the Mendip Cave Registryand the Cambrian Cave Registry. The Mendip CaveRegistry is manned by volunteer registrars who form themembership of the group. Its aim is to record allavailable information relating to the caves and stonemines in the counties of Bristol, Somerset and Wiltshire.It has around 8000 documented sites of speleologicalinterest. These include caves, stream sinks, springs andsites where cavers have dug in their efforts to discovernew caves. The Cambrian Cave Registry is currently

Table 3. Datafields gathered for stream sinks

Stream sinks;record item

Parameters

Sink name Free textElevation MetresProven dye traces Yes, no, no dataMorphology Discrete compound, discrete single

sink, diffuse sink, losing stream,ponded sink, cave entrance,concealed sink, no data

Flow Perennial, intermittent, ephemeral(flood), estavelle, no data

Size Trickle, small stream, mediumstream, large stream, small river,medium river, large river, no data

Primary data source Field mapping, air-photo, siteinvestigation, database, maps andsurveys, literature, Lidar remotesensing, DoE database, no data

Reliability Good, probable, poor, no dataEstimated discharge Litres per second (l s�1)Other data Free textReferences Free text

Table 4. Datafields gathered for natural cavities

Natural cavities,record item

Parameters

Cavity name Free textLength MetresVertical range MetresElevation MetresType Open cave natural, infilled cave

natural, gull cave, lava tube,boulder, peat cave, sea cave,stoping cavity, palaeokarst,hydrothermal borehole cavity,no data

Rock units penetrated(bedrock and superficial)

British Geological Survey rockand stratigraphical codes

Primary data source Field mapping, air-photo, siteinvestigation, database, mapsand surveys, literature, Lidarremote sensing, DoE database,no data

Streamway Yes, no, no dataOther entrance Yes, no, no dataEvidence of mining Yes, no, no dataReliability Good, probable, poor, no dataOther data Free textReferences Free text

Table 5. Datafields gathered for property damage

Property damage,record item

Parameters

Address Free textPostcode Postcode formatElevation MetresDamage survey 1 Date (dd/mm/yyyy), notes, damage

rating (1–7)Damage survey 2 Date (dd/mm/yyyy), notes, damage

rating (1–7)Damage survey 3 Date (dd/mm/yyyy), notes, damage

rating (1–7)Suspected cause Natural subsidence, mining subsidence,

landslip, compressible fill, building defectReliability Good, probable, poor, no dataOther data Free textReferences Free text


being digitized and is scheduled to be available on-linesometime in 2008. The Black Mountain Cave Register,part of the Cambrian Cave Registry, used to be accessi-ble on-line, but is no longer available. The regionalcaving councils, supported by the BCA, also hold data-bases containing information about caves and accessarrangements.

The Cave Database (http://www.cavedatabase.co.uk/)is an open access website to which anyone can addcave-related links. However, its coverage is patchy, andthe data are not complete. In addition, the website maybe at risk of being compromised by the gratuitousinclusion of links that are not cave related, or data thathave not been checked.

CAPRA, the Cave Archaeology and PalaeontologyResearch Archive based in the University of Sheffield,has compiled a gazetteer of English, Scottish and Welshcaves, fissures and rock-shelters that contain humanremains. This is available on the CAPRA website(Chamberlain & Williams 1999, 2000a, b).

The Chelsea Spelaeological Society has compiled adataset of natural and man-made cavities in SEEngland. These have been published as a series ofpaperback publications (some co-authored with theKent Underground Research Group) that are availablefrom the Society. Most of the data relate to dene-holesand Chalk mines, but some natural chalk caves areincluded, such as those at Beachy Head in Sussex(Waltham et al. 1997).

Many of the lesser known karst areas, including muchof the Chalk outcrop, are not covered by these schemes.These areas still contain significant densities of karstgeohazards, and are commonly located in more denselypopulated areas. Moreover, a significant drawback ofthese databases is that they only provide information onknown karst features. Predicting the whereabouts ofkarst features such as buried sinkholes and undiscoverednear-surface cavities is more difficult. Combining avail-able baseline karst data with other digital spatial data,such as bedrock and superficial geology, superficialdeposit thickness and digital terrain models, within aGIS can provide a powerful tool for predicting karstgeohazards.

National Karst Geohazard GIS(GeoSure)

Over the past decade, the BGS has invested considerableresources in the production of digital geological mapdata for the UK. Digital geological maps (DiGMap) arenow available for most of the country at the 1:50 000scale, and complete coverage at the 1:250 000 and1:625 000 scales. A significant part of the country is alsocovered at the 1:10 000 scale. All these datasets includethe bedrock, and the 1:625 000, 1:50 000 and 1:10 000scale datasets also include data for the superficial

deposits. Each polygon of digital geological data isattributed with a two-part code (the LEX-ROCK code)that gives its stratigraphy and its lithology. For example,the Seaford Chalk Formation is represented by the codeSECK_CHLK, whereas the Carboniferous Gully OoliteFormation has the code GUO_OOLM. A comprehen-sive list of all the LEX codes used can be found in theBGS Lexicon of Named Rock Units, which is availableon the Internet at http://www.bgs.ac.uk/lexicon/ lexicon_intro.html. Here, each named unit will eventually bedefined and its upper and lower boundaries described.The lithological codes (ROCK) are based on the BGSRock Classification Scheme, and are also explained andlisted on the Internet at http://www.bgs.ac.uk/data/dictionaries.html and can be searched by name or code.

The availability of the digital geological maps hasallowed the BGS to produce a set of predictive data-sets for geological hazards, known and marketed as‘GeoSure’. Several derived datasets have been producedusing a variety of algorithms to provide geohazarddata for soluble rocks (dissolution), landslides (slopeinstability), compressible ground, collapsible rocks,shrink–swell deposits and running sand.

The GeoSure dissolution dataset can be used toidentify areas where there is potential for a range ofkarstic features to develop. Such features include poten-tial geohazards such as sinkholes, dissolution pipes,natural cavities and areas prone to dissolution-generatedirregular rockhead, as well as hydrological features suchas stream sinks, estavelles and karstic springs. It can alsobe used to highlight areas where karstic groundwaterflow might be significant to the public interest. Thisis a distinct, separate dataset that complements andvalidates the NKD.

