Prediction models for estimating available fodder of two savanna tree species (Acacia dudgeoni and Balanites aegyptiaca) based on field and image analysis measures

Abstract

Browse production is difficult to measure non-destructively without some level of subjectivity combined with a lack of accuracy and reproducibility. This study examined the possibility of using ground-based photographs to estimate browse production. Thirty five sample trees of Acacia dudgeoni and Balanites aegyptiaca were selected. Tree crown area (CA) was measured from two-dimensional images using computer-based image analysis. Browse biomass from destructive sampling was correlated with CA and measured tree variables. The results from the regression analysis indicated that all models were significant (all p values < 0.05) for the two species but the predictive power was low for Acacia (r² < 0.50) compared to Balanites (r² > 0.75). The establishment of a relationship between browse and crown area (CA) estimated from photographs or the field measured crown area (PCA) indicated that the fit of the relationship was better for PCA (r² adjusted = 0.75) compared to CA (r² adjusted = 0.73) for Balanites. For Acacia, regression coefficients for CA and PCA were 45% and 33%, respectively. The image analysis technique could offer an objective method for estimating biomass changes by browsing, lopping and seasonal leaf fall when coupled with dendrometric measures.