A remote sensing-based approach to management zone delineation in small scale farming systems
In collaboration with researchers from the James Hutton Institute, China Agricultural University, the University of Minnesota and Auburn University (USA), ADAS has developed an approach for dividing smallholder farming villages in the North China Plain into management zones for more precise agronomic prescriptions using remotely sensed data. The work has recently been published in the journal Agronomy.
Small-scale farming systems (<0.5ha per farm) represent about 80% of the farmed area in China, and the over-application of Nitrogen fertiliser is common. Science-based evidence is therefore needed to help farmers make more efficient decisions in relation to nutrient management. Management zones are defined as sub-regions within the field that have similar combinations of yield-limiting factors and are managed accordingly. In the context of small-scale farming systems, field sizes are too small to divide them into management zones, but at a village scale, this approach is possible, particularly in these wheat and maize growing areas of China where fields are separated by just a narrow ridge of soil (see image above). Understanding the factors that lead to the spatial and temporal variability of a crop within the village is the first step for optimal agronomic management.
The approach that we took used a ten-year time series of optical images from around the flowering stage of the wheat crop from sensors on the Landsat series of satellites. The Green Normalised Difference Vegetation Index (GNDVI) was calculated for each pixel location and year. The GNDVI is closely related to the photo synthetically absorbed radiation and has shown a linear correlation with the Leaf Area Index (LAI) and biomass in other studies. The spatial and temporal variability in the GNDVI were quantified across the village.
Soil brightness imagery, which is an indicator of soil texture, organic matter and soil moisture, was also obtained from a commercial source for testing as an additional variable for defining management zones. Sampling data for soil inorganic nitrogen (N) and organic carbon (OC) were also available for validation purposes.
A clustering algorithm, known as partitioning around medoids, was used to group fields into three or four clusters, or management zones. The clustering algorithm aims to minimise the amount of variation (in the clustering variables) within clusters and maximise the amount of variation between clusters. The accuracy of this approach for delineating management zones was evaluated by calculating the relative variance of the measured soil N and OC. Soil brightness alone was a fairly poor predictor of measured N and OC in this study, explaining only up to 9% of the variability. The GNDVI variability metrics were a reasonable predictor (up to 39%) of variability in measured N and OC. The three-zone solution in the combined model (soil brightness + GNDVI) was the best predictor of N and OC variability at up to 45%.
The results of this study can be considered as a preliminary method based on the integration of different remotely sensed data to delineate management zones at the village scale. More studies are needed to further refine them for guiding site-specific management in small scale farming systems. In addition, incorporating measurements of field level yields would aid in validating this approach in the future.
This project was funded by the Agri Tech in China Newton Network+ (ATCNN) and Norwegian Ministry of Foreign Affairs (SINOGRAIN II, CHN-17/0019).