Delineating management zones based on historical yield maps

1. A method comprising:

using instructions programmed in a computer system comprising one or more processors and computer memory, receiving, from physical sensors via a computer network, digital yield data representing yields of crops that have been harvested from an agricultural field;

using the instructions programmed in the computer system, transforming the digital yield data, by applying an empirical cumulative density function to the digital yield data, to generate transformed digital yield data;

based on the transformed digital yield data, using the instructions programmed in the computer system, smoothing and imputing, missing yield values to the transformed digital yield data to result in generating and storing smooth transformed digital yield data;

wherein imputing missing yield values to the transformed digital yield data comprises:

plotting the transformed yield data onto a virtual geometric digital grid constructed for the agricultural field and enclosed by boundaries of the agricultural field;

determining a particular grid, of the virtual geometric digital grid, that has no corresponding yield data record in the transformed digital yield data;

determining whether one or more yield data observations for a location corresponding to the particular grid are available in one or more observed yield value sets collected in one or more previous years;

in response to determining that one or more yield data observations for a location corresponding to the particular grid are available in One or more observed yield values sets collected in one or more previous years;

determining, using spatiotemporal modeling, a missing yield data record by cross imputing the one or more yield data observations for the location corresponding to the particular grid from the one or more observed yield value sets collected in one or more previous years; and

imputing the missing yield data record into the transformed digital yield data as the corresponding yield data record;

in response to determining that no yield data observations for the location corresponding to the particular grid is available in any observed yield value sets collected in previous years:

determining, for a plurality of neighboring grids that are adjacent to the particular grid, a plurality of neighboring yield data records from the transformed digital yield data;

determining, a plurality of weights for each of the neighboring yield data records;

computing a weighted sum based on the plurality of neighboring yield data records and the plurality of weights; and

imputing the weighted sum into the transformed digital yield data as the corresponding yield data record;

using the instructions programmed in the computer system, determining a first count value of a plurality of management classes, wherein the plurality of management classes includes areas in the agricultural field that have homogeneous yield limiting factors, but are not restricted to be spatially contiguous;

using the instructions programmed in the computer system, generating a plurality of first management zones for the agricultural field by clustering the smooth transformed digital yield data and using the first count value;

using the instructions programmed in the computer system, generating a set of first merged management zones by merging one or more small management zones, of the plurality of first management zones, with their respective similar neighboring large zones;

using the instructions programmed in the computer system, storing, in a computer-based database, the set of first merged management zones and the first count value, in a set of management zone metrics; and

using the instructions programmed in the computer system, automatically controlling, using the set of first merged management zones, a computer control system of one or more of seeding, irrigation, nitrogen application, and harvesting apparatus.

2. The method of claim 1 , further comprising:

determining a second count value for the plurality of management classes;

generating a plurality of second management zones by clustering the smooth transformed digital yield data and using the second count value;

generating a set of second merged management zones by merging one or more small management zones, of the plurality of second management zones, with their respective similar neighboring large zones; and

storing the set of second merged management zones and the second count value in the set of management zone metrics.

3. The method of claim 2 , further comprising: generating a management zone recommendation based on the set of management zone metrics by evaluating a delineation quality of management zones stored in the set of management zone metrics.

4. The method of claim 3 , wherein the digital yield data comprise data representing yield information collected for a multi-year time period, and wherein the method further comprises determining a recommended class count for the plurality of management classes based on the set of management zone metrics.

5. The method of claim 3 , wherein the digital yield data comprise data representing yield information collected for one year, and wherein the method further comprises determining a recommended class count for the plurality of management classes based on the set of Management zone metrics.

6. The method of claim 1 , further comprising obtaining the digital yield data from a plurality of historical yield maps.

7. The method of claim 1 , further comprising preprocessing the digital yield data by performing one or more of: removing yield maps that correspond to multiple crops planted in the same season in the agricultural field, removing yield maps that include yield records outside the boundaries of the agricultural field, marking yield records of absolute zeros as missing values, removing yield records for fields that have less than two years of yield maps, or removing yield maps that have more than a certain count of missing values.

8. The method of claim 1 , further comprising using the empirical cumulative density function to transform the digital yield data into the transformed digital yield data including transformed digital yield data records, each of which falls in a particular range.

9. The method of claim 1 , further comprising generating the smooth transformed digital yield data by performing one or more of: removing outliers data from the transformed digital yield data, determining one or more missing values in the transformed digital yield data and including the one or more missing values in the smooth transformed digital yield data, or performing a spatial smoothing on the transformed digital yield data.

