Evaluating topographic wetness indices across central New York agricultural landscapes

M. T. Walter,B. K. Richards,Z. Qiu,M. Fleming,B. P. Buchanan,J. Archibald,R. L. Schneider

doi:10.5194/hess-18-3279-2014

Abstract

Abstract. Accurately predicting soil moisture patterns in the landscape is a persistent challenge. In humid regions, topographic wetness indices (TWIs) are widely used to approximate relative soil moisture patterns. However, there are many ways to calculate TWIs and very few field studies have evaluated the different approaches – especially in the US. We calculated TWIs using over 400 unique formulations that considered different digital elevation model (DEM) resolutions (cell size), vertical precision of DEM, flow direction and slope algorithms, smoothing via low-pass filtering, and the inclusion of relevant soil properties. We correlated each TWI with observed patterns of soil moisture at five agricultural fields in central NY, USA, with each field visited five to eight times between August and November 2012. Using a mixed effects modeling approach, we were able to identify optimal TWI formulations applicable to moderate relief agricultural settings that may provide guidance for practitioners and future studies. Overall, TWIs were moderately well correlated with observed soil moisture patterns; in the best case the relationship between TWI and soil moisture had an average R2 and Spearman correlation value of 0.61 and 0.78, respectively. In all cases, fine-scale (3 m) lidar-derived DEMs worked better than USGS 10 m DEMs and, in general, including soil properties improved correlations.

Highlights

Soil moisture is a key variable controlling a host of important hydrological and biogeochemical processes and, imposes a considerable ecohydrological fingerprint on the landscape
A comparison of the means and overall distributions of Akaike information criterion (AIC) values reveals that lidar-based topographic wetness indices (TWIs) consistently provide a better fit to the patterns of observed soil moisture than USGSbased TWIs across the full range of parameter combinations (Fig. 2)
The AIC distribution of the lidar data set is substantially greater than the United States Geological Survey (USGS), indicating that the different parameter combinations had a greater influence on the performance of lidar TWIs

Summary

Introduction

Soil moisture is a key variable controlling a host of important hydrological and biogeochemical processes and, imposes a considerable ecohydrological fingerprint on the landscape. The detailed numerical approach is typically incorporated into distributed hydrologic modeling frameworks and has been shown to provide reasonable simulations of soil moisture patterns (Frankenberger et al, 1999; Motovilov et al, 1999; Mehta et al, 2004; Cuo et al, 2006). Such models often require extensive data input and calibration, are generally prohibitively complex for conservation planners to use (Lane et al, 2006; White et al, 2010) and frequently suffer from equifinality issues (Beven, 2006)

Methods

Results

Conclusion