Agriculture In Uruguay: New Methods For Drought Monitoring and Crop Identification Using Remotely Sensed Data

Wednesday, 17 December 2014
Jerrod Lessel, NASA DEVELOP National Program, International Research Institute for Climate and Society, Palisades, NY, United States and Pietro Ceccato, International Research Institute for Climate and Society, Columbia University, Palisades, NY, United States
Agriculture is a vital resource in the country of Uruguay. Here we propose new methods using remotely sensed data for assisting ranchers, land managers, and policy makers in the country to better manage their crops. Firstly, we created a drought severity index based on the climatological anomalies of land surface temperature (LST) data from the Moderate Resolution Imaging Spectroradiometer (MODIS), precipitation data from the Tropical Rainfall Monitoring Mission (TRMM), and normalized difference water index (NDWI) data also using MODIS. The use of the climatological anomalies on the variables has improved the ability of the index to correlate with known drought indices versus previously published indices, which had not used them. We applied various coefficient schemes and vegetation indices in order to choose the model which best correlated with the drought indices across 10 sites throughout Uruguay’s rangelands. The model was tested over summer months from 2009-2013. In years where drought had indeed been a problem in the country (such as 2009) the model showed intense signals of drought. Secondly, we used Landsat images to identify winter and summer crops in Uruguay. We first classified them using ENVI and then used the classifications in an ArcMap model to identify specific crop areas. We first created a polygon of the classifications for soils and vegetation for each month (omitting cloud covered images). We then used the crop growing cycle to identify the times during the year for which specific polygons should be soil and which should be vegetation. By intersecting the soil polygons with the vegetation polygons during their respective time periods during the crop growing cycle we were able to create an accurately identify crops. When compared to a shapefile of proposed crops for the year the model obtained a kappa value of 0.60 with a probability of detection of 0.79 and a false alarm ratio of 0.31 for the south-western study area over the 2013-2014 summer.