Precipitation forecast in a region of Ecuador by inclusion

The prognosis of precipitation represents a scientific challenge of great relevance due to its complexity and countless applications it can have. Precipitation is a phenomenon that is determined by several space -temporal physical factors. Consequently, for your forecast it is important to introduce both spatial and temporal information of the relevant variables. Sometimes, even consider teleconexions (interrelation between distant phenomena). To this end, in this project, convolutional neuronal networks will be used that consider spatial information around the point in which the prognosis is wanted to make, as well as temporary information. This information on climatic and meteorological variables will be obtained from satellite data and reanalysis models, which will be used for training of the aforementioned neural networks. In the first instance, the precipitation prognosis for a single pixel will be carried out taking into account the spacetime data of a region surrounding said pixel (punctual prognosis). From the results of this first phase, the forecast will be carried out in a region, consisting of several pixels, that is, a sequence of future precipitation maps (regional forecast) will be obtained. The results of this project can be used to benefit the population in several areas: agriculture, livestock, energy production, drinking water supply, risk management for droughts and floods, tourism, etc. These potential applications show the relevance of this research.

20th UNIVERSITY RESEARCH PROJECT COMPETITION