Modeling the Agricultural Drought Vulnerability of Pennsylvania Soils

Open Access
Mebane, Valerie Jean
Graduate Program:
Soil Science
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • Rick Lane Day, Thesis Advisor
  • crop model
  • agricultural drought
  • SSURGO soils
Catastrophic crop losses in Pennsylvania emphasize the need for an enhanced understanding of drought and its effects on agriculture. Agricultural drought occurs when there is inadequate soil moisture to meet a plant’s needs, ultimately resulting in a reduction in crop yields. Consequently an accurate drought vulnerability assessment tool should have the ability to account for the sensitivity of crops to soil moisture deficit conditions during critical stages of crop development. Unfortunately many drought indices and assessment methods widely used today fail to provide an accurate assessment of soil variability across the landscape and may neglect the interaction of crops on soil moisture reserves. The goal of the research presented in this paper was to develop a long-term agricultural drought risk assessment tool for Pennsylvania that was applicable at the field scale to help farmers make educated decisions based on their potential vulnerability to agricultural drought. The assessment tool was developed using field based variables and a crop growth model and was structured to have limited data requirements to allow for a regional and statewide implementation. Because the effects of soil moisture stress on crop yield are specific to the crop variety and the stage of growth during which the stress occurs, this research focused specifically on agricultural drought risk associated with the effects of soil moisture stress on corn. The AquaCrop crop simulation model was utilized to simulate the effects of soil moisture stress on corn production under a variety of soil and climate conditions over a span of thirty years. Although the model was already validated under diverse environmental conditions, a validation study was conducted to evaluate AquaCrop’s ability to capture the relationship between corn growth and the effects of soil moisture stress under Pennsylvania conditions. Data obtained from previous studies conducted in Rock Springs, PA and Landisville, PA were utilized to evaluate AquaCrop’s ability to simulate the progression of cumulative biomass and grain yield with time, final biomass and harvestable yield, and volumetric water content at six depths. The results indicated that AquaCrop was able to accurately simulate the progression of cumulative biomass and grain yield with time, with index of agreement values ranging from 0.96 to 0.99. Comparisons between simulated and measured final biomass and final harvestable yield produced biomass deviations ranging from 2.4 to 20.7%, and yield deviations of 2.9 and 15.3%. The water balance evaluation indicated that the model showed a tendency to underestimate soil water content at shallower depths. However, for a model intended to assess a location’s potential vulnerability to drought, a bias towards the underestimation rather than overestimation of soil water content was preferred to encourage the adoption of appropriate drought avoidance practices. Averaged across all depths the results were consistent with other validation studies of soil water balance models, with RMSE ranging from 1.5 to 9.8 vol%. SSURGO (Soil Survey Geographic) data were used to satisfy the crop model’s requirement for horizon-specific detailed soil characteristics for each location where corn growth was simulated. However, the vast quantity of soils contained within the SSURGO database can make it extremely difficult to evaluate all soils at the field scale in a regional or statewide modeling analysis. A second study was conducted to develop a structured approach of reducing the dimensionality of the SSURGO soils data while still preserving data variability. Cluster analysis was used to group the 637 agricultural SSURGO soils mapped within the nine county study region using seventeen soil properties considered important for an assessment of a soil’s vulnerability to agricultural drought. Forty-three clusters were formed and found to be satisfactory using the evaluation criteria of cubic clustering criterion, pseudo F, and pseudo t2 statistics. A preliminary principal component analysis was conducted to transform the original seventeen correlated variables into a smaller number of uncorrelated variables to be entered into the cluster analysis. The first four principal components explained 85.98% of the total variance. A representative soil was selected for each cluster based on the smallest Euclidean distance to the cluster centroid. Comparisons of simulated biomass yield were made between the clusters to assess differences and similarities in the behavior of the clusters in an agricultural setting. Certain clusters showed minimal differences while others showed considerable differences, providing an indication that the soil clusters were able to represent the variability of the original soils in terms of biomass yield. The final study utilized the AquaCrop model validated in the first study and the representative soils determined in the second study to simulate the effects of soil moisture deficit conditions for each soil under a variety of climate conditions over a span of thirty years. The effects of soil moisture stress on corn production were quantified to develop a worst-case scenario baseline drought vulnerability index with the ability to capture the vulnerability of Pennsylvania’s diverse soils to agricultural drought. An index representing conditions likely to occur once every four years was mapped over a subset of nine Pennsylvania counties to provide farmers with a tool to detect potential problematic conditions. A relationship was developed between the index and relative yield reductions to put the index in terms more meaningful to farmers. The relative yield reductions likely to occur once in four years were then mapped over the nine county study area. The results demonstrated that the drought vulnerability of a given location stems from a complex interaction of both climate and soil properties.