Techniques for monthly rainfall regionalization in southwestern Colombia

Abstract

The knowledge of rainfall regimes is a relevant requirement for many activities such as water resources planning, risk management, agriculture activities management, and other hydrologic applications. The present study has consisted of validating four techniques (one linear, one non-linear, and two hybrids) that allow identifying homogenous regions. We take the monthly rainfall in the Southwestern Colombia (Nariño), an area of 33,268 km 2 characterized by complex topography and local factors that can infl uence the rainfall behavior, to test all techniques. The results showed overall the best performance for the approach related to non-linear principal component analysis and self-organizing map. However, in all mainly prevail two regions: the Andean Region and Pacifi c Region with a bimodal and unimodal regime, respectively. The bimodal one dominates over the Andes mountains range and the unimodal one the coastal zone. The application of non-linear approaches provided a better understanding of the seasonality of rainfall, and the results may be useful for water resource management.

Estimation of Monthly Reference Evapotranspiration with Scarce Information Using Machine Learning in Southwestern Colombia

“…The southwest of Colombia (Nariño) is one of the most biodiverse regions of the country and the world, located between 0°21' and 2°40' north latitude and 76°50' and 79°02' west longitude, with approximately 33,268 km2 (Canchala et al, 2020;Ocampo-Marulanda et al, 2022). The Pacific region (14,754 km2) accounts for 52 %, the Andean region (15,466 km2) for 40 % and the Amazon region (3,048 km2) for the remaining 8 % (See Figure 1) (Gobernación de Nariño, 2019).…”

“…For this reason, models to estimate ETo have been developed during the last decades, using machine learning techniques such as Artificial Neural Networks (ANN) (Zanneti et al, 2008;Alves et al, 2017;Fonseca et al, 2018;Laqui et al, 2019;Meneses et al, 2020) Autoregressive Integrated Moving Average Model (ARIMA) (Jordan et al, 2008;Gautam and Sinha, Mossad and Alazba, 2016;Bouznad et al, 2020), and Multiple Linear Regression (MLR) (Yirga, 2019). In Colombia, the spatial distribution of climatological stations is uneven due to several factors, such as complex topography (e.g., the Andean Mountain range), areas affected by the armed conflict, and low investment in technological resources, among others (Urrea et al, 2019;Canchala et al, 2022). In southwestern Colombia (Nariño), 76 % of the rainfall stations are in the Andean region, with a density of one station every 470 km2, covering 40 % of the total area of Nariño.…”

“…of one station every 1,442 km2, accounting the 52 % of the total area of Nariño. In the case of climatological stations, the scenario is worse since, in the Andean region, there is one every 1,720 km2 and only one in the Pacific region (Barbacoas) (Ocampo-Marulanda et al, 2022). The rainfall stations are those available Instituto de Hidrología, Meteorología y Estudios Ambientales (IDEAM).…”