Elsevier

Journal of Hydrology

Volume 312, Issues 1–4, 10 October 2005, Pages 207-222
Journal of Hydrology

Impact of DEM mesh size and soil map scale on SWAT runoff, sediment, and NO3–N loads predictions

https://doi.org/10.1016/j.jhydrol.2005.02.017Get rights and content

Abstract

The accuracy of agricultural nonpoint source pollution models depends to a great extent on how well model input spatial parameters describe the relevant characteristics of the watershed. It is assumed that reducing the precision of spatial input parameters affects the simulation results of runoff and sediment yield from the entire watershed. However, there may be no significant increase in the accuracy of models, as a result of more precise topographic or soil information, which increase the input data collection and preparation. The objective of this study was to determine the impact of the mesh size of the digital elevation model, DEM (from 20 to 500 m) and the soil map scale (1/25,000; 1/250,000 and; 1/500,000 scale) within the Soil and Water Analysis Tool (SWAT) to simulate runoff, sediment, and NO3–N loads at the outlet of an agricultural watershed. Results of the Lower Walnut Creek (21.8 km2, central Iowa) showed that an upper limit to DEM mesh size of 50 m is required to simulate watershed loads. Decreasing the mesh size beyond this threshold does not substantially affect the computed runoff flux but generated prediction errors for nitrogen and sediment yields. Whatever the DEM mesh size considered, a detailed soil map has to be considered to accurately estimate the loads. Finally, the impacts of DEM mesh size and soil map scale on the modeling results are discussed in respect of the relevant characteristics of the watershed and included in SWAT.

Introduction

Water, its availability and quality, and sediment delivery have become challenging issues threatening food supply, and security, human health and natural ecosystems. Moreover in the light of recent concern on climate and/or land-use changes, this quantitative information is even more crucial since it allows the description of the expected impact of changes on water supply, water quality and soil erosion. As a result and to satisfy the needs of environmental scientists and planners to make informed decisions, there have been attempts to develop predictive models. The quality of the spatial information is presumed to directly affect the simulation results of hydrologic models. The accuracy of which depends to a great extent on how well model input spatial parameters describe the relevant characteristics of the watershed. Fine digital elevation models (DEMs) and detailed or high-scale land-use and soil maps (e.g. 1/25,000) would generate accurate estimations. But such information is very costly to obtain. This is especially true in the case of soil maps which require numerous soil observations and laboratory analyses, and to a lesser extent fine DEMs. This paper investigates the impact of the mesh size of the digital elevation model and the soil map scale on simulated runoff, sediment, and NO3–N loads of an agricultural watershed by using the Soil and Water Analysis Tool (SWAT).

Several references exist on the interaction between the accuracy of the topographical description and the prediction quality for spatial prediction (Quinn et al., 1991, Farajall and Vieux, 1995, Brasington and Richards, 1998, Chaplot et al., 2000, Thompson et al., 2001). Although DEMs describe the landscapes' shape over large areas, high prediction errors mostly occur due to the misestimation of terrain attributes such as the elevation above the stream (Chaplot et al., 2000) and the slope curvature (Thompson et al., 2001). Zhang and Montgomery (1994) suggested a spatial resolution of 10 m to represent hydrologic processes based on simulations with TOPMODEL (Beven and Kirkby, 1979). Bloschl and Sivapalan (1995) stated that high resolution DEMs are required when local processes are dominant. In a watershed of Iowa, Kalin et al. (2003) by using the KINEROS model observed that increasing the DEM resolution increased the water flux at the peak runoff without affecting its time occurrence. The precision of the DEM is also expected to affect the delineation of watersheds which in turn would influence models' prediction quality. FitzHugh and Mackay (2000) using the Soil and Water Assessment Tool (SWAT), indicated in a watershed of Wisconsin a 44% drop of sediment estimates from the coarsest to the finest watershed delineations. Vieux and Needham (1993) found that outputs from the AGNPS model (Young et al., 1987) also varied with changes in the cell size. A decrease of sediment yield was observed with an increase of DEM mesh from 100 to 200 m due to decreasing channel erosion. In addition Brown et al. (1993) showed that the prediction of sediment yield using the ANSWERS model began to change at DEM meshes greater than 120 m. Such a result was attributed to the impacts of increasing amounts of aggregation on the distribution of overland soil, land use, and terrain parameters.

