Abstract

MARTINEZ, Alejandro et al. Forecasting precipitation over the tropical Andes: Lessons from convection-permitting simulations. Rev. acad. colomb. cienc. exact. fis. nat. [online]. 2024, vol.48, n.186, pp.145-168.  Epub Jan 24, 2024. ISSN 0370-3908.  https://doi.org/10.18257/raccefyn.1965.

We evaluated 625 daily precipitation forecasts over parts of the Colombian Andes using the Weather Research and Forecasting (WRF) model between 2020 and 2022. We run the simulations in "convection-permitting" mode using a grid spacing of 4 km. WRF's performance was compared with forecasts from the Global Forecasting System (GFS) and observational data. On average, WRF's forecasts produced a pattern of nighttime maxima within the inter-Andean valleys and lowlands like that of the Global Precipitation Measurement Mission (GPM). According to the WRF model, daytime precipitation mostly occurred over the mountains. Compared to GPM, the WRF model overestimated precipitation in the mountains and underestimated it in the lowlands. In the GFS model, the biases and the absolute precipitation were smaller in magnitude than in the WRF. Correlation values between daily precipitation in the GPM and the forecasts of both models tended to be higher during dry seasons, with maximum values of 0.62 and 0.70 in the WRF and the GFS, respectively. The indices based on contingency tables were similar in the WRF and the GFS, with Bias, FAR, and POD values close to 1.1, 0.10, and 0.96, respectively. The evaluation of heavy rainfall events in two places in Antioquia established that the WRF and several global models provided forecasts up to 12 hours in advance. However, higher-resolution simulations are better for representing structures and gradients typical of the complex terrain of the tropical Andes.

Keywords : Weather forecast; Mountain regions; Mesoscale meteorology; Numerical models; WRF.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )