Remote sensing of aboveground grass biomass between protected and non‐protected areas in savannah rangelands

Abstract

AbstractAccurate information on the amount of aboveground grass biomass (AGBg) stored in savannah rangelands remains unknown. Therefore, in this study we assessed the capability of a new multispectral sensor (Landsat‐8 OLI) in estimating aboveground grass biomass stored within two different land use management units in savannah rangelands using the stepwise multiple linear regression methods. Specifically, we evaluated the performance of different Landsat‐8 OLI derivatives (i.e. (a) spectral bands, (b) derived vegetation indices and (c) a combination of spectral bands and vegetation indices) to quantify aboveground grass biomass. The results highlighted that the use of spectral bands as stand‐alone model variables yielded considerably high accuracies in terms of the coefficient of determination (r2) and the root mean square error (RMSE). An r2 of 0.73 (72.81%) and RMSE of 37.88% for the protected area and an r2 of 0.75 (75.18%) and with a RMSE of 33.16% for the nonprotected area were attained. The use of vegetation indices, however, demonstrated a very weak model performance yielding an r2 of 0.31 (30.68%) and a RMSE of 51.51%, for the protected area and an r2 of 0.50 (50.43%) and RMSE of 40% for the nonprotected areas. Comparatively, combining Landsat‐8 OLI derived spectral bands and vegetation indices demonstrated improved model performance, yielding an r2 of 0.92 (91.90%), and RMSE of 41.3% for the protected area and an r2 of 0.94 (93.82%) with a RMSE of 33.14% for the nonprotected area. Therefore based on the observed results, AGBg in savannah rangelands can be satisfactorily estimated using broadband multispectral derivatives.