Location: Dale Bumpers Small Farms Research Center
Project Number: 6020-21500-001-027-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Sep 1, 2026
End Date: Aug 31, 2028
Objective:
The major goal is to develop dynamic functional soil moisture maps with high accuracy and precision for field-scale precision agricultural applications using a combination of direct measurements, proximal sensor, remote sensing, machine learning and artificial intelligence. Precision Agriculture has been identified as one of the major priorities for the ARS national programs. Many agriculture input resources such as fertilizers, water, insecticides/pesticides, machinery, etc., are spent annually by farmers. The effectiveness of these inputs is primary driven by the availability of water, especially moisture stored in the soil profile during the growing season. Consistent with the objective of this research, the project is to develop daily interactive soil moisture maps at high accuracy that respond to weather drivers such as precipitation, evapotranspiration and soil characteristics among many factors. The identification of major factors and drivers of the in-field variability of soil water dynamics, especially soil moisture, would better inform management decisions in support of daily farm operations. The project will link performance of various crops, mostly yield with spatial and temporal availability and variability soil moisture.
Negative Impact:
Currently, most of the agricultural inputs are applied uniformly throughout fields leading to under or over application rates. This in turn increases cost of production, causing economic harm and negatively impacts resources and environment. The lack of such dynamic soil moisture maps that support a better and more effective use of the agricultural inputs leads to great losses of resources estimated in millions and billions of dollars. It will also not take advantage of current developments in multiple data collection platforms, powerful machine learning and data analysis platform such as SciNet to fully explore the information locked in many years of data collected by government and other research institutions over decades throughout the US and internationally.
Approach:
Multiple soil moisture data collection methods from direct measurement, in-situ sensors, proximal and remote sensing will be used in combination with distributed hydrological models to create high resolutions spatial and temporal soil moisture maps at various depths. High resolution input data for weather, soils, and vegetation will be developed for running distributed hydrological models for generating high spatial temporal soil moisture maps. The maps will be validated with direct soil moisture measurements when available and in-situ soil moisture sensors. In addition, yield maps generated from harvesting combines will be used to assess the usefulness and accuracy of soil moisture maps and develop functional soil water movement maps. The distributed hydrological models are complex, data driven with high computational needs. SciNet platform will be utilized to overcome these challenges. Machine learning and artificial intelligence approaches will be used to train the distributed hydrological models with readily available data from satellites that are widely used and available for larger areas. The digital twin approach will be used to allow for an expansion of the usability of such models for wide agricultural applications throughout the country. The maps can be also available through online platforms and apps.