Paper No. 95-2701An ASAE Meeting Presentation

The Use Of A GIS For Assessing Management AlternativesTo Reduce Over-land Sediment Production And Transport

S.G. McNulty1, L.W. Swift, Jr., J. Hays, A. Clingenpeel

1. Coweeta Hydrologic Laboratory, 3160 Coweeta Lab Rd., Otto NC 28763

Written for Presentation at the1995 ASAE ANNUAL INTERNATIONAL MEETING

Sponsored by ASAE

HYATT Regency HotelChicago, IllinoisJune 19-23, 1995

Summary:Soil erosion models can be coupled with a geographic information system (GIS) toprovide increased spatial resolution of soil erosion prediction, while reducing the timeneeded for model parameterization. We used a GIS, a simple soil sediment predictionmodel (i.e. USLE), and a simple sediment transport model, to predict soil erosion on900 m2 grids across a 1500 ha watershed, for two alternative logging roadmanagement practices. Although simple sediment production models have few inputparameters, and do not provide extremely accurate estimates of absolute soil loss,these models have utility for comparing alternative management practice effects onrelative changes in soil loss. Additionally, the combination of the USLE with a GIS,to predict areas with a high potential for soil loss, could be used to alert landmanagers to areas where additional soil erosion mitigation is necessary. Themanagers can then use more detailed and accurate, although data intensive, modelssuch as WEPP in these potential problem areas to predict soil erosion rates moreaccurately. We concluded that a GIS can be coupled with a simple soil sedimentmodel to assess relative differences in soil erosion given two management scenarios,and to help locate areas within watersheds where more detailed assessment of soil lossis needed.

Keywords:Soil erosion, GIS, Erosion Modeling, Forest Management

The authors(s) is solely responsible for the content of this technical presentation. The technical presentation does not necessarily reflect theofficial position of ASAE, and its printing and distribution does not constitute an endorsement of views which may be expressed.

Technical presentations are not subject to the formal peer review process by ASAE editorial committees; therefore, they are not to bepresented as refereed publications.

Quotation from this work should state that it is from a presentation made by (name of author) at the (listed) ASAE meeting.

EXAMPLE - From Author's Last Name, Initials. Title of Presentation.' Presented at the Date and Title of meeting, Paper No. x. ASAE,2950 Niles Rd., St. Joseph, MI 49085-9659 USA.

For information about securing permission to reprint or reproduce a technical presentation, please address inquires to ASAE.

ASAE, 2950 Niles Rd., St. Joseph, MI 49085-9659 USAVoice:616.429.0300 FAX 616.429.3852

INTRODUCTION

Soil erosion due to road construction, log removal, and site preparation can be themajor forestry impact on water quality. The ability to estimate soil loss from the watersheddue to past and proposed forest activities is an important tool for cumulative effectsassessment and management planning. Soil erosion models developed for agriculturalpractices have been adapted to the forest situation with varying success (Dissmeyer andFoster, 1980). Unlike in agriculture, forest soil disturbance is often patchy anddiscontinuous at the site level and the disturbed sites are separated over a watershed. Inaddition to being spatially discontinuous, forest disturbances may be separated by longperiods of recovery and non-disturbance. These factors make soil erosion modeling at thewatershed scale difficult.

A Geographic Information System (GIS) is a powerful tool which can be utilized forall phases of ecosystem modeling (Verbyla, 1990; Everham et al., 1991), including soilerosion modeling (McNulty et al., 1995). A GIS has existed since the early 1960's(Tomlinson et al., 1976), but only since the mid-1980's have reductions in software andhardware costs and increased program flexibility made them viable for ecosystem research(Iverson and Risser, 1987; Johnston, 1987; Johnson, 1990; Osborne, 1990; Davis et al.,1991; Drayton et al., 1992). For these reasons, Lanfear (1989) concluded that the impact ofGIS in understanding environmental issues may be as great as that of the introduction of theFORTRAN programming language.

A soil erosion model supported within a GIS enables the land manager to examine thespatial distribution of disturbance on soil erosion and transport for both current (baseline) andproposed management scenarios. The system could be used for risk assessment on theeffects of alternative management scenarios on stream water quality. This paper discussesthe development and use of a simple CIS-based soil erosion and terrestrial transport modelfor use by land managers.

