Capability of models to predict leaf N and P across four seasons for six sub-tropical forest evergreen trees

Nutrient phenology of evergreen subtropical forests of southern Africa is poorly understood. Foliar nitrogen (N) and phosphorous (P) forms key components of photosynthesis and are vulnerable to global change stressors. Remote sensing techniques can potentially map and monitor nutrient phenology, yet models to predict across species, seasons and climatic regions are deficient. This study evaluates the capability of various models, developed from leaf spectra of selected spectral regions and seasons, to predict nutrient concentration across season and species. Seasonal differences in foliar N and P were assessed using a one-way ANalysis Of VAriance (ANOVA). The relationship between leaf spectra and nutrients was assessed using linear regressions between the foliar nutrients and spectral indices. The predictive capability of three models was compared using root mean square error (RMSE) values. Amongst the four seasons, winter leaves showed the highest mean N (2.16%, p < 0.01). However, winter showed the lowest variability of foliar N (coefficient of variation = 8%) compared to the variability of the other three seasons (coefficient of variance > 35%). In fact, between winter and spring, the variability in foliar N increased by 294%. Foliar P did not significantly differ between the four seasons. Predictive models for leaf N concentration developed for each season showed a higher level of accuracy, particularly for winter, whereas predictive models for leaf P showed low accuracies. Models developed from a single season showed a slight increase in error for the summer and autumn, however a larger increase in error for the winter season for the evergreen trees. The results suggest that spectral measurements can be potentially be used to quantify nutrient phenology at regional scale and monitor the impacts of global change on nutrient phenology and photosynthesis.