SPATIO-TEMPORAL DESCRIPTION OF THE RAINFALL FOR COLOMBIAN ANDEAN MOUNTAINOUS REGION FOR WEATHER FORECASTING PURPOSES. CASE STUDY: MANIZALES - CALDAS, COLOMBIA

Abstract:

Manizales is a city located in west-central Colombian Andes in the Caldas province, whose spatial location coincides with one of the most threatened areas of Colombia (landslides, earthquakes, volcanic eruptions, other).

As a middle Andean mountainous city and for being located in the area of influence of the ITCZ presents an equatorial mountain climate with a bimodal rainfall regime, and with an average annual rainfall around 2000 mm, it shows very significant rates of precipitation, on average, 70% of the days of the year it is rainy. This situation favors the formation of large masses of clouds and the presence of macroclimatic phenomena such as ENSO, which has historically caused large-scale impacts in both warm and cold phase.

Since last decade different entities have implemented a hydro-meteorological network which measures and transmits telemetrically every five minutes hydro-climatic variables. In general, the real-time weather monitoring should be used for a better understanding of our environmental urban environment and to establish indicators of quality of life and welfare for the community.

Despite the city has telemetric data on atmospheric and hydrological variables, there is still no tool or a methodology able to generate a spatio-temporal description of these variables. So, the aim of this work is to establish guidelines to sort all this information of atmospheric variables monitored in real time with the help of data mining techniques, machine learning tools to improve the knowledge of atmospheric patterns at Manizales and to serve for territorial planning and decision makers.

To reach this purpose the current data warehouse available at the National University of Colombia at Manizales will be used, and it will be fed with observed variables from hydro-meteorological monitoring stations that transmit in real-time. Then, as mentioned this information will make the corresponding processing with data mining techniques to describe the rainfall patterns. All this complemented with the application of statistical techniques for data analysis and exploration.

The main contribution of this research is the creation of tools to be used in numerical modeling with forecasting purposes, aiming to improve the resolution given by mesoscale models, which are currently used for weather forecast in Colombia.