Relating Crop Yield Patterns to Terrain Attributes Under Water-Limited and Waterlogged Conditions

Authors: Erskine, Robert - Rob; Green, Timothy; Starr, Gordon

Submitted to: International Conference on Precision Agriculture Abstracts & Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 12/7/2007
Publication Date: 7/21/2008
Citation: Erskine, R.H., Green, T.R., Starr, G.C. 2008. Relating Crop Yield Patterns to Terrain Attributes Under Water-Limited and Waterlogged Conditions. International Conference on Precision Agriculture Abstracts & Proceedings.

Interpretive Summary: Terrain attributes derived from high-resolution digital elevation models (DEMs) can be useful for explaining spatial patterns of soil moisture and crop yields. Assuming landscape topographic controls on soil moisture variability, we correlated soil moisture and crop yield with a suite of terrain attributes (elevation, slope, aspect, curvature, potential solar radiation index, specific catchment area, and wetness index) on rolling agricultural fields in Colorado and Maine. Centimeter-level accurate global positioning system (GPS) methods were employed to produce 5-m grid DEMs. Soil moisture, as measured by time domain reflectometry (TDR) over the top 0.2 m in Maine and top 0.3 m in Colorado, was correlated to terrain attributes at both locations. Soil moisture was positively correlated with curvature (concave upward being positive), specific catchment area, and wetness index, while it was negatively correlated to elevation, slope, cos(aspect), and potential solar radiation index. Maximum correlations were typically observed with wetness index. Winter wheat crop yields as measured by a yield monitor in Colorado, where water-limited conditions prevailed, were correlated with the terrain attributes in a similar manner as soil moisture. In Maine, potato crop yields were measured by hand-harvested plots and by a yield monitor. Maximum correlations were observed with elevation, now a positive correlation, while yield was negatively correlated topographic curvature, specific catchment area, and wetness index due to waterlogged conditions. Correlations were higher based on the hand-harvested plots than with the yield monitor. Correlations with yield monitor data were improved by normalizing the data to the mean yield of each individual swath, which removes variability caused by yield bias or monitor errors affecting entire swaths. This cross-comparative study provides insight to the landscape processes controlling soil moisture and yield variability in two different environments.

Technical Abstract: Terrain attributes derived from high-resolution digital elevation models (DEMs) can be useful for explaining spatial patterns of soil moisture and crop yields. Assuming landscape topographic controls on soil moisture variability, we correlated soil moisture and crop yield with a suite of terrain attributes (elevation, slope, aspect, curvature, potential solar radiation index, specific catchment area, and wetness index) on rolling agricultural fields in Colorado and Maine. Centimeter-level accurate global positioning system (GPS) methods were employed to produce 5-m grid DEMs. Soil moisture, as measured by time domain reflectometry (TDR) over the top 0.2 m in Maine and top 0.3 m in Colorado, was correlated to terrain attributes at both locations. Soil moisture was positively correlated with curvature (concave upward being positive), specific catchment area, and wetness index, while it was negatively correlated to elevation, slope, cos(aspect), and potential solar radiation index. Maximum correlations were typically observed with wetness index, with r = 0.61 in Maine and r = 0.40 in Colorado. Winter wheat crop yields as measured by a yield monitor in Colorado were correlated with the terrain attributes (maximum r = -0.53 with slope), where water-limited conditions prevailed. In Maine, potato crop yields were measured by hand-harvested plots and by a yield monitor. Maximum correlations were observed with elevation, now a positive correlation (r = 0.51), while yield was negatively correlated topographic curvature, specific catchment area, and wetness index due to waterlogged conditions. Correlations were higher based on the hand-harvested plots than with the yield monitor. Correlations with yield monitor data were improved by normalizing the data to the mean yield of each individual swath, which removes variability caused by yield bias or monitor errors affecting entire swaths. This cross-comparative study provides insight to the landscape processes controlling soil moisture and yield variability in two different environments.