METHODS

Construction of the sediment production and transport models occurred in fourseparate stages. First, the models were examined, adapted or developed. For example,over-land sediment transport models have not traditionally contained a spatial distributioncomponent. A GIS allows for eroded soil to be spatially disturbed across the watershed, asopposed to moving in mass to a single endpoint. Second, data necessary to define and latervalidate the models were assembled or measured. Third, the models were placed into a userfriendly context to make data input and model operation simple for land managers withlimited resources and training in the use of models. Finally, the model outputs should beeasily interpreted and useful as a planning tool. Each of these stages will be discussedseparately.

Site locationThe data necessary to develop and validate the models are derived from a combination

of pre-existing digitized maps (e.g. soil series, forest compartment boundaries, roads,streams and topography) and field collected measurements (e.g. soil disturbance, streamsedimentation, and over-land soil transport rates). The Wine Springs EcosystemManagement Project is a 1143 ha pilot study for model development and validation (Fig. 1).

ForKtSwIca D>M Inttrtac* Vtaw

iTf

AKCH.OT

Pan/Z»»m r) x.y: 3.94931.3.314961 ' 4x.dr. 334331.S.31438 dbfc 162182

QShefterwood CutHi Group Selection CutS§Two Aged CutSi Wildlife forage BurnH Community Restoration Bum• Control Plots\M Streams ffl Forest Access Roadsfflftwed forest Roads B2 Wine Springs Watershed 1MJU

nnaasfsiaaaasf

Wayah H«ng»r PIrt-kt)

9«5«llneH2OJ

SwImHZO)narlal PhotearaphJw«t»nhedjStraams)Compartnwnts j

Figure 1 The location and various proposed management activities, streams and current roadsystems of the Wine Spring Ecosystem Management Area are presented in this CIS dashboard.

Basin elevation ranges from 915 m at Nantahala Lake to 1655 m at Wine SpringBald. This forested watershed is part of the Wayah Ranger District. About one-half thebasin has been undisturbed by forestry activities for 60 or more years. Portions of the upperwatershed have been actively managed for timber production within the last 25 years.

The focus of current management planning for the Wine Spring Ecosystem Project ison whole ecosystem management using 36 Desired Future Condition (DFC) statements toguide activities (Swank and Culpepper, 1993). These broad-based DFC's were developed bya team of land managers, forest user groups, environmental interest groups, and ecosystemscientists. To monitor the Ecosystem Management Project's achievement of DFC's, a widevariety of physical, chemical and biological information is being accumulated describingpresent conditions. Some of these data are being applied to this study to develop, test andvalidate the sediment production and transport models using a CIS.

Hardware requirementsThe databases and models are being developed in a work-station environment using

the ARC-INFO GIS. The work-station environment is both needed and practical for modeloperation. Due to the intensive computations for each cell required by the models and thelarge hard disk capacity necessary to store input and output data, a work-station is the mostfeasible method for providing these resources.

User Friendly Model InterfacingA major hinderance for the use of sediment models by land managers is the difficulty

of applying models. The two largest obstacles are the collection of input databases andinterpretation of the results (Engel et al., 1993; McNulty et al., 1995). By placing thedatabases and models in a GIS with a user friendly environment, data input and modeloutputs can be more easily used and interpreted (Fig 1.). Most of the GIS databases neededto run the models are already digitized from other sources. If a database is not available fora particular watershed, the manager could either collect the watershed data from the field, oruse a regional database for that model parameter (e.g. the rainfall run-off factor (R)).

With a GIS, model outputs are displayed as maps. Maps of soil erosion and sedimenttransport will assist the land manager in identifying areas where sections of the watershed aremost susceptible to soil erosion, and where over-land sediment transport trails have thepotential to reach a stream. The model could be re-run through a succession of alternativemanagement scenarios, until a management strategy (e.g. using filter strips, brush barriers,or altering the season in which management activities are conducted) is found whichminimizes (or reduces to an acceptable level) soil erosion and stream sedimentation. Thus,output maps would be a planning tool for cumulative effects assessment and ecosystemmanagement.

Sediment Production ModellingThe universal soil loss equation (USLE) is used to predict erosion for each grid cell.

