Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Research Project #434702

Research Project: Mapping Water Use, Stress and Productivity in Agricultural Landscapes by Fusing Multi-Sensor Data Products

Location: Hydrology and Remote Sensing Laboratory

Project Number: 8042-13610-029-29-I
Project Type: Interagency Reimbursable Agreement

Start Date: May 1, 2018
End Date: Apr 30, 2019

Objective:
The ultimate goal in this project is to prototype a multi-source remote sensing system with operational potential for global agricultural monitoring at field scale. The fused multi-source data products from this project will be used to accomplish the following tasks: (1) Map daily water use and stress at field scale by fusing evapotranspiration (ET) timeseries developed from MODIS/VIIRS/Landsat/Sentinel-2 satellite data. (2) Map daily vegetation indices at field scale as an indicator of crop progress and biomass accumulation. (3) Extract phenological metrics at field scale from the daily vegetation index timeseries. (4) Combine field-scale stress, vegetation index and phenological data to estimate yield in a prototype application for improved operational yield monitoring/forecasting tools. (5) Map water productivity (WP) at field-to-regional scales to investigate variability in agricultural water use efficiency and evaluate coarser resolution global WP assessments.

Approach:
This proposed project aims to prototype methods for routine production of high spatiotemporal resolution ET, vegetation index (VI) and derived phenology products for crop monitoring applications using a multi-sensor data fusion approach. This approach fuses moderate resolution/near-daily retrievals of ET and surface reflectance (SR) from sensors like MODIS and VIIRS with periodic finer scale data from Landsat, Sentinel-2 (S2), ECOSTRESS and other Landsat-like sensors to generate multi-year timeseries of gridded products at daily timesteps and 30-m spatial resolution. ET will be estimated using a well-established surface energy balance algorithm based on thermal infrared (TIR) retrievals of land-surface temperature. For high-resolution sensors like Sentinel-2 that lack TIR imagers, a novel thermal sharpening approach will be employed to sharpen 375m TIR data from VIIRS on the sensor overpass days using multi-band SR data. Similarly, ECOSTRESS TIR data will be supplemented by fused SR datastreams. Collectively, these high spatiotemporal resolution “datacubes” provide valuable field-scale diagnostics of water use, moisture stress, phenology and biomass accumulation required for monitoring agricultural production systems and forecasting yield. The accuracy of these products will be evaluated over diverse agricultural landscapes, including both crop and rangelands in the U.S. and internationally. ET retrievals will be compared with flux tower measurements, and performance will be assessed both in terms of standard statistical metrics and ability to capture episodic changes in moisture conditions resulting from, e.g., rainfall, irrigation, harvest, and rapid stress onset. VI data and derived phenological metrics will be assessed at full resolution in comparison with biophysical data collected in-field, and at larger scales using county and state-level crop progress reports. Finally, we will demonstrate utility of the 30-m daily ET/VI datasets and derived phenological parameters for operational assessments. Cloud computing technique will be tested for the large area mapping. Improvement in moisture stress and yield mapping capabilities will demonstrated over highly managed agricultural systems in the U.S., Brazil, Czech Republic and Lebanon – where performance of coarser resolution timeseries are known to be degraded due to mixed pixel effects. We will also map water use and water productivity (“crop-per-drop”) over these landscapes to quantify differences in efficiencies between regions, climates and cropping systems. This study will motivate the value to agriculture of routine, daily multi-sensor satellite products developed at Landsat scale, where vegetation water use and development can be differentiated by crop type and land management practice.