The first iterations of the GeoSure dissolution datasetused the bedrock codes as the basis for a purelylithology-based system that was enhanced by localknowledge and manual subdivision. More recently, thetechnique used to create the GeoSure GIS has beenupgraded, and now employs a scoring system in a GISenvironment using a range of additional datasets. Thishas several advantages over the purely lithology-basedsystem. It allows greater flexibility in rating hazardlevels, and can give a more precise hazard rating for agiven area of actual ground being considered for somepublic purpose. It allows differential weighting to begiven to different controlling factors depending on theunderlying karstic lithology. It is a more robust systemand can be tailored to the quality of the input data.

Methods

The GeoSure dissolution layer is but one of five layers ofgeoscience data that can be brought into considerationfor general and specific land classification, planning anddevelopment purposes. The dissolution layer has been


created by identifying a suite of factors that can influ-ence the style and degree of karst features likely todevelop at any one place. Each of these factors, such asbedrock lithology, geomorphological domain and super-ficial deposit thickness is represented in the GIS as aseparate theme. Each of these themes is then categorizedand given a score, to give an indication of the contribu-tion it might make to the overall degree of soluble rockhazard, in a similar manner to the more detailed site-specific method for the Chalk proposed by Edmonds(2001). Those with a strong influence will have a highscore, those with a slight influence will have a low score.Some factors, such as very thick superficial depositcover, might even have a negative score if they signifi-cantly reduce the particular hazard or damage potentialfor karst features to develop.

The karst geohazard layer is created in the GISsoftware by intersecting all the datasets into one, creat-ing a mosaic of intersected polygons. For each resultingpolygon, the sum total of each of the contributingfactors is then used to give an overall potential hazardscore. These scores can then be categorized to give ahazard rating (A–E) for a particular area. The hazardrating categories can be tailored to suit the quantitativeneeds of different end-users; for example, a high hazardrating important for those planning a tunnelling schememight not be pertinent for a home owner. The geologicalmanifestations of carbonate karst compared with thoseof gypsum and salt karsts are different, so the evaporiteshave a slightly different set of determinative parametersand are thus software-calculated separately. In addition,the actual scores used for each factor can be updatedwhen new or more detailed baseline information, such asa new version of the digital geological map, becomesavailable.

Bedrock lithology

The type, style and scale of karst geohazard is stronglyinfluenced by its intrinsic bedrock lithology, which cancommonly be used as a proxy for a particular rockunit’s overall mass fracture properties and hydrologicalcharacteristics.

All the known soluble and karstic rocks (with theexception of units containing gypsum and halite, whichare described below) were extracted from the BGSDiGMap 1:50 000 scale digital geological databaseand grouped on the basis of their LEX-ROCK codesinto four broad lithological groupings: Cretaceouschalks; Palaeozoic limestones (including Neoproterozoicmetacarbonates); Jurassic limestones and oolites, andTriassic conglomerates and Permian dolomites.

These broad groupings reflect the different style ofkarst landscape evolution that these rock units support.For example, the style and type of karst features devel-oped on the Chalk is very different from that developedon the Carboniferous Limestone.

Each of the 480 geological units extracted from theDiGMap database was then given a score based on itsLEX-ROCK code. Those rock units that either areknown to be, or have the potential to be, highly karstic,such as some of the Carboniferous Limestone units, aregiven a high score, whereas those rock units that are lesskarstic, but still within the range of ‘soluble’ rocks, suchas some of the Jurassic oolites, have a lower score. Therest of the guiding factor datasets were then cut againstthis modified bedrock layer to create the final hazardlayer specific to the karst areas of the UK. The scoresused to weight each of the guiding factors can beadjusted slightly according to the lithological groupingsto reflect the style and degree of karstification, as well ason a local basis, where and if new, project or use-relateddata are available.

In addition, a set of typically non-soluble lithologiesthat are known to host karstic features were alsoextracted. In areas of interstratal karst, the overlyingcover rocks are liable to subside into cavities in theunderlying limestone, creating cover-collapse sinkholes.The Twrch Sandstone Formation (part of the MarrosGroup, which replaces the term Millstone Grit in SouthWales) is a classic example of a non-soluble lithologythat locally hosts sinkholes and stream sinks. Somemapped formations contain both soluble and non-soluble lithologies, and this is reflected in the GIS by theoverall score given to the particular formation. Moredetailed geological datasets may allow further subdivi-sion of the individual soluble or non-soluble lithologieswithin a particular formation.

Superficial deposit domains

The superficial geology of an area has a strong influenceon the development potential of karst features. Areasunderlying only a thin cover of superficial depositscommonly have a greater incidence of near-surface karstfeatures such as suffosion sinkholes. This situation iscaused mainly by concentration of surface runoff on thecover strata, thereby generating discrete point rechargeinto the susceptible underlying bedrock and, ultimately,concentrating dissolution. The unconsolidated depositscan also slump and ravel into dissolutionally enlargedfissures in the underlying limestone, creating dissolutionpipes, cavities and, ultimately, sinkholes. These types ofsinkhole are common in areas with a thin cover ofglacial deposits overlying the Carboniferous Limestone,as in the Yorkshire Dales, or areas of clay-with-flintsover the Chalk in southern England.

Details of all the superficial deposits that overliesoluble bedrock polygons were extracted from the BGSDiGMap 1:50 000 scale digital geological database andcut against the soluble bedrock polygons. The superficialpolygons were then grouped into a set of super-ficial deposit domains that reflect their genetic origins(Table 6), based on LEX-ROCK codes. Each of these


domains is given a score. In the case of marine super-ficial deposits the score is negative, which reflects thelower potential for karst to develop in areas covered bymarine deposits such as estuarine alluvium.

The accuracy of both the solid and superficial poly-gons is based on the baseline geological data, which inturn are dependent on the quality and quantity ofexposure, and the antiquity of the original geologicalmapping.

Superficial thickness

Each superficial unit will also have a score dependingupon how thick it is. Areas with a very thin superficialcover will have a high score, reflecting the greaterlikelihood of point recharge effects and their subsequentinfluence on sinkhole formation, with lower scoresobtaining where the superficial units are relatively thick.For superficial deposit thicknesses greater than a speci-fied limiting value the score will become negative,reflecting their negative influence on the development ofsinkholes. This value is based on observations of thedensity and size of known sinkholes in areas with asuperficial cover. The influence of superficial deposits onthe development of karst features is clearly demon-strated in the Yorkshire Dales. Here, limestone areaswith a thick, dense, and low-permeability glacial tillcover (for example, beneath the boulder clay of drum-lins), generally have very few sinkholes (although older,relict sediment-filled palaeokarst features may be pre-served beneath the basal till), whereas areas with a thinsuperficial cover are commonly pockmarked with suffo-sion sinkholes. The superficial thickness is taken from

the BGS Superficial Deposits Thickness Model, dividedinto six categories, and scored accordingly.

