Publication: TELEDETECCIÓN SATELITAL Y ESPECTROMETRÍA DE CAMPO COMO HERRAMIENTAS PARA LA ESTIMACIÓN DEL CONTENIDO DE HUMEDAD DE LA ESPACIE PINUS RADIATA EN LA V REGIÓN DE VALPARAISO.
Date
2017
Authors
VALDIVIA GAVILÁN, JORGE ALBERTO
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El presente trabajo abarca la intersección de dos áreas de la ciencia: la teledetección satelital y laespectrometría de campo, con la finalidad de estimar el FMC (Fuel Moisture Content), indirectamente,a partir de modelos estadísticos con índices tanto a nivel híperespectral (espectrometría) utilizandomodelos de transferencia radiativa (PROSPECT) para ampliar el rango espectral, como índices espectralesa nivel satelital. El FMC es una de las variables más relevantes al momento de evaluar lagestión del riesgo de incendios forestales, explicando variables críticas como el proceso de ignición yla propagación del fuego.La obtención de muestras de vegetación viva (pinus radiata) se realizó en la Reserva Lago Peñuelasde la CONAF (V Región de Valparaíso) durante el verano del 2017. Estas muestras se analizaron enel laboratorio EC2G (UTFSM, Casa Central) donde se obtuvo el FMC, el EWT (Equivalent WaterThickness), el DMC (Dry Matter Content), el espectro de reflexión comprendido entre los 650–1.100nm, entre otras variables. Todas las pruebas se realizaron un día después del paso del satélite, con el finde corroborar que la imagen no tuviese ruido atmosférico en la zona bajo análisis y así hacer el procesomás eficiente. Se simuló el espectro reflectivo de la vegetación comprendido entre los 400–2.500 nmutilizando PROSPECT, con el fin de contar con un rango más amplio de índices espectrales.Las variaciones del FMC obtenidos en laboratorio de las 11 muestras estudiadas variaron entre los122,19% a los 157,05% con una media de 139,25% de humedad de los combustibles para el períodode muestreo. A su vez, se empleó una metodología para determinar el FMC a partir del cociente entreel EWT y DMC a nivel de hoja (obtenidos en laboratorio), que consistió en determinar el peso fresco,seco y área de 10 agujas representativas de pino de la muestra total, donde se obtendría, de esta forma,dichas variables para una aguja de pino representativa de la muestra. El rango de valores para el FMCmetodológico varió entre 123,30% y 154,15% con una media de 135,74 %. El FMC con respecto a lametodología empleada se explicó con una correlación de R2=0,76, donde el FMC metodológico tendióa subestimar los valores del FMC medido. Se calcularon los índices para determinar, indirectamente,los valores del EWT y DMC a nivel de hoja. En el caso del EWT los mejores índices fueron el NMDI(R2=0,872) y el WI (R2=0,869) con RMSE de 0,001899 (g=cm2) y 0,001921 (g=cm2), respectivamente.En el caso del DMC se propone el WI como el mejor índice, donde el modelo se explicó con unacorrelación de R2=0,933 y RMSE de 0,000988 (g=cm2). A nivel satelital, tanto el EWTLeaf como elDMCLeaf se multiplicaron por el LAI (Leaf Area Index), que se obtuvo del producto MCD15A2Hdel sensor Modis de NASA, con el fin de obtener sus valores a nivel del dosel. Así, el EWTCanopyy DMCCanopy obtuvieron sus mejores resultados con el NDII (R2=0,837) y el NBR (R2=0,8356),respectivamente, donde los valores RMSE fueron de 0,00682 (g=cm2) para el EWTCanopy y 0,00477(g=cm2) para el DMCCanopy.Aplicando los modelos del EWT y DMC se obtiene la estimación del FMC a partir de informaciónespectral. Tanto en el caso híperespectral como satelital el modelo se ajustó de mejor forma alcomparar los valores con el FMC determinado por la metodología empleada. En el caso del FMC porespectrometría se obtuvo un RMSE de 14,20% y 13,73% para el FMC medido en el laboratorio y elFMC metodológico, respectivamente. En el caso satelital, se obtuvo un RMSE de 10,80% y 8,50%para el FMC medido en el laboratorio y el FMC metodológico, respectivamente. En ambos casos elmodelo tendió a generar una mayor dispersión de los datos experimentales.
