APPLICATION OF REMOTE SENSING TECHNIQUES TO OBTAIN DETAILED SOIL MOISTURE DATA IN LARGE AREAS

Estimating soil moisture (SM) has an important role in diverse hydrological processes because of its importance as a variable in the hydrological cycle. The correct estimation and the understanding of its dynamic temporal-space changes within the basin under study is critical while developing meteorological and climate predictive models in flood warning systems, allowing improved calibration in the runoff models that contribute to formulate and plan for prevention policies both for flooding prevention and mitigation and early alert of droughts and global warning.

However, measuring SM in the field is not frequent due to the fact that in-situ techniques require much effort and resources, while it is also frequent that they are not carried in a routine way because the large areas to cover and heterogeneous nature of the soils. The resarch seeks to capitalize from current offer of remote sensing products to develop a model to estimate the SM through the processing of satellite imaging; in this particular case using Landsat 5 TM.

The methodology developed to obtain SM from these satellite images correlates soil moisture data taken with time-space variability in a pilot plot located 2km northeast of Rosario city (Argentina), using reflectance data from Landsat 5 TM images taken the same day and time that the soil samples were taken in the plot for correlation. After applying digital imaging processing to the satellite data, and using statistic regression technique, the data was analyzed verifying different sets of variables and combinations. Finally, it was possible to obtain a satisfactory statistical model (r²=0.88) using the variables NDVI, NDWI, bands 1, 4 and 5.

The model was later applied to the basin area, where the net runoffs are calculated from rain data taken from a network of monitoring stations during storm events, by using an existing hydrological–hydraulical model and compared with the real runoffs measured during those events, concluding that the data obtained through the model helps predict accurately the runoff during those events. In addition, the amount of data obtained from the satellite image by applying the model is more detailed and dense than field campaigns to collect data manually, allowing better accuracy and lower costs to collect and process the SM data.