10. The method of claim 1 , wherein the set of first merged management zones includes contiguous subregions within the agricultural field that have homogeneous yield limiting factors and are uniformly managed; wherein the first count value for the plurality of management classes indicates a count of management classes in the plurality of management classes.

11. The method of claim 1 , further comprising applying any one of: a K-means approach, a fuzzy C-means approach, or a region merging approach.

12. The method of claim 1 , further comprising generating the set of first merged management zones by executing computer instructions to perform one or more of: applying hierarchical agglomeration; evaluating a delineation quality of management zones by applying one or more clustering validation measures to the set of management zone metrics.

13. A data processing system comprising:

a memory; and

one or more processors coupled to the memory and programmed to:

receive digital yield data representing yields of crops that have been harvested from an agricultural field;

transform the digital yield data, by applying an empirical cumulative density function to the digital yield data, to generate transformed digital yield data;

based on the transformed digital yield data, smooth the transformed digital yield data to result in generating and storing smooth transformed digital yield data;

wherein smoothing the transformed digital yield data comprises:

plotting the transformed digital yield data onto a virtual geometric digital grid constructed for the agricultural field and enclosed by boundaries of the agricultural field;

determining a particular grid, of the virtual geometric digital grid, that has no corresponding yield data record in the transformed digital yield data;

determining whether one or more yield data observations for a location corresponding to the particular grid are available in one or more observed yield value sets collected in one Or more previous years;

in response to determining that one or more yield data observations for a location corresponding to the particular grid are available in one or more observed yield value sets collected in one or more previous years:

determining, using spatiotemporal modeling, a missing yield data record by cross imputing, the one or more yield data observations for the location corresponding to the particular grid from the one or more observed yield value sets collected in one or more previous years; and

imputing the missing yield data record into the transformed digital yield data as the corresponding yield data record;

in response to determining that no yield data observations for the location corresponding to the particular grid is available in any observed yield value sets collected in previous years:

determining, for a plurality of neighboring grids that are adjacent to the particular grid, a plurality of neighboring yield data records from the transformed digital yield data;

determining, a plurality of weights for each of the neighboring yield data records;

computing a weighted sum based on the plurality of neighboring yield data records and the plurality of weights;

and imputing the weighted sum into the transformed digital yield data as the corresponding yield data record;

determine a first count value of a plurality of management classes, wherein the plurality of management classes includes areas in the agricultural field that have homogeneous yield limiting factors, but are not restricted to be spatially contiguous;

generate a plurality of first management zones for the agricultural field by clustering the smooth transformed digital yield data and using the first count value;

generate a set of first merged management zones by merging one or more small management zones, of the plurality of first management zones, with their respective similar neighboring large zones;

store the set of first merged management zones and the first count value in a set of management zone metrics; and

automatically control, using the set of first merged management zones, a computer control system of one or more of seeding, irrigation, nitrogen application, and harvesting apparatus.

14. The data processing system of claim 13 , wherein the one or more processors are programmed to:

determine a second count value for the plurality of management classes;

generate a plurality of second management zones by clustering the smooth transformed digital yield data and using the second count value;

generate a set of second merged management zones by merging one or more small management zones, of the plurality of second management zones, with their respective similar neighboring large zones; and

store the set of second merged management zones and the second count value in the set of management zone metrics.

15. The data processing system of claim 14 , wherein the one or more processors are programmed to: generate a management zone recommendation based on the set of management zone metrics by evaluating a delineation quality of management zones stored in the set of management zone metrics.

16. The data processing system of claim 13 , wherein the digital yield data representing crop yields harvested from the agricultural field are is-obtained from a plurality of historical yield maps.

17. The data processing system of claim 13 , wherein the digital yield data representing crop yields harvested from the agricultural field are preprocessed by performing one or more of: removing yield maps that correspond to multiple crops planted in the same season in the agricultural field, removing yield maps that include yield records outside the boundaries of the agricultural field, marking yield records of absolute zeros as missing values, removing yield records for fields that have less than two years of yield maps, or removing yield maps that have more than a certain count of missing values.

18. The data processing system of claim 13 , wherein the empirical cumulative density function transforms the digital yield data to the transformed digital yield data including transformed digital yield data records, each of which falls in a particular range.

19. The data processing system of claim 13 , wherein the smooth transformed digital yield data are generated by performing one or more of: removing outliers data from the transformed digital yield data, determining one or more missing values in the transformed digital yield data and including the one or more missing values in the smooth transformed digital yield data, or performing, a spatial smoothing on the transformed digital yield data.