The goal of a Yield estimation map is to predict the levels of yields that can be achieved at harvest time.
How we get to the estimations
To get to the estimations, we will:
- Based on field observations, get yield estimations for particular points in the field.
- Establish a correlation between those estimations and the NDVI values from or near the date of the field observations
- Based on the NDVI values for the rest of the field, extrapolate the yield estimations.
Visiting the field and getting the yield estimations
The first step in the Yields estimation process is the field visit. In the visit, for each field we want to estimate yields, a number of locations will be visited and, based on a number of criteria and forecast variables which will depend on the crop, geographic area, etc, an estimation of the expected yields will be given.
The number of points to be visited and estimated for a field may vary depending on the area and the historical variability, but it's advisable to have no less than 20 points, scattered all across the area of interest for each field.
The field visit should be between 45 and 30 days before the harvest, this being the usual timeframe for the crop flowering and the dates of maximum NDVI value.
Establishing the correlation
Having the estimations, the next step will be to establish a correlation between the estimations for a particular field and the NDVI values for the same field, for the date of the observations. To do this, we obtain satellite imagery from that date (approximately), process the NDVI maps and match the yield estimations with the corresponding NDVI values, based on their geographical location.
Finally, a curve is adjusted to those NDVI values and yield estimations. Several different curve types are tested; based on statistical criteria, the curve which provides a better model for the data is selected.
Producing the map
The final step is the production of the yield estimation map. The curve obtained in the previous step is applied to the rest of the NDVI values in the map, resulting in yield estimation values for the rest of the field, and the process is then carried out in the rest of the field