Superficial permeability

The permeability of the superficial deposit also has adefinite controlling effect on the likelihood of dissolutionfeatures developing in the underlying bedrock. Highlypermeable units such as sandy river terrace depositshave a higher score than lower-permeability clay-richdeposits. This reflects the ease with which the water canflow through the deposit, creating an irregular rockheadand dissolution pipes pocking the bedrock interface. Thepermeability information is derived from the existingBGS Superficial Permeability database and has beendivided into four categories and scored accordingly.

Slope angle

A nationwide digital slope model has been created usingNextMap data. The slope angles have been cate-gorized into a range of values, cut against the bedrocktheme and scored accordingly. Flat or gently slopingareas will have the highest score, principally because it isin these areas in the UK that sinkholes, dissolution pipesand areas of irregular rockhead are known to be mostlikely to develop. On steeper terrains, slope processestend to erode karst features rather than enhance them.

Glacial limit

The limits of the extents of the Devensian and Anglianglaciations have exerted only a minor degree of influenceon the magnitude of karstification in the UK. In areassouth of the glacial limits, many relict karst featuresoccur, preserved on the landscape. In contrast, areaswithin the limit of the last Devensian glaciation havebeen subjected to locally extensive erosion and manykarstic features developed in the Pleistocene interglacialshave been eroded away. This dataset divides the countryinto polygons defined by the various recognizable glaciallimits (Anglian, Devensian and Loch Lomond). Areasoutside the maximum glacial limit will have the highestscore as they will be most weathered.

Expert GIS polygons

In addition to the polygons derived from existing digitaldatasets, three new datasets characterizing the relation-ship between a soluble rock and adjacent impermeablestrata were created (Fig. 4). These three polygon setshave been created manually using digital geological andtopographical maps, cave surveys, detailed local knowl-edge, the National Karst Database and overviews ofkarst geomorphology in general.

Table 6. Superficial deposit domain groupings

Domain Characteristics

Unmantled domain Areas with no superficialcover, or where the cover hasbeen eroded away

Marine domain All marine and intertidalsuperficial deposits

Raised beach domain All types of raised beachdeposits

River valley domain Fluvial deposits such asalluvium, valley peat andvalley gravels

River terrace domain All types of river terracedeposits

Weathered mantled domain Residual and/or weatheringdeposits such as clay-with-flints, loessic deposits andcover-sands, and hill peat

Dry valley domain Head, coombe and valleygravels

Glacial domain Tills and glacial morainedeposits


Runoff margin

The greatest concentrations of potentially hazardous,surface karst features are commonly associated withcontact margins between a relatively impermeable rockbody and one or more soluble lithologies, creating a lineof infusion along which surficial runoff is directed ontothe soluble rock mass. There are several broad types ofsituation where this relationship can arise. It occurswhere younger sandstones or mudstones overlie thesoluble lithology; for example, where the PalaeogeneReading Formation overlies the Chalk in southernEngland, or where the Twrch Sandstone Formationoverlies the Carboniferous Limestone in South Wales. Italso occurs in dipping strata where older rocks form thehigher ground; a classic example is the contact betweenthe Avon Group (Lower Limestone Shales) and theCarboniferous Black Rock Limestone Formation in theMendips. Alternatively, a similar karst contact mightarise as a result of faulting or an unconformity.

This zone of enhanced surface karst is represented inthe GIS by a manually picked polygon along the marginbetween a non-karstic rock and a karstic unit where thenon-karstic rock is topographically higher. The relativeaccuracy of the zone is partially dependent on theunderlying geological data. Although depicted as adistinct zone in the GIS, in reality its margins are likelyto be gradational.

Feather edge

Karstic features also occur in non-karstic rocks thatoverlie a soluble lithology. Examples include the numer-

ous, large, well-developed sinkholes (Fig. 5) devel-oped on the Twrch Sandstone Formation on MynyddLangynidr in South Wales (Waltham et al. 1997). Theseare caused by the collapse of the overlying sandstoneand mudstone into cavities in the underlying Carbonif-erous Limestone, and they can develop through a con-siderable thickness (perhaps as much as 30 m) of coverstrata. Similar, smaller features occur in the Reading

Fig. 4. A schematic cross-section through the north crop of the South Wales coalfield, showing the relationships between the runoff

margin and the feather edge zones along the contact between relatively soluble and insoluble rocks. An area of interstratal karstoccurs where the Twrch Sandstone overlies karstified Carboniferous Limestone.

Fig. 5. A large sinkhole and stream sink developed on theSouth Wales Lower Coal Measures Formation, MynyddLlangattwg, South Wales [SO 1766 1527]. Subsidence here hasbeen caused by the collapse of the Carboniferous Limestone atdepth. The collapse column has migrated up more than 40 mthrough the overlying Twrch Sandstone Formation into the‘Coal Measures’. The water reappears in the Cascade Inlet inthe 31 km long Agen Allwedd cave system, over 200 m belowthe surface. Photo A. Farrant, copyright NERC.


and Thanet Sand Formations overlying the Chalk(Edmonds 1983). Karst features such as sinkholes,stream sinks and dissolution pipes are particularly com-mon where the cover strata thin to a feather edge at thecontact with the underlying soluble strata.

This zone is also represented in the GIS by a manuallypicked polygon along the ‘feather-edge’ of the imper-meable deposit. This zone will be given a high score, asthe bedrock lithology is generally insoluble and wouldotherwise have a very low score; for example, theReading Formation on the Chalk. As with other manu-ally picked lines, the relative accuracy of the zone ispartially dependent on the underlying geologicaldata, and, again, in reality its margins are likely to begradational.

Interstratal karst

In parts of South Wales, the Forest of Dean and theMendip Hills, significant areas of interstratal karstoccur. This is where karstic drainage systems havedeveloped in the Carboniferous Limestone below asignificant thickness of impermeable rocks such as theTwrch Sandstone Formation or the Cromhall SandstoneFormation in the Forest of Dean. In this scenario, theremight be little or no surface expression of karst, butextensive caves or karstic groundwater flow systems canoccur at depth (sometimes over 100 m below the sur-face), such as Ogof Draenen near Abergavenny (Farrant2004), and the Wet Sink–Slaughter Risings system nearJoyford in the Forest of Dean (Waltham et al. 1997).Areas of significant interstratal karst have beenidentified and digitized, making use of geological andtopographical maps, groundwater tracing experimentresults, cave surveys and assistance of expert localknowledge.