The present work covers the intersection of two science areas: remote sensing and field spectrometryof vegetation with the objective of estimating the FMC (Fuel Moisture Content), indirectly, fromstatistical models with both hyperspectral (spectrometry) using radiative transfer models (PROSPECT)to expand the spectral range, and multispectral (satellite spectral indices). The FMC is one of the mostrelevant variables at the moment of evaluating the risk management of forest fires, explaining criticalvariables such as the time to ignition and fire behaviour.The collection of samples of live vegetation (pinus radiata) was held in the Reserva Lago Peñuelasof CONAF (V Region of Valparaíso, Chile), during the summer of 2017. These samples were analyzedat the EC2G laboratory (UTFSM, Casa Central) where the FMC, EWT (Equivalent Water Thickness),DMC (Dry Matter Content), the reflective spectrum between 650–1,100 nm, among other variables,were obtained. All sampling was done a day after the passage of the satellite in order to corroborate thatthe image did not have atmospheric noise in the area under analysis and make the process more ecient.The reflective spectrum of vegetation was simulated between 400–2,500 nm using PROSPECT inorder to have a broader range of spectral indices.The variations of the FMC obtained in the laboratory of the 11 samples under study varied between122.19% to 157.05% with an average of 139.25% moisture content of the fuels in the samplingperiod. At the same time, a methodology for the determination of the FMC from the quotient betweenthe EWT and DMC at leaf level (obtained in the laboratory) was calculated, which consistedin measuring the fresh weight, dry weight and area of 10 pine needles, representative of the totalsample, with the objective to estimate the same variables for one pine needle, which would representan average pine needle from the total sample. The range of values for the methodological FMC variedbetween 123.30% and 154.15% with an average of 135.74 %. The relation of FMC with respect tothe methodology used (FMCMetodología) was explained with a correlation of R2=0.76, where the lattertended to underestimate the measured FMC values. Spectral indices were calculated to determineindirectly the values of the EWT and DMC at leaf level. In the case of the EWT, the best indiceswere the NMDI (R2=0.872) and the WI (R2=0.869) with RMSE of 0.001899 (g=cm2) and 0.001921(g=cm2), respectively. In the case of the DMC, the proposed index is the WI, where the model is explainedwith a correlation of R2=0.933 and RMSE of 0.000988 (g=cm2). At satellite level, both EWTLeafand DMCLeaf were multiplied by the Leaf Area Index (LAI), obtained from NASA’s Modis productMCD15A2H, in order to obtain those values at canopy level. Thus, the estimation of EWTCanopy andDMCCanopy had its best results with the NDII (R2=0.8371) and NBR (R2=0.8356) indices, respectively.The RMSE values were 0.00682 (g=cm2) for the EWTCanopy and 0.00477 (g=cm2) for the DMCCanopy.The FMC is equal to the quotient of EWT and DMC. Therefore, by applying the models of both, theEWT and DMC, the FMC can be estimated from hyperspectral and satellite indices. In both cases, theestimated FMC adjusted better with the methodological FMC. The FMC estimated from hyperspectralinformation had RMSE values of 14,20% and 13,73% for the measured FMC and methodologicalFMC, respectively. At satellite level, the FMC model obtained RMSE values of 10.80% and 8.50%for the measured FMC and methodological FMC, respectively. In both cases, the model tended togenerate a greater dispersion of the experimental data.
The present work covers the intersection of two science areas: remote sensing and field spectrometryof vegetation with the objective of estimating the FMC (Fuel Moisture Content), indirectly, fromstatistical models with both hyperspectral (spectrometry) using radiative transfer models (PROSPECT)to expand the spectral range, and multispectral (satellite spectral indices). The FMC is one of the mostrelevant variables at the moment of evaluating the risk management of forest fires, explaining criticalvariables such as the time to ignition and fire behaviour.The collection of samples of live vegetation (pinus radiata) was held in the Reserva Lago Peñuelasof CONAF (V Region of Valparaíso, Chile), during the summer of 2017. These samples were analyzedat the EC2G laboratory (UTFSM, Casa Central) where the FMC, EWT (Equivalent Water Thickness),DMC (Dry Matter Content), the reflective spectrum between 650–1,100 nm, among other variables,were obtained. All sampling was done a day after the passage of the satellite in order to corroborate thatthe image did not have atmospheric noise in the area under analysis and make the process more ecient.The reflective spectrum of vegetation was simulated between 400–2,500 nm using PROSPECT inorder to have a broader range of spectral indices.The variations of the FMC obtained in the laboratory of the 11 samples under study varied between122.19% to 157.05% with an average of 139.25% moisture content of the fuels in the samplingperiod. At the same time, a methodology for the determination of the FMC from the quotient betweenthe EWT and DMC at leaf level (obtained in the laboratory) was calculated, which consistedin measuring the fresh weight, dry weight and area of 10 pine needles, representative of the totalsample, with the objective to estimate the same variables for one pine needle, which would representan average pine needle from the total sample. The range of values for the methodological FMC variedbetween 123.30% and 154.15% with an average of 135.74 %. The relation of FMC with respect tothe methodology used (FMCMetodología) was explained with a correlation of R2=0.76, where the lattertended to underestimate the measured FMC values. Spectral indices were calculated to determineindirectly the values of the EWT and DMC at leaf level. In the case of the EWT, the best indiceswere the NMDI (R2=0.872) and the WI (R2=0.869) with RMSE of 0.001899 (g=cm2) and 0.001921(g=cm2), respectively. In the case of the DMC, the proposed index is the WI, where the model is explainedwith a correlation of R2=0.933 and RMSE of 0.000988 (g=cm2). At satellite level, both EWTLeafand DMCLeaf were multiplied by the Leaf Area Index (LAI), obtained from NASA’s Modis productMCD15A2H, in order to obtain those values at canopy level. Thus, the estimation of EWTCanopy andDMCCanopy had its best results with the NDII (R2=0.8371) and NBR (R2=0.8356) indices, respectively.The RMSE values were 0.00682 (g=cm2) for the EWTCanopy and 0.00477 (g=cm2) for the DMCCanopy.The FMC is equal to the quotient of EWT and DMC. Therefore, by applying the models of both, theEWT and DMC, the FMC can be estimated from hyperspectral and satellite indices. In both cases, theestimated FMC adjusted better with the methodological FMC. The FMC estimated from hyperspectralinformation had RMSE values of 14,20% and 13,73% for the measured FMC and methodologicalFMC, respectively. At satellite level, the FMC model obtained RMSE values of 10.80% and 8.50%for the measured FMC and methodological FMC, respectively. In both cases, the model tended togenerate a greater dispersion of the experimental data.
Description
Catalogado desde la version PDF de la tesis.
Keywords
ESPECTROMETRIA , FMC , TELEDETECCION