Modelamiento numérico de la distribución espacio temporal del permafrost en el Perú mediante el uso de variables topo-climáticas
Cargando...
Código QR
Autores
Villavicencio Guillen, Eduardo Emer
Correo de contacto
Resumen
Con el objetivo de determinar la cobertura de permafrost mediante el modelamiento numérico de su distribución espacio temporal en el Perú mediante las variables topo climáticas, para lo cual en esta tesis se usaron las variables del Índice Normalizado de Vegetación (NDVI por sus siglas en ingles), pendiente, radiación solar, temperatura del aire, temperatura del suelo, orientación y altitud, generando dos modelos, donde los resultados obtenidos muestran que el primer modelo de permafrost denominado exploratorio (90 m) utilizó como variables de identificación al NDVI, radiación solar y la altitud debido a que sus valores de acuerdo a la distribución de Poisson son inferiores a 0.005 y la correlación entre estas variables es inferior a 0.5, el modelo llegando a tener un valor de curva ROC AUC de 0.81 y una validación cruzada 0.74. Por su parte el segundo modelo de permafrost denominado avanzado (30 m) tuvo una divergencia en los datos es por ello que de acuerdo a la distribución de Poisson la pendiente y radiación solar superan el valor de 0.005, siendo este el motivo de un mayor análisis, donde la radiación solar es una variable determinante en la identificación del permafrost como lo demostraron Boeckli, Gruber, Marcer entre otros (Boeckli et al. 2012, Gruber 2012, Marcer et al. 2017), lo que conllevo a realizar un análisis más profundo del conjunto de variables determinando el desarrollo de dos submodelos, el primer submodelo consideró las variables de NDVI, pendiente, temperatura, radiación solar y altitud (NPTRA), por su parte el otro submodelo consideró las variables de orientación, NDVI, temperatura y altitud (ONTA) donde la comparación de ambos submodelos dieron el valor de la curva ROC-AUC en el modelo NPTRA fue de 0.72 mientras que para el modelo ONTA fue de 0.73, por último en la validación cruzada obtuvieron valores de 0.68 y 0.67 respectivamente. Por lo cual se determinó que el modelo NPTRA es el adecuado en la generación del modelo, de este modo se obtuvo las áreas probables de permafrost, se obtuvo que la mayor presencia de permafrost está en las regiones de Arequipa, Cusco, Apurímac y Ayacucho mientras que en la región de Ancash presentan zonas de permafrost más pequeñas.
With the objective of determining permafrost coverage through numerical modeling of its spatiotemporal distribution in Peru using topo-climatic variables, this thesis employed variables such as the Normalized Difference Vegetation Index (NDVI), slope, solar radiation, air temperature, soil temperature, aspect, and altitude. Two models were developed, and the results showed that the first permafrost model, referred to as the exploratory model (90 m), used NDVI, solar radiation, and altitude as identification variables. These variables were selected because their Poisson distribution values were below 0.005, and their correlations were less than 0.5. The model achieved a ROC-AUC curve value of 0.81 and a cross-validation score of 0.74. The second permafrost model, referred to as the advanced model (30 m), exhibited data divergence. According to the Poisson distribution, slope and solar radiation exceeded the 0.005 threshold, prompting further analysis. Solar radiation was identified as a key variable in permafrost identification, as demonstrated by studies such as those by Boeckli, Gruber, and Marcer (Boeckli et al., 2012; Gruber, 2012; Marcer et al., 2017). This led to a deeper analysis of the variables, resulting in the development of two sub-models. The first sub model considered the variables NDVI, slope, temperature, solar radiation, and altitude (NPTRA), while the second sub-model included aspect, NDVI, temperature, and altitude (ONTA). Comparative analysis of these sub-models showed that the ROC-AUC curve value for the NPTRA model was 0.72, while the ONTA model achieved 0.73. Cross-validation scores were 0.68 and 0.67, respectively. Ultimately, the NPTRA model was determined to be the most suitable for generating the permafrost model. This model indicated the probable permafrost areas, with the highest permafrost presence in the regions of Arequipa, Cusco, Apurímac, and Ayacucho, while smaller permafrost zones were identified in the Ancash region.
With the objective of determining permafrost coverage through numerical modeling of its spatiotemporal distribution in Peru using topo-climatic variables, this thesis employed variables such as the Normalized Difference Vegetation Index (NDVI), slope, solar radiation, air temperature, soil temperature, aspect, and altitude. Two models were developed, and the results showed that the first permafrost model, referred to as the exploratory model (90 m), used NDVI, solar radiation, and altitude as identification variables. These variables were selected because their Poisson distribution values were below 0.005, and their correlations were less than 0.5. The model achieved a ROC-AUC curve value of 0.81 and a cross-validation score of 0.74. The second permafrost model, referred to as the advanced model (30 m), exhibited data divergence. According to the Poisson distribution, slope and solar radiation exceeded the 0.005 threshold, prompting further analysis. Solar radiation was identified as a key variable in permafrost identification, as demonstrated by studies such as those by Boeckli, Gruber, and Marcer (Boeckli et al., 2012; Gruber, 2012; Marcer et al., 2017). This led to a deeper analysis of the variables, resulting in the development of two sub-models. The first sub model considered the variables NDVI, slope, temperature, solar radiation, and altitude (NPTRA), while the second sub-model included aspect, NDVI, temperature, and altitude (ONTA). Comparative analysis of these sub-models showed that the ROC-AUC curve value for the NPTRA model was 0.72, while the ONTA model achieved 0.73. Cross-validation scores were 0.68 and 0.67, respectively. Ultimately, the NPTRA model was determined to be the most suitable for generating the permafrost model. This model indicated the probable permafrost areas, with the highest permafrost presence in the regions of Arequipa, Cusco, Apurímac, and Ayacucho, while smaller permafrost zones were identified in the Ancash region.
Descripción
Universidad Nacional Agraria La Molina. Escuela de Posgrado. Maestría en Recursos Hídricos
Palabras clave
Permafrost
Citación
Fecha
2025
Colecciones
Seleccionar año de consulta:
Licencia de uso

Excepto si se señala otra cosa, la licencia del ítem se describe como info:eu-repo/semantics/openAccess