Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Estimating critical grassland vegetation moisture parameters using topoclimatic variables and remotely sensed data in relation to fire occurrence

Auteur : Shinga, Wenzile.

Université de soutenance : University of KwaZulu-Natal

Grade : Master Degree (Geography) 2021

Résumé
Quantifying grassland Fuel Moisture Content (FMC) and Equivalent Water Thickness (EWT) is critical for establishing early fire-warning systems as well as encouraging proactive fire management strategies. This also facilitates the preservation of grassland functions such as carbon sequestration under the influence of climate change. Fire danger has been monitored using local weather information from multiple stations, which is tedious and lacks spatial representation. Meanwhile, using remote sensing and statistical algorithms to estimate grass moisture elements such as FMC and EWT could facilitate a better understanding of fire regimes even in inaccessible areas. In this regard, this work sought to i) assess the utility of topo-climatic and Sentinel 2 Multispectral Instrument (MSI) satellite data in estimating grass FMC and ii) to estimate EWT using Sentinel 2 MSI derived variables in the rangelands of Southern Africa using the Random Forest (RF) algorithm. Results of this study showed that FMC could be estimated to an R 2 and RMSE of 0.68 and 0.039 % m2 , respectively. The optimal variables in this model were channel networks and elevation. In estimating EWT using Sentinel 2 MSI variables only, the RF results yielded an R 2 and RMSE of 0.75 and 0.019 g/m2 , respectively. The important variables identified using RF were Modified Normalised Difference Vegetation Index (NDVI) (Short Wave Infra-red (SWIR)1/Band 2), Band 2, Soil Adjusted Vegetation Index, and the Modified Simple Ratio (SWIR1/Band 2). This study demonstrates the prospects of utilizing Sentinel 2 MSI satellite remotely sensed and top-climatic data in estimating EWT and FMC as fire risk indicators in South African grasslands.

Présentation

Version intégrale (3,0 Mb)