Proper implementation of environmental models requires decisions to be made about not only the size of the DEM mesh, but also about the precision of the soil and land-use maps. DEM mesh size and scale of the soil map are inter-related, thus decisions about an optimal mesh cannot be made without a consideration of the scale of the available soil map and vice-versa. Finally, it is important to understand not only the level to which model prediction errors are affected by the precision of spatial input data, but also the mechanisms involved in these changes. The authors' belief is that fine topographic or detailed soil information could unnecessarily increase the effort required for data collection and preparation without any significant benefit for models accuracy.

The Soil and Water Assessment Tool (SWAT, Neitsch et al., 2000), an agricultural nonpoint source pollution model, was designed to predict the impact of agricultural management practices on water outflow, sediment, nutrient and pesticide loads for large ungauged sub-basins (Arnold et al., 1998). Input information for each sub-basin are climate; ponds/wetlands; groundwater; and the main channel, or reach, draining the sub-basin as well as hydrologic response units or HRUs with unique land cover, soil, and management combinations. Among the required input data for SWAT, the precision of the DEM and the soil map scale would greatly affect its prediction results. Indeed, the detail of the sub-basin or reaches delineation as well as the estimation at the sub-basin or HRU levels of topographic properties would for instance, depend on the spatial accuracy of the available input topography. This should be high enough to capture significant topographic variability within the watershed.

The study was conducted in the Lower Walnut Creek watershed, an agricultural basin of central Iowa (USA) where SWAT was previously validated using observed monthly water and NO3–N fluxes collected over the 1991–1998 period and using a 20 m DEM and a 1/25,000 soil map. In this study, the estimated loads for runoff, sediment, and NO3–N by using 12 different DEM mesh size (from 20 to 500 m) and three soil map scales (1/25,000; 1/250,000; and 1/500,000) were compared over 1991–1998.

Section snippets

The soil and water assessment tool

The Soil and Water Assessment Tool (SWAT) is a physically based and continuous time model developed to predict, over long periods, the impact of land management practices on water, sediment and chemical yields in watersheds with varying soils, land-use, and management conditions (Arnold et al., 1998). The model simuls both the land phase of the hydrologic cycle, controlling the amount of water, sediment, nutrient and pesticide loadings to the main channel in each sub-basin, and the water or

The rainfall over 1991–1998

Mean annual rainfall during 1991–1998 was 807 mm. A minimum of 551 mm was observed in 1994 whereas 1993 showed the maximum of 1024 mm. The remaining years exhibited amounts close to the mean amount over the study period, e.g. 852 mm in 1991; 895 mm in 1996; and 940 mm in 1998. Maximum monthly rainfall mainly occurred in late spring and summer with a mean amount over the period of 112 mm in June and 131 mm in July. June 1998 showed 302 mm. December, January and February with 26, 29, and 27 mm were the

Discussion

Simulation results over a 9-year period showed that runoff predictions at the outlet of the 21.8 km2 Lower Walnut Creek watershed of central Iowa were consistently accurate irrespective of the DEM mesh used. This contradicts previous researches which showed that simplifications in describing the watershed topography greatly affect the runoff production by using the NRCS curve number method. Kalin et al. (2003) attributed the observed decrease of the runoff volume and peak at coarser DEMs to the

Acknowledgements

This research was supported by the Texas Institute for Applied Environmental Research (TIAER), Stephenville (Texas), and the USDA-ARS, National Soil Tilth Laboratory of Ames (Iowa). The author is grateful to Ali Saleh from TIAER and Jeff Arnold from USDA-ARS, Temple (Texas) for their advice in SWAT implementation and to Dane Jaynes from the National Soil Tilth Laboratory for providing spatial data. I am also indebted to the critical suggestions these collaborators gave during this study.

References (26)

  • M. DiLuzio et al.

    SWAT Arcview Interface User's Guide For SWAT, 98.1: 1998 Beta Release

    (1998)
  • ESRI, 1997. Understanding GIS. The ArcView GIS 3.2', ESRI, 380 New York Street, Redlands, CA, USA,...
  • N.S. Farajall et al.

    Capturing the essential spatial variability in distributed hydrological modeling infiltration parameters

    Hydrological Processes

    (1995)
  • Cited by (0)

    View full text