Floods are the most common type of naturaldisaster , accounting for more than half of the total amount of reportednatural disasters in 2006 (OFDA/CRED International Disaster Database,Université catholique de Louvain – Brussels – Belgium), however countriesin regions such as South America keep very little amount of historicaldata about these disasters, which impacts the ability of the government tooffer help to the ones that most need it.

We set two goals for our solutionto achieve, the first goal was to obtain an indicator to determine whichcountries had vulnerable population, our second goal was to develop a modelthat asses the risk of flooding of given area, depending on variousindicators from nasa's available datasets.

In order to achieve our firstgoal we recovered datasets on GDP per capita, Mortality Rate, Inflation Rate,Internet usage and Crime of South American countries from http://data.un.org, to form country groupsfrom this data and classify the countries we used a K means clusteringalgorithm to form clusters, after this process we ended up with a clearseparation of countries with high vulnerability and countries more resilient.

To achieve our secondgoal we used data from SMAP L3 Radiometer Global Daily 36 km EASE-GridSoil Moisture V005 in order to collect data about soil moisture ,soilmoisture/water content and ground temperature in order to develop a modelthat would have these indicators as an input and a flood risk level as anoutput, this is where we faced our first difficulty due to the lack of solidand detailed historical data on floods in the region, this prohibited usfrom doing supervised training, to overcome this challenge we decided to use aK means clustering Algorithm to group the different regions, and later useother years datasets and some historical data recoveredfrom OFDA/CRED to confirm our models effectiveness.