Estimación del contenido de humedad de vegetación mediterránea a partir de imágenes MODIS
- Autores: Marta Yebra Álvarez
- Directores de la Tesis: Emilio Chuvieco Salinero (dir. tes.)
- Lectura: En la Universidad de Alcalá ( España ) en 2008
- Idioma: español
- Tribunal Calificador de la Tesis: Fernando Valladares Ros (presid.), María Inmaculada Aguado Suárez (secret.), Francisco Javier Salas Rey (voc.), Juan Ramón de la Riva Fernández (voc.), F. Mark Danson (voc.)
- Menciones:
- Europea
- Calificación: Sobresaliente Cum Laude
- Materias:
- Enlaces
- Tesis en acceso abierto en: e_Buah
-
Dialnet Métricas: 1 Cita
- Resumen
- español
La presente tesis doctoral se centra en la estimación operativa del contenido de humedad de combustibles vivos (LFMC) en regiones de clima mediterráneo utilizando información obtenida de imágenes captadas por el sensor Terra-MODIS, en el ámbito de la estimación de condiciones de peligro de incendio. En concreto, se plantea como alternativa a los tradicionales modelos empíricos o índices meteorológicos, el uso de modelos de simulación de la reflectividad (RTM), puesto que estos últimos poseen una mayor capacidad de generalización y rigor físico. Así, se exploran diferentes formas de adaptación de los parámetros de entrada de estos modelos a dos tipos de combustibles mediterráneos (pastos y matorral) y varias técnicas de inversión, con el propósito final de conseguir una metodología operativa, precisa, robusta y generalizable de estimación de LFMC. Las estimaciones de LFMC con los RTM son siempre contrastadas con las obtenidas mediante modelos empíricos calibrados con datos de humedad tomados en campo, con el objetivo de verificar si los primeros son más idóneos que los segundos. Finalmente, se analizan las relaciones entre el LFMC de pastizales y matorrales con la probabilidad de ocurrencia de incendio (número de incendios, total de superficie quemada y ocurrencia de grandes incendios) con el objetivo de resaltar el interés de esta variable como indicadora de ocurrencia de incendios en zonas mediterráneas. Con esta tesis doctoral se demuestra que los RTM tienen mayor potencial que los modelos empíricos e índices meteorológicos para la estimación operativa de LFMC a gran escala, ya que, correctamente parametrizados, conducen a estimaciones más robustas en diferentes zonas de validación. Todas las conclusiones están respaldadas por once años de trabajo de campo, llevado a cabo sistemáticamente y bajo protocolos estandarizados en diferentes zonas dominadas por vegetación mediterránea.
- English
This doctoral thesis is focussed on the operational estimation of Live Fuel Moisture Content (LFMC) of Mediterranean vegetation species from Terra-MODIS imagery in the context of fire risk assessment. The thesis is organised around seven papers. The first paper (in Spanish) is an exploratory analysis of the potential of meteorological data and Terra-MODIS data to assess LFMC by examining the correlations between those datasets and field LFMC. The paper demonstrates that among the meteorological fire danger indices and satellite derived indices that are examined, the latter produce better correlations with LFMC and, therefore, Terra-MODIS data is potentially more suitable for the estimation of LFMC of the vegetation species studied. The second paper (in English) deals with the estimation of LFMC with Terra-MODIS data by comparing the interest of empirical relations and of statistical models based on simulated reflectances derived from radiative transfer models (RTM). The paper demonstrates that both models produce good results, but that the RTM present a greater interest as they produce more robust estimations, and consequently are expected to have a greater generalization power. This hypothesis is tested in the third paper (in Spanish), with ground measurements obtained at other areas dominated by grasslands in Central Spain and Australia. The hypothesis is confirmed as both models provide similarly accurate results for grasslands with analogous types of vegetation, and RTM are more accurate than empirical models when applied to grasslands with structure and composition different from the calibration site. Once RTM have demonstrated their interest, remains the difficulty to adequately parameterise these models. The fourth (in English), fifth and sixth (in Spanish) papers propose to use ecological rules to avoid simulating unrealistic spectra, where the parameters present values never met simultaneously on the field. The fourth paper deals with woodlands, addressing the problem of fixing parameters of RTM and demonstrating that a choice of co-occurring observed parameters provides better results. The fifth paper analyses how the different parameters used in the RTM co-evolve in shrublands species which leads to the designation of reduced and realistic co-evolution ranges. This approach proves its usefulness by significantly decreasing the residual estimation error when compared to models run with unrestricted ranges. In the sixth paper, the RTM are parameterized using some of the criteria explained in fifth paper, and two merit function to compare observed and simulated spectra (relative minimum quadratic distance and minimum spectral angle) are explored. This paper concludes that the use of the spectral angle criteria provides a more consistent measure of spectral similarity which leads to a more accurate estimation of LFMC. The estimations with these criteria are also compared with the estimations derived from the shrublands model calibrated in paper two for shrublands (empirical and RTM) based on field data measured in four different locations of Spain., concluding that the estimations have been highly improved and that the RTM have greater generalization potential are therefore they are more suitable for LFMC at global scale. The last paper (in English), examines the interest of LFMC as a predictor of fire occurrence in a Mediterranean ecosystem. It demonstrates that LFMC is a good predictor of number of fires, total burned areas and occurrence of large fires. Three different logistic regression models are then built to predict these three dimensions of fire occurrence. The discussion concludes that the proposed models provide significant results but do not provide sufficiently accurate predictions at all periods of low or high fire risk since, as expected, fire occurrence also depends on many other factors. After the seven papers there is an extra chapter (chapter 8) which deals with the methodology which should be followed to generate the cartography of LFMC by mean of the models calibrated and tested in this thesis and presents some examples. As a conclusion, in this thesis three methods (meteorological data, empirical models and RTM) have been considered to operatively assess LFMC over a reasonably large area, with the advantages and disadvantages of these three methods then compared. The potential of using RTM is demonstrated and a way to improve on the problem of parameterisation of these models is proposed and validated, making use of the ecological rules governing the relationships between the different parameters and imposing restrictions to their co-variations. The thesis ends by confirming the potential of the LFMC parameter for fire occurrence prediction. All this work is supported by field work, systematically carried out along different periods, used for the calibration and validation steps.
- español
