Assessment of the Standardized Precipitation and Evaporation Index (SPEI) as a Potential Management Tool for Grasslands

Early warning of detrimental weather and climate (particularly drought) on forage production would allow for tactical decision-making for the management of pastures, supplemental feed/forage resources, and livestock. The standardized precipitation and evaporation index (SPEI) has been shown to be correlated with production of various cereal and vegetable crops, and with above-ground tree mass. Its correlation with above-ground grassland or forage mass (AGFM) is less clear. To investigate the utility of SPEI for assessing future biomass status, we used biomass data from a site on the Konza Prairie (KP; for years 1984-1991) and from a site at the United States Department of Agriculture-Agricultural Research Service's (USDA-ARS) Grazinglands Research Laboratory (GRL; for years 2009-2015), and a publicly-available SPEI product. Using discriminant analysis and artificial neural networks (ANN), we analyzed the monthly timescale SPEI to categorize AGFM into above average, average, and below average conditions for selected months in the grazing season. Assessment of the confusion matrices from the analyses suggested that the ANN better predicted class membership from the SPEI than did the discriminant analysis. Within-site cross validation of the ANNs revealed classification errors ranging from 0 to 50%, depending upon month of class prediction and study site. Across-site ANN validation indicated that the GRL ANN algorithm better predicted KP AGFM class membership than did the KP ANN prediction of GRL AGFM class membership; however, misclassification rates were 25% in all months. The ANN developed from the combined datasets exhibited cross-validation misclassification rates of 20% for three of the five months being predicted, with the remaining two months having misclassification rates of 33%. Redefinition of the AGFM classes to identify truly adequate AGFM (i.e., average to above average forage availability) improved prediction accuracy. In this regard, results suggest that the SPEI has potential for use as a predictive tool for classifying AGFM, and, thus, for grassland and livestock management. However, a more comprehensive investigation that includes a larger dataset, or combinations of datasets representing other areas, and inclusion of a bi-weekly SPEI may provide additional insights into the usefulness of the SPEI as an indicator for biomass production.