All the above datasets are intersected with each otherand the scores for each summed to create the finalhazard layer. For example, a flat area of the Carbonif-erous Black Rock Limestone situated adjacent to anarea of impermeable strata within the runoff area, withno superficial deposits in the Mendip Hills would scorehighly, and be given an E hazard rating. In contrast,an area of bare Carboniferous Black Rock Limestoneon a moderate slope would generate a more modestscore corresponding to a C rating (Fig. 6). An area onthe feather edge of Twrch Sandstone overlying theCarboniferous Limestone scores highly, putting it in a Dhazard category. Similarly, an area of gently slopingSeaford Chalk with a thin cover of clay-with-flintswould also score highly and be given a D rating, whereasan area of New Pit Chalk on a scarp slope wouldgenerate a low score, and be given an A rating. It isimportant to note that the relative scale, style and typeof karst features present in each area may be different.Some smaller scale, less obvious karst features such aschalk dissolution pipes may pose more of an engineering

hazard than some larger karst features such as deepcaves.

Figure 6a and b shows an example from the MendipHills in Somerset where a mature karst landscape hasdeveloped on the Carboniferous Limestone. The greatestdensity of karst features predictably occurs along themargin of the Avon Group mudstone and within a thincover of Mesozoic strata capping the limestone, butsome lesser degree of surface karst should also be notedto occur across this zone, and within the Triassic MerciaMudstone Group conglomerates.

The information given to accompany the overallhazard ratings can then be tailored to suit the type ofend-user, be it a home owner, insurer or estate manager(Table 7).

Gypsum and salt karst areas

Gypsum and salt are so much more soluble than carbon-ate rocks that they tend to form buried and interstratalkarst and the rocks themselves are rarely seen at thesurface. For this reason many of the influences that canbe factored into the karst prediction for the carbonaterocks do not apply to such interpretations for gypsumand salt terrains. The lithology of the superficial depositshas less influence and slope angle largely becomesirrelevant. Superficial deposit thickness does have amitigative bearing, but not until it is very thick (greaterthan about 30 m) and relatively less permeable. Thethick superficial deposits also tend to occur in areaswhere groundwater movement is already restricted orbelow sea level. For salt- and gypsum-bearing strataa slightly different delimitation approach is beingemployed; the scoring method is being investigated forrevision in the gypsum or salt application, but has notyet been quantified to make it practical. The techniqueused relates almost exclusively to the use of manuallypicked polygons created by someone with extensiveknowledge of the local geology and karst. The initialdata from which the polygons are created are based onthe BGS digital geological map information, andenhanced by information from the National KarstDatabase and local or published knowledge, to identifyand delimit the subject dissolution problem areas.

Permian gypsum-bearing strata

The Permian gypsum sequence in eastern England formsan interstratal karst with thick sequences of gypsum,sandwiched between fractured-dolomite aquifers. Themain gypsum-bearing strata are the Hayton andBillingham anhydrite–gypsum sequences that equate inpart to the Edlington and Roxby Formations at out-crop. The Brotherton Formation dolomite, sandwichedbetween the two gypsum sequences, is heavily affectedby karstic subsidence and collapse (Fig. 7) emanating


from dissolution within the underlying EdlingtonFormation–Hayton gypsum, but the dolomite itself onlylocally contains dissolution features. The amount ofdissolution is controlled by the amount of water flowthrough either or both of the sequences. The Permiangypsum sequence is a useful case example showing that

both river valleys and buried river valleys can have aprofound effect on the concentration of dissolutionfeatures. The most subsidence-prone areas occur whereeither or both these features cut through the sequenceand where water feeds down-dip in the carbonateaquifers, to discharge as sulphate-rich artesian springs

Fig. 6. (a) The digital 1:50 000 bedrock and superficial geological map of the North Hill area of the Mendip Hills, near Bristol [ST540 514], with data from the National Karst Database superimposed; purple circles are caves or stream sinks; small green circles aresinkholes. The blue colours are the Carboniferous Limestone formations, the grey is the Avon Group (see b). North Hill comprisesDevonian Portishead Formation (brown), with pink Triassic conglomerates to the NE. A thin deposit of valley head runs along thedry valley though Priddy. (b) The GeoSure hazard map for this area with data from the National Karst Database superimposed;purple circles are caves or stream sinks; small green circles are sinkholes. Much of the bare Carboniferous Limestone plateau givesa rating of C, although locally unmapped loessic cover sands give rise to higher densities of sinkholes north of Priddy. A thin zonearound the margins of the Avon Group give a rating of E; this area is where the majority of the caves and stream sinks are found.Many suffosion sinkholes occur in the valley head deposits in Priddy, which has a D rating. The Avon Group consists of a sequenceof interbedded limestones and mudstones, which is weakly karstic overall, hence the A rating. The Triassic conglomerates to the NEof North Hill host several significant cave systems and stream sinks, and are rated C, except along the margins of the PortisheadFormation, where they have a D rating.


Tab

le7.

Pos

sibl

ete

xtto

acco

mpa

nyha

zard

rati

ngs

tailo

red

for

diff

eren

tpo

tent

ial

end-

user

s

Haz

ard

rati

ngG

ener

alch

arac

teri

stic

sIn

dica

tive

impl

icat

ions

for

Pla

nner

sD

evel

oper

sor

geot

echn

ical

engi

neer

s

Hou

seho

lder

sIn

sura

nce

orfin

anci

alin

stit

utio

ns

Env

iron

men

tal

heal

thor

was

tedi

spos

al

Far

mer

san

des

tate

man

ager

s

1or

AA

reas

whe

reso

lubl

ero

cks

are

pres

ent,

but

unlik

ely

toca

use

prob

lem

sex

cept

unde

rex

cept

iona

lco

ndit

ions

No

cons

trai

nts

tola

ndus

eca

used

byla

ndin

stab

ility

wit

hin

site

Nor

mal

desk

stud

yan

dw

alko

ver

surv

eyof

site

No

mai

nten

ance

orus

eim

plic

atio

nsca

used

byla

ndin

stab

ility

No

incr

ease

dco

stbe

caus

eof

land

inst

abili

typr

esen

t

Pos

sibl

eco

ncer

nab

out

grou

ndw

ater

cont

amin

atio

n

No

rest

rict

ions

onla

ndus

ebe

caus

eof

land

inst

abili

ty,

conc

erns

abou

tgr

ound

wat

erco

ntam

inat

ion

2or

BA

reas

wit

hsi

gnifi

cant

solu

ble

rock

s,bu

tfe

wdi

ssol

utio

nfe

atur

esan

dno

subs

iden

ce;

