Predicting seed yield of moist-soil plants

Estimates of seed production are important for basic and applied ecological research, calculation of carrying capacity of waterfowl foraging habitats, and evaluation of wetland management. M. K. Laubhan and L. H. Fredrickson predicted seed yield of moist-soil plants using phytomorphological measurements and multiple linear regression (Journal of Wildlife Management 56:329-337). We evaluated their general methodology in Mississippi, tested if different morphological measurements would predict seed mass accurately and precisely, and evaluated their significant predictor variables in our models. We measured phytomorphological variables of 6 moist-soil plant species in 1993 and 1994. Significant variation in seed mass was explained by our models (P < 0.001, R-adj(2) = 0.78-0.97); however, our models never contained the same variables as Laubhan and Fredrickson's models. Models containing significant (P less than or equal to 0.01) variables detected by Laubhan and Fredrickson, but developed using our data, had lower predictability and greater collinearity than our models. Although multiple regression models may adequately predict seed yield of moist-soil plants, our results suggested inconsistency in predictor variables, model site-specificity and need for a simpler predictive approach.