Ecosystem factors regulating the production of sediment are input to the model as GISdatabases (e.g. Digital Elevation Model (DEM), streams, soil series K factors, roads, skid-trails, and landing locations). The vector based information is converted into a 30 x 30 mgrid format in ARC-INFO.

The USLE, which is a simple model to parameterize, is described in equation 1

A = R x K x L S x C x P (1)

where A is seasonal or annual soil loss; R is the rainfall runoff factor; K is thesoil erosivity factor; LS it a topographic factor which combines slope length(L) with slope steepness (S); C is the forest cover management factor; and P isthe soil conservation practice factor.

1MILE

Figure 2 This map projects the soil erosivity factor (K) at a 30 x 30 m grid across the Wine SpringEcosystem Management Area. The data were transferred into the model from pre-digitized soil seriesmaps.

In the soil erosion model, each factor is digitized into the GIS as an attribute from pre-existing data. For example, rain-fall intensity (R) is derived from regional values or data fromclimate stations located near the watershed. The data can be extrapolated across the basin based onelevation and location of the rainfall collectors, the data can be extrapolated across the basin. Otherfactors such as the LS factor can be derived from pre-existing DEM's, while the K factor is anattribute attached to digitized soil series maps. Figure 2 illustrates how the K factor which variesacross the Wine Spring watershed is displayed using a GIS.

Within the GIS system, forest managers can attempt to alter the sediment production ratesthrough changes in the cover management factor (C), and the soil conservation practice (P). In ourmodeling project, forest management practices (e.g. road building, burning, timber harvesting) areentered into the GIS database (Fig. 1). Depending on the type of practice, the values for C x P arechanged within affected cells. Forest managers can alter the predicted rates of soil loss by changingwhere, when, and how various forest management practices are conducted.

Given that the USLE is a simple model, estimations of sediment loss should not be expectedto precisely match measured rates. Instead, we envision the use of the model as a tool forpredicting relative soil loss. Forest managers will be able to use this model to predict areas withpotentially high soil loss relative to other sections of the management area. By running the modelmultiple times under different management practices and seasons, this model would be useful fordetermining which conditions minimize total soil loss or soil loss in sensitive areas (i.e. nearstreams). This modeling work is being conducted in a modular framework (i.e. the sedimentproduction and transport models are separate). As more sophisticated models (e.g. WEPP) becomeavailable, the substitution of WEPP for the USLE into the modeling framework will be facilitatedbecause some of the necessary databases already exist in the GIS.

Over-land Sediment TransportAfter the USLE model predicts amounts of soil loss, the GIS routes sediment out of the cell.

Near Wine Spring, numerous sediment trails have been measured across otherwise undisturbedforest floor on slopes ranging in gradient from 20 to 40%. Sediment volume was measured every 3m down slope from the sediment source to the end of the sediment trail. Based on these fieldmeasurements, and other more extensive data (Ketcheson and Megahan, submitted), we developed asimple relationship between sediment transport distance and sediment volume (or mass). Inpreliminary results from measured sediment trails near Wine Spring, percent sediment volume wasfound to be linearly related to percent sediment distance for large volumes (> 20 m3) of sediment.However, with smaller volumes of sediment (<5 m3), proportionally more of the sediment isdeposited closer to the sediment source. These results are similar to the results of Ketcheson andMegahan (submitted) who measured over 300 sediment trails in the Pacific Northwest and found thattotal sediment volume was highly correlated (r2>0.95, P< 0.0001) with total sediment traveldistance (per. comm).

Using an average bulk density of 1.4 g cm"3 for sediment (Farmer and Van Haveren, 1971),total sediment volume can be converted into sediment mass. Data collected at Wine Spring alsoshowed that total travel distance was a function of total sediment mass (^=0.93, P<0.05, n=4(including zero point)).

Given these two relations (i.e., total sediment mass v. total sediment travel, and % sedimenttravel v. % sediment mass), a rough approximation of sediment transport and the amount ofsediment deposited in each downslope cell can be calculated (McNulty et al., 1995). At each gridcell, the mass of soil loss is calculated from the USLE and added to any sediment transported intothat cell. From this mass of soil, sediment transport distance is calculated by equation (2) where

L= 5.1 + (M)x 1.97 x 10-3 (2)

L = sediment travel distance (m); and M = sediment mass (kg).

