Erosion Susceptibility Assessment using Statistical Models at a Tropical Watershead in Costa Rica
Guardado en:
| Autores: | , |
|---|---|
| Formato: | artículo original |
| Estado: | Versión publicada |
| Fecha de Publicación: | 2017 |
| Descripción: | The aim of this study is to assess the state of conservation of a tropical watershead in Costa Rica in terms of its capacity to provide the ecossytem service of control of erosion. Firstly a spatial inventory of erosion occurrence areas was made by the interpretation of aerial photographs (CARTA Project). It was previously made under the condition that its spatial distribution is not random but depends on the complex interaction of natural and human factors. The methodology applied to quantify that spatial relationship was based on the stadistical analysis of erosion susceptibility. To do that it were applied two comparative methods: 1) a multivariate logistic regression and 2) a joint condicional probability model under Bayesian theorem. The two alternatives were combined with an indirect bivariate statistical analysis base on the weights of evidence method. The positive weight of evidence was assigned to each of the different classes into which a factor map is classified: land use, the slope gradient, geomorphology, and distance to streams. The results of the calculation of the weights of evidence and the interpretation of the coeficientes of the ecuation of logistic regression demostrated that the most relevant factor map was the land use. The both susceptibility maps were evaluated through two different methodologies: the success rate and the ROC curve. The validation results revealed that the joint conditional model was slightly better at predicting erosion features than the logistic regression model. |
| País: | Portal de Revistas UNA |
| Institución: | Universidad Nacional de Costa Rica |
| Repositorio: | Portal de Revistas UNA |
| Lenguaje: | Español |
| OAI Identifier: | oai:ojs.www.una.ac.cr:article/10110 |
| Acceso en línea: | https://www.revistas.una.ac.cr/index.php/ambientales/article/view/10110 |
| Palabra clave: | ponderaciones de evidencia regresión logística binaria probabilidad condicional conjunta curva ROC Sistemas de Información Geográfica (SIG). weights of evidence modelling multivariate logistic regression joint conditional probability ROC curve Geographical Information Systems (SIG) ponderações de evidência regressão logística binária probabilidade condicional conjunta Sistemas de Informação Geográfica (SIG) |