Skip to main content

High Resolution Imagery for Crop Yields Approximation

Crop yield forecasting is an essentially the most important aspect in the agriculture which can benefit in understanding the yield gaps, aiding in suggesting better practices. Current availiablities of different geoinfomational products have extended the ways in the crop monitoring. While the crop monitoring process is the continuous process, both the high temporal resolution as well as finer spatial resolution might not be possible at many cases. For instance, Planet/Dove is equipped with daily revist cycle and WorldView-3 provides finer spatial resolution. Incorporation of both has the possibilites for understanding the crops at the field levels. Satellite data, if properly available in growing periods can be helpful in precision farming and can be implemented in guiding the samping based estimation of yields.

Satellite imageries (VHR) themselves are recieved untimely in unperiodic manner. For this, sensors with high spatial and temporal can expand the applications in the crop monitoring process. A research attempts to explain the capabilites of spatial resolution to crop yields, extents of the space borne sensors to explain the crop yield variabilites, and the particular timing for acquiring the image which helps in further explaining within-field variability.

The study collected field crop yield data from (30 fields) combine harvesting machines, each from the differnce from 3 m distance whose yield maps were generated at 10, 20 and 30m resolution. The studied imagery included WorldView-3 (1.25 m) - 8 spectral bands, Planet/Dove-Classic (3.25 m), and harmoized Landsat 8/ Sentinel-2 HLS (30 m). Field values and satellite features were correlated, rergression model was developed between field data and satellite features.

In 30 m resolution, only 59% of field variability were found to be explained due to coarseness and mixtured effects. The author notes the difficulty of making a single crop cut within 30 m pixel to be compared with the satellite data.

WV-3 could explain 48% (corn) and 36% (soybean) of crop variability in the yield variability which were months after the planting. Also, only the red, green and NIR bands were suggestive in explaining the yield variance. The authrs while analyzing the change in R2 in the fields with Planet and HLS data, with varied results unable to explain field yield variability. Different models were incorporated within for the results; positive correlations were seen between HLS and Planet based models. The noted point is that the increase in yield values decreased the models explainability towards yield variability, probably saturating over high yields.

The authors suggest in focusing in incorporation of water stress indicators and active remote sensing sensors.

The article is more concisley summarized here in

Link for the research paper: 


Recent Posts Widget