Author(s): Eric Sandi; Rafael Oreamuno-Vega; Steven Sandi

Linked Author(s): Eric Sandi , Steven Sandi

Keywords: Water resources management; Abangares; Neural network; SAC-SM

Abstract: Definition and implementation of water use plans require the use of diverse approaches that allow for a better description of the water availability. The use of long-term hydrologic data series is ideal for defining current trends and changes to water availability in response to management practices, but unfortunately, such long-term data are not always available and in some regions data is almost a luxury. The use of modelling techniques can help overcome the limitations of scarce data as they provide a useful tool for exploring different management scenarios and help in the definition of water use plans, but at the same time good quality data is also necessary to setup such models. The Abangares River catchment, located in the northwest region of Costa Rica, presents a significant challenge for water resources management. The use of water for human consumption and agriculture is managed by the local government that must also ensure environmental water for ecologically protected areas within the catchment. In the past decades, water scarcity would commonly occur in the catchment, particularly during the dry season (December to May) when rainfall abruptly decreases. Significant efforts to improve the water management and planning of the catchment have been carried out over the last decade, with instruments for hydrologic data collection being installed and continuous research aimed at improving the water infrastructure and hydrologic description of the catchment. In this contribution, we present the calibration of a daily runoff model (SAC-SMA) for one of the sub-catchments of the Abangares River making the best use of the data recently collected in the catchment. Data analysis revealed gaps and deficiencies in rainfall data, limiting the implementation of the hydrologic model. To overcome this, we applied a neural network algorithm to reconstruct missing rainfall data using observed data series from 2015 to 2019. The model was then calibrated for the period from 2015 to 2017, which corresponds to the available runoff data periods. While the implementation of this model is yet to be validated as more data becomes available over the next few years, the model represents a keystone for the water management of the catchment, especially given the pressures of climate change and variability. The period of 2015 to 2017 was one of the driest periods in the region. We discuss the implications of the use of such modeling tools in locally governed catchments and the importance of improving such techniques in the face of climate change.

DOI: https://doi.org/10.3850/IAHR-39WC2521711920221344

Year: 2022