unlik

ely

toca

use

prob

lem

sex

cept

wit

hco

nsid

erab

lesu

rfac

eor

subs

urfa

cew

ater

flow

No

cons

trai

nts

tola

ndus

eca

used

byla

ndin

stab

ility

wit

hin

site

Nor

mal

desk

stud

yan

dw

alko

ver

surv

eyof

site

.C

onsi

dera

tion

ofst

abili

tyof

site

surr

ound

ings

No

mai

nten

ance

orus

eim

plic

atio

nsca

used

byla

ndin

stab

ility

Incr

ease

dco

stbe

caus

eof

land

inst

abili

tyun

likel

y.Sl

ight

liabi

lity

beca

use

ofgr

ound

wat

erpo

lluti

onpo

ssib

le

Con

cern

abou

tgr

ound

wat

erco

ntam

inat

ion

No

rest

rict

ions

onla

ndus

ebe

caus

eof

land

inst

abili

ty,

conc

erns

abou

tgr

ound

wat

erco

ntam

inat

ion

3or

CA

reas

wit

hsi

gnifi

cant

solu

ble

rock

s,w

here

ther

ear

edi

ssol

utio

nfe

atur

es,

and

noor

very

littl

ere

cord

edsu

bsid

ence

,bu

ta

low

poss

ibili

tyof

itoc

curr

ing

natu

rally

orin

adve

rse

cond

itio

nssu

chas

high

surf

ace

orsu

bsur

face

wat

erflo

w

Rep

ort

onim

plic

atio

nsfo

rst

abili

tysh

ould

besu

bmit

ted

ifch

ange

sto

surf

ace

drai

nage

orne

wco

nstr

ucti

onar

epr

opos

ed

Site

inve

stig

atio

nsh

ould

cons

ider

spec

ifica

llyth

ela

ndst

abili

tyof

the

site

and

surr

ound

ings

.C

are

shou

ldbe

take

nw

ith

loca

ldr

aina

gein

toth

ebe

droc

k

Con

side

rati

onof

impl

icat

ions

for

stab

ility

shou

ldbe

mad

eif

chan

ges

tosu

rfac

edr

aina

geor

new

cons

truc

tion

are

plan

ned

Incr

ease

dco

stbe

caus

eof

land

inst

abili

typo

ssib

le.

Som

elia

bilit

ybe

caus

eof

grou

ndw

ater

pollu

tion

poss

ible

Pot

enti

alfo

rsi

tein

tegr

ity

tobe

dam

aged

bym

inor

grou

ndm

ovem

ents

.C

once

rnab

out

grou

ndw

ater

cont

amin

atio

n

Con

side

rm

inor

chan

ges

inla

ndus

e,su

rfac

eru

noff

and

drai

nage

topr

even

tgr

ound

wat

erco

ntam

inat

ion

and

redu

ceth

elik

elih

ood

ofsu

bsid

ence

4or

DA

reas

wit

hve

rysi

gnifi

cant

solu

ble

rock

s,w

here

ther

ear

enu

mer

ous

diss

olut

ion

feat

ures

and/

orso

me

reco

rded

subs

iden

cew

ith

am

oder

ate

poss

ibili

tyof

loca

lized

subs

iden

ceoc

curr

ing

natu

rally

orin

adve

rse

cond

itio

nssu

chas

high

surf

ace

orsu

bsur

face

wat

erflo

w

Lan

dus

ech

ange

sin

volv

ing

load

ing,

infil

ling,

exca

vati

onor

chan

ges

tosu

rfac

edr

aina

gem

ayaff

ect

stab

ility

,an

das

sess

men

tan

dm

itig

atio

nm

easu

res

shou

ldac

com

pany

appl

icat

ion.

Con

serv

atio

nm

easu

res

shou

ldbe

cons

ider

ed

Spec

ialis

tsi

tein

vest

igat

ion

for

stab

ility

asse

ssm

ent

mig

htbe

nece

ssar

ybe

fore

cons

truc

tion

.C

onst

ruct

ion

wor

km

ight

caus

esu

bsid

ence

.Su

rfac

edr

aina

gesh

ould

not

beal

low

edto

affec

tth

eka

rst

syst

emor

grou

ndw

ater

Do

not

load

the

land

,an

dob

tain

spec

ialis

tad

vice

befo

reun

dert

akin

gbu

ildin

gw

ork.

Do

not

disp

ose

ofsu

rfac

edr

aina

geto

the

grou

nd.

Mai

ntai

ndr

aina

gein

fras

truc

ture

Incr

ease

dco

stbe

caus

eof

land

inst

abili

typr

obab

le.

Lia

bilit

ybe

caus

eof

grou

ndw

ater

pollu

tion

poss

ible

Pos

sibl

eda

mag

eto

cont

amin

ativ

est

ruct

ures

,ta

nks,

drai

nage

,se

wer

s,pi

pelin

es,

etc.

Insp

ecti

onof

stru

ctur

esre

com

men

ded.

Are

aspr

one

topo

lluti

onan

dgr

ound

wat

erco

ntam

inat

ion

Con

side

rso

me

chan

ges

inla

ndus

e,su

rfac

eru

noff

and

drai

nage

topr

even

tgr

ound

wat

erco

ntam

inat

ion

and

redu

ceth

elik

elih

ood

ofsu

bsid

ence


Tab

le7

Con

tinu

ed.

Haz

ard

rati

ngG

ener

alch

arac

teri

stic

sIn

dica

tive

impl

icat

ions

for

Pla

nner

sD

evel

oper

sor

geot

echn

ical

engi

neer

s

Hou

seho

lder

sIn

sura

nce

orfin

anci

alin

stit

utio

ns

Env

iron

men

tal

heal

thor

was

tedi

spos

al

Far

mer

san

des

tate

man

ager

s

5or

EA

reas

wit

hve

rysi

gnifi

cant

solu

ble

rock

s,w

here

ther

ear

enu

mer

ous

diss

olut

ion

feat

ures

and/

orco

nsid

erab

lere

cord

edsu

bsid

ence

wit

ha

high

poss

ibili

tyof

loca

lized

subs

iden

ceoc

curr

ing

natu

rally

orin

adve

rse

cond

itio

nssu

chas

high

surf

ace

orsu

bsur

face

wat

erflo

w

Lan

dus

ech

ange

sin

volv

ing

load

ing,

infil

ling,

exca

vati

onor

chan

ges

tosu

rfac

edr

aina

gem

ayaff

ect

stab

ility

.P

erm

issi

onfo

rde

velo

pmen

tm

ight

requ

ire

inve

stig

atio

nan

dre

med

ial

wor

ksas

part

ofde

velo

pmen

t.P

erm

issi

onfo

rde

velo

pmen

tm

ight

not

bepo

ssib

le.

