Satellite surveillance of soil moisture in farms.

Project description

Satellite surveillance of crops is probably the link that best strengthens the agribusiness chain. The use of satellite monitoring and online analytics fills gaps in traditional auditing methods. Satellite exploration of thousands of farms in a region increases efficiency and lowers costs, and is applicable to all types of crops. The analysis of historical satellite images allows the comparison between the current growing season with that of previous years, to detect or anticipate possible problems in the current season.

For the development of this project, three data sources were used, including the optical images of Sentinel 2, the VV polarization maps of Sentinel 1, and the climatic variables of NOAA. The developed application assumes soil moisture as the most important variable for satellite surveillance, because it is an extremely critical variable for countries with arid soils and little annual precipitation, such as Chile in Latin America. This variable is complemented by the NDVI index which has proven useful in assessing crop health and other crop conditions and is also understood by many farmers.

A satellite model of soil moisture has been created for each NOAA cell, using historical NOAA data and historical Sentinel 1 data, over a full crop cycle, to associate NOAA predicted moisture with Sentinel 1 VV reflectance at the NOAA cell. The data is sampled over time with the frequency of satellite visits. This model created from temporal information is applied at the spatial level to all the farms located in the same NOAA cell.

A second moisture model is customized for each farm, using spatial disaggregation techniques on satellite data, in this procedure the previous crop cycle is used to generate the farm moisture model that will be applied to the current cycle. These two models are combined to predict farm moisture over time, from the NOAA reading every six hours.

The figure above shows the image of the NDVI coefficient of a farm, and the two forecast curves for surface moisture (red curve) and 40 cm depth (blue curve), for the next 5 days. The figure below shows the moisture image calculated from the VV reflectance channel, and the surface soil moisture curve during the last year (red curve), and the soil moisture curve at 40 cm depth (blue curve). The two models explained have been used to generate the soil moisture curves.

The two examples show a great difference between soil moisture in both farms, the top farm has maximum moisture values in the order of 0.17 and the bottom farm 0.46. We conclude that satellite surveillance of crop soil moisture is most important where it is most needed.