Near-infrared spectroscopy for estimating soil burn severity

• Autor(a) principal: Pedrosa,Elisabete Trindade
• Data de Publicação: 2012
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/10708
• topic: Estudos ambientais; Incêndios florestais; Temperatura; Solos; Espectroscopia de infravermelhos

Forest fires are a natural phenomenon and occurred long before human kind was around, serving important ecosystem functions. In the past decades, however, some parts of the world have seen marked increases in the frequency and spatial extent of wildfires. This includes Portugal, where forest fires have, on average, affected 100.000 ha of rural lands per year since the mid1970s. In general, the direct and indirect effects of fires depend strongly on the temperatures to which vegetation and soil are exposed. In the case of wildfires – as opposed to prescribed burning or experimental fires - these temperatures can hardly ever be measured. Therefore, wildfire impacts are commonly assessed using proxies based on the consumption of the vegetation and the colour of the ashes deposited on the soil surface. These so-called burn severity indices typically provide qualitative estimates, distinguishing between low, medium and high severity. Recently, however, near- infrared (NIR) reflectance spectroscopy was successfully applied to estimate the maximum temperatures reached (MTR) by soils heated under laboratory conditions. The present study wanted to explore the potential of NIR for estimating MTR in soils burnt by wildfires. To this end, the work addressed two main topics: (i) spatial variability in the relationships between soil heating temperatures in a muffle and the corresponding NIR-based MTR estimates, both between and within study sites; (ii) the importance of this spatial variability in estimating MTRs of wildfire-burnt soil samples. A number of NIR-based models was constructed and used to predict the known MTR of laboratory-heated soil samples. One of the two long-unburnt study sites revealed marked variability over short distances, whereas the other did not. The models based on larger sample numbers, however, provided robust MTR predictions, even when these models involved samples from the two study sites. This probably reflected the sites comparable parent materials, soils and land cover (eucalypt plantations in schist soils). The best achieved models were used to estimate MTR by soil samples from a wildfire occurred in the central-north of Portugal, in the year 2010.. According to the index proposed in this work and the maximum temperatures reached estimations, the soil burn severity of the studied sites was moderate to high in surface samples, and low to moderate in the sub-surface samples.