GIScience & Remote Sensing (Dec 2022)

Model-driven estimation of closed and open shrublands live fuel moisture content

  • Gengke Lai,
  • Xingwen Quan,
  • Marta Yebra,
  • Binbin He

DOI
https://doi.org/10.1080/15481603.2022.2139404
Journal volume & issue
Vol. 59, no. 1
pp. 1837 – 1856

Abstract

Read online

Live fuel moisture content (LFMC) is a crucial variable affecting the ignition potential of shrublands. Different remote sensing-based models (either empirical or physical) have been adopted to estimate LFMC in shrublands but with mixed success potentially owing to differences in vegetation cover (closed vs. open shrublands). This study aimed to evaluate and discuss LFMC estimation in open and closed shrublands using different remote sensing approaches. For each case, three broadly used radiative transfer models (RTMs) (PROSAILH, PROGeoSail, and PROACRM), and two empirical models were selected and compared. The empirical models were calibrated by a stepwise regression approach using a spectral index (SI) and its normalized form (SImax-min). Results showed that both RTMs and empirical models performed well in retrieving LFMC of closed shrublands (RTMs: R2 = 0.60–0.66, RMSE = 14.96–18.51%, bias = −5.99–4.36%, and empirical models: R2 = 0.69–0.72, RMSE = 10.67–11.30%, bias = 0.18–0.35%). However, all RTMs failed to retrieve LFMC for open shrublands (R2 = 0.01–0.09, RMSE = 45.21–48.66%, bias = 9.76–14.75%) potentially due to the high heterogeneity of vegetation in this vegetation type. In contrast, the SImax-min-based model outperformed the RTMs for the open shrublands LFMC estimation (R2 = 0.44, RMSE = 32.32%, bias = −0.34%) but saturated at high LFMC values (> 120%). In conclusion, PROACRM and the empirical model using SImax-min as an explanatory variable are recommended to model closed and open shrublands LFMC, respectively. This study gives insights into developing effective models for improving shrubland LFMC estimation by considering various fractions of covers of shrublands that were not previously considered.

Keywords