Estimation of grassland biophysical parameters using hyperspectral reflectance for fire risk map prediction

Abstract

In remote sensing, the reflectance of vegetation has been successfully used for the assessment of grassland biophysical parameters for decades. Several studies have shown that vegetation indices that are based on narrow spectral bands significantly improve the prediction of vegetation biophysical characteristics. In this work, we analyse the relationships between the biophysical parameters of grasslands and the high-spatial-resolution hyperspectral reflectance values obtained from helicopter platform data using both a spectral vegetation index and a regression approach. The regression approach was favoured as it had optimal results with respect to producing higher R² values than the spectral index approach (water content, 0.91 v. 0.90; leaf-area index, 0.88 v. 0.61; and green ratio, 0.90 v. 0.83). These three parameters were selected to obtain a fire risk map for the Bosco della Fontana grassland areas. The extreme spatial variability of the fire risk confirmed the hypotheses regarding the importance of obtaining scale-appropriate biophysical maps to model fire risk in fragmented landscapes and ecosystems. More studies are needed in order to investigate both the limits and the opportunities of high-spatial-resolution sensors in highly fragmented landscapes for the remote detection of fire risk and to generalise the obtained results to other grassland vegetation types.