In the CIS, the slope of each cell is derived from the DEM. This determines which directionsediment is moved toward the next adjoining cell. The amount of sediment deposited on a adjoiningsink cell is a function of the percent of total sediment mass, and the maximum amount of sedimentthat a grid cell can absorb (McNulty et al. 1995). Once a fraction of the total sediment mass from asource cell has been deposited onto a sink cell, the remaining mass of sediment will be calculatedand the sink cell will become the new source cell. Again the travel distance of the remainingsediment will be predicted based on remaining sediment mass. The interaction between sedimentproduction, travel, and deposition continues until the sediment encounters a stream cell or all of thesediment is deposited onto the forest floor (McNulty et al., 1995).

Management ScenariosTwo management scenarios were run using the USLE model with a GIS interface. In the

first scenario, all forest logging roads were well graveled surfaces with no ruts or exposed soil.Road maintenance in the first scenario is consistent with Forest Service "Best managementpractices". In the second scenario, forest logging roads were classified as rutted, with someexposed soil. This classification is typical of poorly constructed roads, or roads which have notbeen properly maintained. Both scenarios were run with average spring precipitation intensity andamount, and a variety of management cuts and regeneration burns (Fig. 1). The only differencebetween scenario 1 and 2 was the degree of road maintenance. Within the model, changes in roadmaintenance, and logging and burning practices are entered as variations in the C x P Factor.

RESULTS AND DISCUSSION

Sediment ProductionNo sediment loss was predicted for most of the watershed using either management scenario.

Small amounts of sediment loss (< 0.2 t ha"1) were predicted from the proposed prescribed burnareas and proposed timber cutting sites. Burning is prescribed for pine-oak community restorationand to improve wildlife habitat. The majority of the predicted soil loss was came from roadsurfaces (Fig. 3 and 4).

No Soil LossLess Than 0.1 tLess Than 1 tLess Than 12 tMore Than 12 t 1MILE

Figure 3 USLE predictions of soil erosion for forest management scenario 1 (a three month springperiod of a average precipitation year, given well maintained logging roads).

Soil erosion rates were higher for the poorly maintained road (Fig. 4), compared to the wellmaintained road soil erosion rates (Fig. 3). In the well maintained road scenario (scenario 1), only6 cells had estimated sediment production rates > 12 t cell'1, while in the poorly maintained roadscenario (scenario 2), several sections of road had estimated sediment production rates > 12 t cell"1.The maximum amount of predicted sediment was 32 t cell"1, and occurred on a poorly graveled roadcell (scenario 2) which was located on a steep slope.

Sediment TransportGiven a flat 30 x 30 m grid cell, the sediment transport equation predicts that at least 12.6 t of

sediment needs to enter a cell (i.e. if L = 30, then bye eq. 2, 30 = 5.1 + (M) x 1.97 x 10"3, .'.M= 12.6t) for the sediment to travel further than one grid cell over the forest floor. If the slope ofa grid cell > 0, the slope length would be > 30 m and additional soil mass would be required tomove the

No Soil LossLess Than 0.1 tLess Than 1 tLess Than 12 t 1MILE

figure 4 USLE predictions of soil erosion for forest management scenario 2 (a three month springperiod of a average precipitation year, given poorly maintained or construction logging roads).

sediment trail the increased distance. If the sediment mass was _<. 12.6 t, the forest floor wouldabsorb the mass within the first cell below the source and the sediment trail would end.

Therefore, this model predicts that very few of the sediment sources will have significantmovement given the proposed management practices. However, in the few cases where sedimentmovement may be significant, further analysis should be conducted to determine if sediments couldreach nearby streams. If the model predicted that the sediment would reach the stream givencurrent management practices, brush barriers, filter strips or relocation of management activitiescould reduce or eliminate sediment movement. The model could be re-run given changes inmanagement practices to determine if sediment transport to the stream is still likely.

CONCLUSIONSA GIS facilitates the linking of separate models and databases for the prediction of soil

erosion and over-land sediment transport. Using a modular approach, models can easily be

10

exchanged within the larger CIS framework. In this example, initial use of the USLE model on theWine Spring Ecosystem Management Area predicts little soil erosion across most of the watershedand little soil movement. Although the results of this research are preliminary, forest land managerscould use this modeling structure to minimize sediment production and stream water impacts givenalternative forest management practices. The utility of GIS in forest erosion production andtransport modeling will increase as a tool for land managers during the coming years.

