Remote sensing algorithms for mapping regenerative agriculture

1. A method of predicting a cover crop practice, the method comprising:

receiving a time series of satellite imagery, the time series of satellite imagery covering at least a geographic region, during a predetermined time period, wherein the predetermined time period begins with an end of a first crop season and ends with an end of a following crop season;

generating a first set of field level zonal summary time series from the time series of satellite imagery for at least one field within the geographic region, wherein the field level zonal summary time series comprises a time series of Normalized Difference Vegetation Index (NDVI) values;

for each field in the geographic region,

determining a crop type of a field;

counting a number of peaks in the time series of NDVI values during the predetermined time period;

determining the number of peaks in the time series of NDVI values during winter months of the predetermined time period;

estimating a length of spring by calculating a number of days between a day of year when NDVI reaches 15% of a normalized amplitude of the last NDVI peak during the predetermined time period and a day of year when NDVI reaches 50% of the normalized amplitude of the last NDVI peak during the predetermined time period;

calculating a winter greenness value;

for fields of each crop type within a same county, calculating a time series of median NDVI values across all fields of the crop type; and

for each field in the county, applying a decision tree classifier to a field, wherein a field is predicted not to have a cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is less than a threshold greenness, or

ii. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is less than a threshold number of days, or

iii. fewer than two peaks in NDVI are detected during the predetermined time period and the determined crop type is winter wheat; and

a field is predicted to have a wintergreen cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is greater than a threshold number of days and the determined crop type is not winter wheat, or

ii. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness, or

iii. a Euclidean distance between the time series of NDVI values for the field and the time series of median NDVI values across all fields of the crop type and within the same county is greater than or equal to a threshold distance; and

a field is predicted to have a winterkill cover crop if

i. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness is less than a threshold greenness, or

ii. the Euclidean distance between the time series of NDVI values for the field and the median NDVI time series values across all fields of the crop type and within the same county is greater than or equal to a threshold distance.

2. The method of claim 1 , wherein the geographic region is one or more states, counties, farms, or fields.

3. The method of claim 1 , wherein the field level zonal summary time series are interpolated at daily time resolution using a Savitzky-Golay filter.

4. The method of claim 1 , wherein the crop type is a summer crop and the winter months comprise at least one month starting with a month following the end of the crop season for the crop type and ending with a month after a beginning of the following crop season for the crop type.

5. The method of claim 1 , wherein the winter greenness value is 10th percentile of NDVI values over the winter months.

6. The method of claim 1 , wherein the winter greenness value is 90th percentile of NDVI values over the winter months.

7. The method of claim 1 , wherein the threshold number of days is 30 days.

8. The method of claim 1 , wherein the threshold distance is 2.

9. A non-transitory computer-readable medium comprising memory with instructions encoded thereon for predicting a cover crop practice, the instructions, when executed by one or more processors, causing the one or more processors to perform operations comprising:

receiving a time series of satellite imagery, the time series of satellite imagery covering at least a geographic region, during a predetermined time period, wherein the predetermined time period begins with an end of a first crop season and ends with an end of a following crop season;

generating a first set of field level zonal summary time series from the time series of satellite imagery for at least one field within the geographic region, wherein the field level zonal summary time series comprises a time series of Normalized Difference Vegetation Index (NDVI) values;

for each field in the geographic region,

determining a crop type of a field;

counting a number of peaks in the time series of NDVI values during the predetermined time period;

determining the number of peaks in the time series of NDVI values during winter months of the predetermined time period;

estimating a length of spring by calculating a number of days between a day of year when NDVI reaches 15% of a normalized amplitude of the last NDVI peak during the predetermined time period and a day of year when NDVI reaches 50% of the normalized amplitude of the last NDVI peak during the predetermined time period;

calculating a winter greenness value;

for fields of each crop type within a same county, calculating a time series of median NDVI values across all fields of the crop type; and

for each field in the county, applying a decision tree classifier to a field, wherein a field is predicted not to have a cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is less than a threshold greenness, or

ii. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is less than a threshold number of days, or