Con

serv

atio

nm

easu

res

shou

ldbe

cons

ider

ed

Spec

ialis

tla

ndst

abili

tyas

sess

men

tne

cess

ary.

Inve

stig

atio

n,re

med

iati

onan

d/or

mit

igat

ion

wor

ksm

ight

bene

cess

ary

tost

abili

zeth

ear

ea.

Con

stru

ctio

nw

ork

mig

htca

use

subs

iden

ce.

Surf

ace

drai

nage

mus

tno

taff

ect

the

kars

tsy

stem

orgr

ound

wat

er

Con

side

rob

tain

ing

spec

ialis

tad

vice

toad

vise

onne

edfo

rst

abili

zati

onw

ork

and/

orla

ndm

anag

emen

tpl

anto

mai

ntai

nst

abili

ty.

Do

not

disp

ose

ofsu

rfac

edr

aina

gein

toth

egr

ound

.M

aint

ain

drai

nage

infr

astr

uctu

re

Incr

ease

dco

stbe

caus

eof

land

inst

abili

tyve

rypr

obab

le.

Lia

bilit

ybe

caus

eof

grou

ndw

ater

pollu

tion

prob

able

Sign

ifica

ntpo

ssib

ility

ofda

mag

eto

cont

amin

ativ

est

ruct

ures

,ta

nks,

drai

nage

,se

wer

s,pi

pelin

es,

etc.

Reg

ular

insp

ecti

onof

stru

ctur

esre

com

men

ded.

Are

asve

rypr

one

topo

lluti

onan

dgr

ound

wat

erco

ntam

inat

ion

Con

side

rm

ajor

chan

ges

inla

ndus

e,su

rfac

eru

noff

and

drai

nage

topr

even

tgr

ound

wat

erco

ntam

inat

ion

and

redu

ceth

elik

elih

ood

ofsu

bsid

ence


in the valleys. Not enough digital information is cur-rently available to characterize the hydrogeology, so themanually picked polygon approach is currently taken byutilizing the National Karst Database to define the areasthat appear to be most susceptible to subsidence.

The point-specific information contained within theNational Karst Database allows the manually pickedpolygons for the feather edge zone to be delimited, sothat the lower parts of the overlying non-karsticSherwood Sandstone Group, which are prone to inter-stratal subsidence, are also included in the susceptiblebelt. The subsidence-prone zone is determined by com-bining the digitized geological outcrop limits of theEdlington, Brotherton and Roxby Formations with thisfeather edge zone of the Sherwood Sandstone Group.The superficial deposits across much of the susceptiblebelt are very thick, but the thickness of gypsum issufficient over most of the area for its dissolutionalremoval to cause subsidence to migrate upwardsthrough the superficial cover to the surface, creatinglarge sinkholes. In some places with thick gypsum(10–30 m), such as to the east of Darlington, where thesuperficial cover is extremely thick (30–40 m) and theground has low relief, there is a lower hydrostatic headdriving the dissolutional water-system. Here there is lesslikelihood of point subsidence, but because the superfi-cial deposits contain water-saturated sand lenses andbeds, subsidence can occur by the limited flow-displacement of loose, overlying (‘running’) sand intothe underlying gypsum karst cavities. The result of this isthe formation of upward-propagating, large shallowbowl-shaped subsidence areas.

In the Vale of Eden the gypsum sequence (consistingof a series of gypsum beds A–D) is sandwiched betweenmudstones. Consequently, less water reaches the gyp-sum, and karstic features occur mainly at the base of the

superficial deposits or in areas where there is concen-trated water flow towards local rivers and streams.Because of this, the outcrop of the A–B–C–D gypsumbeds and the zones between them are manually amalga-mated to produce a combined area. The Permian-agedgypsum of the Vale of Eden and the anhydrite exposedalong the coast are included in this area.

For the Permian gypsum karst areas, the manuallypicked polygons are classified into five groupings, asfollows.

A, very low potential: areas where gypsum is present,but the deposits are known to be thin, where theadjacent rocks are not aquifers and there is no recordedsubsidence.

B, low: areas where gypsum is present in substantialthicknesses, but where the adjacent rocks are notaquifers and where there is no recorded subsidence.

C, moderate: areas where gypsum is present in sub-stantial thicknesses, where the adjacent rocks might ormight not be aquifers, but where there is no recordedsubsidence. Mainly the majority of the Permian gypsumin the Vale of Eden and some of the Permian-agedgypsum of eastern England.

D, high: areas where gypsum is present in substantialthicknesses, where the adjacent rocks are aquifers andwhere there is some recorded subsidence. Mainly thePermian-aged gypsum of eastern England, includingDarlington, Tadcaster and Church Fenton.

E, very high: areas where gypsum is present in sub-stantial thicknesses, where the adjacent rocks are aqui-fers, where buried valleys cut through the sequenceand where there are numerous records of continuingsubsidence. Mainly the Permian-aged gypsum of easternEngland south of the general line Darlington–Ripon–Brotherton.

Triassic gypsum-bearing strata

Triassic gypsum-bearing strata are widespread, butmainly contain thick gypsum beds in the Staffordshire,south Derbyshire and south Nottinghamshire areas.Karstic features are sporadic with some caves, sinkholesand stream sinks present. In addition, some localizedbuilding damage has occurred. The expert zones definedfor the Triassic-aged gypsum terrain are subdivided asfollows.

A, very low potential: areas where gypsum is present,but the deposits are known to be thin, where theadjacent rocks are not aquifers, where the superficialcover is thick and there is no recorded subsidence.Mainly the Triassic Mercia Mudstone Group wherefibrous gypsum and thin-bedded gypsum has beenrecorded and the low-lying areas with relatively thick,lower-permeability superficial cover.

B, low: areas where gypsum is present in substantialthicknesses, but where the adjacent rocks are notaquifers and where there is no recorded subsidence.