LITERATURE CITED

Davis, F.W., D.A. Quattrochi, M.K. Ridd, N. Lam, SJ. Walsh, J.C. Michaelsen,J. Franklin, D.A. Stow, C.J. Johannsen, C.A. Johnston. 1991. Environmental analysis usingintegrated GIS and remotely sensed data: some research needs and priorities. Photogramm. Eng.Rem. Sens. 57(6): 689-697.

Dissmeyer, G.E. and G.R. Foster. 1980. A guide for predicting sheet and rill erosion of forestland. USDA Forest Service Southeastern Area. Atlanta, GA. Tech. Pub. SA-7P 11. 40 pgs.

Drayton, R.S., B.M. Wilde, J.K.H. Harris. 1992. Geographical Information Systemapproach to distributed modelling. Hydrol. Process. 6: 361-368.

Engel, B.A., R. Srinivasan, J. Arnold, C. Rewerts and SJ. Brown. 1993. "Nonpoint source(NPS) pollution modeling using models integrated with geographic information systems (GIS)."Wat. Sci. Tech. 28: 685-690.

Everham, E.M. HI, K.B. Wooster, C.A.S. Hall. 1991. Forest regional climate modeling. In:Proc. of the Symposium on Systems Analysis in Forest Resources Conference, 1991, Charleston,S.C.,p. 11-16.

Farmer, E.E., and B.P. Van Haveren. 1971. "Soil erosion by overland flow and raindrop splash onthree mountain soils." USDA For. Ser. Res. Pap. INT-100. 14 pgs.

Huang, S. and S. Titus. 1993. "An index of site productivity for uneven-aged or mixed-speciesstands." Can. J. For. Res. 23: 558-562.

Iverson, L.R., P.O. Risser. 1987. Analyzing long-term changes in vegetation withGeographic Information System and remotely sensed data. Adv. in Space Res. 7(11):183-194.

Johnson, L.B. 1990. Analyzing spatial and temporal phenomena using geographicalinformation systems. Landscape Ecol. 4(1): 31-43.

Johnston, K.M. 1987. Natural resource modeling in the Geographic Information System

11environment. Photogramm. Eng. Remote Sens. 53(10): 1411-1415.

Kelly, G.D. 1976. "User's guide for the computer program SEDROUTE." USDA Forest Service,Caribou National Forest. Sept. 1976. 36 pp.

Ketcheson, G.L. and W.F. Megahan. submitted. "Sediment production and downslope sedimenttransport from forest roads in granitic watersheds." J. Environ. Man.

Lanfear, KJ. 1989. Editorial: Geographic Information Systems and water resourcesapplications. Water Resour. Bull. 25(3): v-vi.

McNulty, S.G., J.M. Vose, W.T. Swank, J.D. Aber, C.A. Federer. 1994. Regional Scale ForestEcosystem Modeling: Database Development Model Predictions and Validation Using aGeographic Information System. Climate Res. 4:233-231.

McNulty, S., L. Swift, J.Hays, and A. Clingenpeel. 1995. Predicting watershed erosion andterrestrial sediment transport using a GIS. In: Proc. of International Erosion ControlAssociation Symposium, "Carrying the Torch for Erosion Control: An Olympic Task", AtlantaGA. p.395-406.

Osborne, S. 1990. Growth models and geographic information systems. USDA, For. Serv.Gen. Tech. Rep. PNW-263, p. 397-402.

Smith, D.D. and W.H. Wischmeier. 1957. "Factors affecting sheet and rill erosion."Trans. Amer. Geophys. Union 38:889-896.

Tomlinson, R.F., H.W. Calkins, D.F. Marble. 1976. Computer handling of geographicdata. UNESCO Press., Paris, 9 p.

Verbyla, D.L. 1990. Predicting forest changes associated with climate warming: potentialuses of GIS technology. In: Symposium on Management and Productivity of Western-Montane Forest Soils, Boise, Ido., p. 193-195.

Acknowledgements The authors thank Patsy Clinton and Eugene Pape for field support.