Sentinel-2 satellites map crop-water use over Aussie farms

Coala Project > All posts in COALA Project > Blog > Sentinel-2 satellites map crop-water use over Aussie farms

Sentinel-2 satellites map crop-water use over Aussie farms

  • University of Melbourne
  • Blog
  • No Comments

European Union’s Horizon 2020-funded international project, called COALA, produced a suite of crop growth, health and water use information. We compiled this information using high-resolution multispectral images collected from space by the Sentinel-2 satellites. The data we used was collected over irrigated cropping fields in the Goulburn Valley, Victoria during the 2020-2021 summer growing season.

Although growers in Australia had a relatively wet La Niña summer in 2021, water is still one of the major costly resources for irrigated crop growers. COALA’s high-resolution crop water demand and crop water use maps aim to help growers improve water use efficiency. We provide data to support the right amount of irrigation applied at the right time.

Map Time Explorer

Our partners in Ariespace (Italy) and BOKU University (Austria) developed the Map Time Explorer (MTE). The MTE provides a handy visual interface streamlining value-added information products derived from the Copernicus Sentinel-2 satellites. MTE enables quick access to high-resolution maps of crop information and their timeseries over locations chosen by the users. We provide the widely used Normalised Difference Vegetation Index (NDVI). And, users can access various other derived products that utilise the whole range of spectral bands of Sentinel-2 satellites. The MTE crop information layers produced for 2020-2021 summer season can be categorised as:

  • Basic mapping layers: RGB true colours, RGB false colours (aka colour infrared)
  • Crop water information: crop evaporation, rainfall
  • Crop biomass and growth: NDVI, leaf area index (LAI), fractional cover (FCover)
  • Crop photosynthetic activity: fraction of absorbed photosynthetically active radiation (fAPAR), albedo
  • Soil water status: shortwave infrared transformed reflectance (STR), normalised difference moisture index (NDMI)

Validation of satellite-based daily crop water use measurement

While MTE offers various crop information, growers were primarily interested in crop evapotranspiration as a measure of crop water consumption. The near-real-time crop evapotranspiration maps we provide to growers are for individual paddocks. MTE of this temporal and spatial resolution, support growers in optimal water ordering and accurate application of irrigation. Our maps have applications in both summer and winter cropping. But, the near-real-time information provided is more critical for summer cropping when growers need to make daily irrigation decisions.

COALA MTE overcomes the typical data latency of weeks to months for satellite-derived services by directly accessing raw imagery in the Copernicus system. We then internally undertake a suite of pre-processing, and automating value-added product generation. Pilot users in the Goulburn Valley, Victoria were able to access the daily outputs during the summer cropping season.

Our results

The majority of the pilot farms we worked with grew maize, pasture and lucerne. The University of Melbourne team collected biomass, soil water and, most importantly, crop evapotranspiration data during the cropping period. The plot below shows a comparison of crop evapotranspiration generated by COALA MTE and ground-truth values collected by an eddy-covariance system installed in a pilot maize field from December 2020 to April 2021.

During the cropping season, maize crops consumed water with a wide range of daily variation between 2 mm/day and 10 mm/day. And, COALA MTE accurately depicted the daily crop water use with an RMSE (root mean squared error) of 1.41 mm/day. The key weather input data for COALA’s MTE are currently from the Bureau of Meteorology’s AWAP dataset, interpolated from very sparse weather station networks in the pilot region. These results showcase that we have developed one of the best products existing in the region.  

Figure 1. Comparison of remotely sensed crop evapotranspiration by COALA’s MTE and ground-based measurements using an eddy-covariance system in a pilot maize field during Januray-April in 2021.

High resolution crop information

We continuously monitored crop photosynthetic activity and evapotranspiration which demonstrated how we can use high-resolution crop information to detect sub-paddock-scale crop variation. Eventually, we can use the data to predict yield. Below we show a pilot maize field in the fAPAR map. The fAPAR map shows noticeable sub-paddock-scale variability of crop photosynthetic strength from mid-February. The variable crop photosynthesis was caused by variable soil texture and resulting fertiliser variation in the soil within the paddock. The MTE timeseries tools depicted lower crop water consumption over the weak fAPAR patches from the early stage of the growing season, in December 2020 – January 2021 (see image below). The sub-paddock-scale low and high maize water consumption patches coincide with the final yield of the field, in their respective patches. The areas that showed weaker water consumption from the early stage of the season produced significantly less yield.

The COALA team is continuing to explore the MTE outputs, ground data and pilot grower’s feedback from the summer growing season. We are working to see if the within-paddock crop water use patterns can assist growers in taking management actions to improve final yield uniformly over whole paddocks.

By Dongryeol Ryu, Arash Parehkar, Danlu Guo, Andrew W. Western, Oscar Belfore, Guido D’Urso, and Carlo De Michele

Leave a Reply