Fig. 7. Sinkhole formed by the dissolution of Permian gypsumin the village of Sutton Howgrave [SE 3246 7928] near Ripon,North Yorkshire. The hole started to collapse in December2000, the photograph was taken on 14 February 2001 when thehole was 5–6 m in diameter and 11 m deep with water at adepth of 8 m. Photo A. H. Cooper, copyright NERC.


C, moderate: areas where gypsum is present in sub-stantial thicknesses, where the adjacent rocks might ormight not prove to be aquifers, where there is norecorded active subsidence, but where subsidencefeatures are present. Mainly the Triassic-aged MerciaMudstone Group where relatively thick gypsum ispresent.

D, high: areas where gypsum is present in substantialthicknesses, where the adjacent rocks might or might notprove to be aquifers, and where there is some recordedsubsidence.

Permo-Triassic halite-bearing strata

The subdivisions applied to the salt areas are basicallythe same as those applied to the gypsum. The differencecomes in the mechanisms of collapse and the fact thatsalt is much more soluble and much more rapidlydissolved than gypsum. The salt areas are all buriedbeneath considerable thicknesses of overlying brecciatedand collapsed rock, or thick superficial deposits.

Hazard assessment of many of the salt areas is com-plicated by the fact that room-and-pillar salt miningand brine pumping have also commonly occurred inthe areas where natural dissolution has occurred. Thehazard zones are defined as follows.

A, very low potential: areas where salt is present, butthe deposits are known to be thin and they are coveredwith impervious material.

B, low: areas where salt is present in substantialthicknesses, but where the deposits are covered with asignificant thickness of impervious material, or areaswhere there is good evidence that the majority of the salthas naturally dissolved during geological time.

C, moderate: areas where salt is present in substantialthicknesses and present at rockhead (wet rockhead) orbeneath a thin cover of impervious rock.

D, high: areas where salt is present in substantialthicknesses, present at rockhead (wet rockhead) andwhere salt springs are present in the area.

E, very high: areas where salt is present in substantialthicknesses, present at rockhead (wet rockhead) andwhere wild brining or nearby underground salt mininghas occurred, salt springs are present and there is somerecorded subsidence in the vicinity; mainly the Triassic-aged salt of Cheshire, Staffordshire and Worcestershire.

Discussion

Clearly, the GeoSure dataset provides only an indicationof where dissolution features might occur, and does notgive actual locations of karst features, nor does it serveto service the direct need for site classification at singledevelopment locations. Furthermore, the NationalKarst Database does not provide details of all karstfeatures and neither dataset should be used as a substi-

tute for detailed site investigation work, or a moredetailed hazard assessment, such as that proposed byEdmonds (1983). Whether these karst geohazards con-stitute a potential risk depends on the views of theend-user; for example, a cave might pose a problem fora construction company, but be a boon for a cavingenthusiast, whereas an irregular rockhead might be apotential hazard to a construction company excavatinga tunnel, but not pose a risk to a home owner whosehouse has good foundations.

Clearly, given the very localized nature of karst fea-tures, not all areas categorized as being of high risk willhave subsidence or contain karst features, and thisuncertainty needs to be communicated. For propertyowners, the presence of a moderate or high dissolutionrating (class E or D) does not mean that any particularproperty will collapse or subside, but it acts as a warningthat the geological conditions make the area prone tosubsidence under certain circumstance; for example, ifleaking sewers or water pipes wash out clay-infilleddissolution pipes. Karst features are generally stableunder natural or light loading conditions, until excep-tional circumstances such as severe flooding or a burstwater main occur (Fig. 8). Changes to the naturalhydrological regime caused by changes in the nature ofsurface runoff, excavating or loading the ground,groundwater abstraction, leaking services and inappro-priate drainage can also trigger subsidence in otherwisegenerally stable areas (Waltham et al. 2005). Neither ofthese datasets predict when such events will occur, butthey can prove useful for a variety of single users andorganizations (Table 7) who have the obligation of

Fig. 8. Numerous sinkholes formed by a large burst water mainat Littleheath Road, Fontwell, Sussex [SU 944 077] in late1985. At least 63 collapses are visible, eight in gardens, fouron the road and the remainder in the field. The burst watermain was near the digger in the top right of the view. Thesite is on solifluction deposits overlying Culver Chalk, close tothe Palaeogene margin. Photo copyright Sealand Aerial,Chichester, reproduced under licence.


protecting the public from adverse effects of landutilization.

Planners can utilize the datasets to inform the zoningof areas for development and building control, thusprotecting the public from severe subsidence problems(Paukstys et al. 1999). Developers and geotechnicalengineers can use the information to help with linearroute planning, site design and hazard avoidance.Householders can be better informed about potentialgeological hazards when purchasing properties and beaware of situations that can aggravate the natural situ-ation, so that situations such as letting pipes leak, oremptying a swimming pool onto the garden and causinga house to collapse (Edmonds 2005) can be avoided. Forthe insurance and financial industries the datasets canmean they are better informed about the potential risksthey take, but except in exceptional circumstances thedata should not be used as a reason for refusing toprovide insurance, although they might justify a loadingof the premium. Considering the sensitivity of karst totransport of pollution, it is also important that thedatasets provide both related baseline information (suchas the locations of stream sinks and sinkholes) andgeneralized information about the susceptibility ofspecific geological units to develop karstic features.Finally, the datasets can be used to inform farmers andestate managers about their land and some of the con-straints that they should consider when dealing with it.

Acknowledgements. Many thanks go to S. Thorpe, N.Williams, A. Richardson, H. Cullen and S. Doran, whoentered many of the data. K. Adlam is thanked for setting upthe National Karst Database and GIS. J. Walsby is thankedfor supporting the GeoSure project, and R. Newsham isthanked for processing data and providing GIS help. D. Loweand V. Banks kindly read through the text and provided usefulcomments. M. Culshaw, A. Waltham and A. Gibson arethanked for their help. This paper is published with thepermission of the Executive Director, British GeologicalSurvey.

References

A G L 1993. Review of instability due tonatural underground cavities in Great Britain. AppliedGeology Ltd, Royal Leamington Spa.

C, A. T. & W, J. P. 1999. A Gazetteer ofEnglish Caves, Fissures and Rock Shelters ContainingHuman Remains. Capra. 1. World Wide Web Address:http://capra.group.shef.ac.uk/1/caves.html.

C, A. T. & W, J. P. 2000a. A Gazetteer ofScottish Caves, Fissures and Rock Shelters ContainingHuman Remains. Capra. 2. World Wide Web Address:http://capra.group.shef.ac.uk/2/scotland.html.

