Development of a Multivariate Model Focused on the Analysis of Water Availability in Mexico

The present study proposes developing a multivariate model that predicts water availability in Mexico through 26 variables related to aquifers, renewable water, demographic characteristics, rivers and basins, dams, and irrigation factors. The information inherent to them was extracted from the platform of the national water system using records from the 13 administrative hydrological regions between 2010 and 2017. The model is based on the multiple linear regression model and the variable selection method. The results show different versions of the model contrasted concerning the statistical assumptions of the multiple regression. Although the findings presented have implications in the development of strategies focused on a better distribution of the vital liquid, in the face of various projected scenarios based on the variables analyzed, it should be noted that the progressive improvement of the model was carried out through the use of techniques such as the transformation of variables, detection, and elimination of outliers. The final result is water availability in the face of various drought conditions explained by a model with 16 relevant variables. Said prediction model is helpful for the generation of drought mitigation strategies.

Automated summary

The study proposes a multivariate model to predict water availability in Mexico based on 26 variables, with 16 relevant variables identified in the final model. The approach uses multiple linear regression model and variable selection techniques, with progressive improvement through the use of variable transformation, outlier detection, and elimination. The prediction model is helpful for generating drought mitigation strategies.