Exploiting the capabilities of Sentinel-2 and RapidEye for predicting grass nitrogen across different grass communities in a protected area

Abstract

The primary indicator of forage quality in any rangeland environment is grass nitrogen. Remote sensing has been used to map grass quality by predicting nitrogen in order to facilitate effective management and conservation of rangeland ecosystems. Using Telperion Game Reserve as the study site, this study sought to understand the extent to which the application of Sentinel-2 (S2) MSI sensor, together with new red edge bands and vegetation indices can improve the estimation accuracy of grass nitrogen (N) at a landscape scale when compared to the RapidEye sensor. This study was particularly interested in understanding if animal distribution in a rangeland environment is significantly correlated to grass quality across different grass communities at a landscape level. The performances of the simple ratios (SR), normalised vegetation difference indices (NDVI) and random forest (RF) regression models were evaluated and used to map and predict grass nitrogen concentration. The results indicate that the model developed using the SR indices could predict nitrogen with a mean R2 of 0.92 for S2 data and 0.53 from RapidEye data using combined models. The results also showed that nitrogen maps from remote sensing could be used to explain animal aggregation and grazing patterns at a landscape level.