Evaluating the influence of spatial resolutions of DEM on watershed runoff and sediment yield using SWAT

Abstract

Digital elevation model (DEM) of a watershed forms key basis for hydrologic modelling and its resolution plays a key role in accurate prediction of various hydrological processes. This study appraises the effect of different DEMs with varied spatial resolutions (namely TOPO 20 m, CARTO 30 m, ASTER 30 m, SRTM 90 m, GEO-AUS 500 m and USGS 1000 m) on hydrological response of watershed using Soil and Water Assessment Tool (SWAT) and applied for a case study of Kaddam watershed in India for estimating runoff and sediment yield. From the results of case study, it was observed that reach lengths, reach slopes, minimum and maximum elevations, sub-watershed areas, land use mapping areas within the sub-watershed and number of HRUs varied substantially due to DEM resolutions, and consequently resulted in a considerable variability in estimated daily runoff and sediment yields. It was also observed that, daily runoff values have increased (decreased) on low (high) rainy days respectively with coarser resolution of DEM. The daily sediment yield values from each sub-watershed decreased with coarser resolution of the DEM. The study found that the performance of SWAT model prediction was not influenced much for finer resolution DEMs up to 90 m for estimation of runoff, but it certainly influenced the estimation of sediment yields. The DEMs of TOPO 20 m and CARTO 30 m provided better estimates of sub-watershed areas, runoff and sediment yield values over other DEMs.

Appendix

1.1 Nash–Sutcliffe efficiency

Nash–Sutcliffe efficiency (NSE) was used to assess the predictive power of hydrological models (Nash and Sutcliffe 1970). NSE is computed for each specific DEM scenario to measure how well the model predictions represent the observed data, relative to a prediction made using the average observed value. NSE is given by:

$$ \text{NSE}=1-\frac{{\sum}_{i=1}^{n} (O_{i} -S_{i} )^{2}}{{\sum}_{i=1}^{n}(O_{i} -\overline{O})^{2}}, $$

(A1)

where O _i and S _i are the observed and simulated values, n is the total number of paired values, $\overline {{O}}$ is the mean observed value.

1.2 Coefficient of determination

Coefficient of determination (R ²) is interpreted as the goodness-of-fit of a regression. It is simply a measure of variation in the regression explained by the independent variable. Higher coefficient of determination indicates better performance of the model. R ² ranges from 0 to 1 and it is given by:

$$ R^{2}=\frac{\left( {{\sum}_{i=1}^{n} (O_{i} -\overline{O}) (S_{i} -\overline{S})} \right)^{2}}{{\sum}_{i=1}^{n} (O_{i} -\overline{O} )^{2}\times {\sum}_{i=1}^{n} (S_{i} -\overline{S})^{2}}, $$

(A2)

where $\overline {S}$ is the mean of simulated values.

1.3 Percent bias (PBIAS)

Percent bias (PBIAS) measures the average tendency of the simulated data to over- or underpredict the observed data. The desired value of PBIAS is 0.0, with low-magnitude values indicating accurate model simulation. Positive values indicate model underestimation bias, and negative values indicate model overestimation bias. PBIAS is given by:

$$ \text{PBIAS}=\frac{{\sum}_{i=1}^{n} (O_{i} -S_{i} )\times 100}{{\sum}_{i=1}^{n} (O_{i} )}. $$

(A3)