C, A. T. & W, J. P. 2000b. A Gazetteer ofWelsh Caves, Fissures and Rock Shelters ContainingHuman Remains. Capra. 2. World Wide Web Address:http://capra.group.shef.ac.uk/2/wales.html.

C, A.H. 1986. Foundered strata and subsidence resultingfrom the dissolution of Permian gypsum in the Ripon and

Bedale areas, North Yorkshire. In: H, G.M. &S, D.B. (eds) The English Zechstein and RelatedTopics. Geological Society, London, Special Publications,22, 127–139.

C, A.H. 1989. Airborne multispectral scanning of sub-sidence caused by Permian gypsum dissolution at Ripon,North Yorkshire. Quarterly Journal of EngineeringGeology, 22, 219–229.

C, A.H. 1997. Gypsum karst of Great Britain.International Journal of Speleology, 5, 195–202.

C, A.H. 1998. Subsidence hazards caused by the disso-lution of Permian gypsum in England: geology, investi-gation and remediation. In: M, J.G. & E,M. (eds) Geohazards in Engineering Geology. GeologicalSociety, London, Engineering Special Publications, 15,265–275.

C, A.H. 2002. Halite karst geohazards (natural andman-made) in the United Kingdom. EnvironmentalGeology, 42, 505–512.

C, A. & S, J.M. 1999. Road and bridge con-struction across gypsum karst in England. In: B, B.,P, A.J. & H, G. (eds) Hydrogeology andEngineering Geology of Sinkholes and Karst 1999. Proceed-ings of the 7th Multidisciplinary Conference on Sinkholesand the Engineering and Environmental Impacts ofKarst, Harrisburg–Hershey, PA. Balkema, Rotterdam,407–411.

C, A.H. & S, J.M. 2002. Road and bridgeconstruction across gypsum karst in England. EngineeringGeology, 65, 217–223.

C, A.H., F, A.R., A, K. & W, J.C.2001. The development of a national Geographic Infor-mation System (GIS) for British karst geohazards and riskassessment. In: B, B. & H, G. (eds) Geotechni-cal and Environmental Applications of Karst Geology andHydrology. Proceedings of the 8th Multidisciplinary Con-ference on Sinkholes and the Engineering and Environ-mental Impacts of Karst, Louisville, KY. Balkema, Lisse,125–130.

C, M.G. & W, A.C. 1987. Natural and artifi-cial cavities as ground engineering hazards. QuarterlyJournal of Engineering Geology, 20, 139–150.

D E 1990. Planning PolicyGuidance Note 14: Development on Unstable Land.HMSO, London.

D T, L G R 2002. Planning Policy Guidance Note 14. Devel-opment on Unstable Land. Annex 2: Subsidence andPlanning. The Stationery Office, London.

E, C.N. 1983. Towards the prediction of subsidencerisk upon the Chalk outcrop. Quarterly Journal ofEngineering Geology, 16, 261–266.

E, C.N. 2001. Predicting natural cavities in Chalk. In:G, J.S. (ed.) Land surface evaluation for engineer-ing practice. Geological Society, London, EngineeringSpecial Publications, 18, 29–38.

E, C.N. 2005. Subsidence over a chalk pipe atChalfont St. Peter, UK. In: W, A.C., B, F.G.& C, M.G. (eds) Sinkholes and Subsidence. Karstand Cavernous Rocks in Engineering and Construction.Praxis, Chichester, 309–312.

F, A.R. 2001. Karst development in the SouthernEnglish Chalk. In: B, B. & H, G. (eds) Geotech-nical and Environmental Applications of Karst Geology andHydrology. Proceedings of the 8th Multidisciplinary Con-ference on Sinkholes and the Engineering and Environ-mental Impacts of Karst, Louisville, KY. Balkema, Lisse,77–82.


F, A.R. 2004. Ogof Draenen. In: G, J. (ed.)Encyclopedia of Cave and Karst Science. FitzroyDearborn, London, 310–312.

F, T.D. (.) 1989. Limestones and Caves of Wales.Cambridge University Press, Cambridge.

H, R. 1965. Sink-holes on the Lincolnshire limestonebetween Grantham and Stamford. East MidlandGeographer, 3, 454–460.

K, A. & L, D. 2002. Implications of speleologi-cal studies for karst subsidence hazard assessment.International Journal of Speleology, 31.

K, A., L, D., C, A. & S, U. 1997.Gypsum karst of the world. International Journal ofSpeleology, 5.

L-B, J., Y, P.L., F, R.A., C, A.& B, J. 2002. A decision logic framework forinvestigating subsidence problems potentially attributableto gypsum karstification. Engineering Geology, 65,205–215.

L, S.-E. 2001. Marble stripe karst of the Scandi-navian Caledonides, An end member in the contact karstspectrum. Acta Carsologica, 30, 47–92.

L, T.J. 1988. The Limestone Caves of ScotlandPart 2: Caves of Assynt. Grampian Speleological GroupOccasional Publication, 6.

M, L.D., A, T.C., B, J.A., B,J.P., F, A.R. & W, A.T. 2006. Karsticbehaviour of groundwater in the English Chalk. Journalof Hydrology, 330, 63–70.

N, A.J.G. & H, D.E. 1973. Salt. MineralDossier No. 7, Mineral Resources ConsultativeCommittee. HMSO, London.

O, A.D. 1975. The Caves of Scotland, except Assynt.Tony Oldham, Bristol.

O, A.D., O, J.E.A. & S, J. 1986. TheConcise Caves of Devon. Anne Oldham, Crymyrch.

P, B., C, A.H. & A, J. 1999. Planningfor gypsum geohazards in Lithuania and England.Engineering Geology, 52, 93–103.

R, P.F. & C, A.H. 1993. A cave system in Permiangypsum at Houtsay Quarry, Newbiggin, Cumbria,England. Cave Science, 20, 23–28.

S, M.J. 1990. Triassic Palaeokarst in Britain. CaveScience, 17, 93–102.

W, A.C., S, M.J., F, A.R. & G, H.1997. Karst and Caves of Great Britain: GeologicalConservation Review. Chapman and Hall, London.

W, A.C., B, F.G. & C, M.G. 2005.Sinkholes and Subsidence. Karst and Cavernous Rocks inEngineering and Construction. Praxis, Chichester.

Received 29 August 2007; accepted 4 March 2008.


karst geohazards in the uk: the use of digital data...

Documents