iii. fewer than two peaks in NDVI are detected during the predetermined time period and the determined crop type is winter wheat; and

a field is predicted to have a wintergreen cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is greater than a threshold number of days and the determined crop type is not winter wheat, or

ii. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness, or

iii. a Euclidean distance between the time series of NDVI values for the field and the time series of median NDVI values across all fields of the crop type and within the same county is greater than or equal to a threshold distance; and

a field is predicted to have a winterkill cover crop if

i. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness is less than a threshold greenness, or

ii. the Euclidean distance between the time series of NDVI values for the field and the median NDVI time series values across all fields of the crop type and within the same county is greater than or equal to a threshold distance.

10. The non-transitory computer-readable medium of claim 9 , wherein the geographic region is one or more states, counties, farms, or fields.

11. The non-transitory computer-readable medium of claim 9 , wherein the field level zonal summary time series are interpolated at daily time resolution using a Savitzky-Golay filter.

12. The non-transitory computer-readable medium of claim 9 , wherein the crop type is a summer crop and the winter months comprise at least one month starting with a month following the end of the crop season for the crop type and ending with a month after a beginning of the following crop season for the crop type.

13. The non-transitory computer-readable medium of claim 9 , wherein the winter greenness value is 10th percentile of NDVI values over the winter months.

14. The non-transitory computer-readable medium of claim 9 , wherein the winter greenness value is 90th percentile of NDVI values over the winter months.

15. The non-transitory computer-readable medium of claim 9 , wherein the threshold number of days is 30 days.

16. The non-transitory computer-readable medium of claim 9 , wherein the threshold distance is 2.

17. A system comprising:

memory with instructions encoded thereon for predicting a cover crop practice; and

one or more processors that, when executing the instructions, are caused to perform operations comprising:

receiving a time series of satellite imagery, the time series of satellite imagery covering at least a geographic region, during a predetermined time period, wherein the predetermined time period begins with an end of a first crop season and ends with an end of a following crop season;

generating a first set of field level zonal summary time series from the time series of satellite imagery for at least one field within the geographic region, wherein the field level zonal summary time series comprises a time series of Normalized Difference Vegetation Index (NDVI) values;

for each field in the geographic region,

determining a crop type of a field;

counting a number of peaks in the time series of NDVI values during the predetermined time period;

determining the number of peaks in the time series of NDVI values during winter months of the predetermined time period;

estimating a length of spring by calculating a number of days between a day of year when NDVI reaches 15% of a normalized amplitude of the last NDVI peak during the predetermined time period and a day of year when NDVI reaches 50% of the normalized amplitude of the last NDVI peak during the predetermined time period;

calculating a winter greenness value;

for fields of each crop type within a same county, calculating a time series of median NDVI values across all fields of the crop type; and

for each field in the county, applying a decision tree classifier to a field, wherein a field is predicted not to have a cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is less than a threshold greenness, or

ii. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is less than a threshold number of days, or

iii. fewer than two peaks in NDVI are detected during the predetermined time period and the determined crop type is winter wheat; and

a field is predicted to have a wintergreen cover crop if

i. fewer than two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness and the estimated length of spring is greater than a threshold number of days and the determined crop type is not winter wheat, or

ii. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness value is greater than a threshold greenness, or

iii. a Euclidean distance between the time series of NDVI values for the field and the time series of median NDVI values across all fields of the crop type and within the same county is greater than or equal to a threshold distance; and

a field is predicted to have a winterkill cover crop if

i. at least two peaks in NDVI are detected during the predetermined time period and the winter greenness is less than a threshold greenness, or

ii. the Euclidean distance between the time series of NDVI values for the field and the median NDVI time series values across all fields of the crop type and within the same county is greater than or equal to a threshold distance.

18. The system of claim 17 , wherein the geographic region is one or more states, counties, farms, or fields.

19. The system of claim 17 , wherein the field level zonal summary time series are interpolated at daily time resolution using a Savitzky-Golay filter.

20. The system of claim 17 , wherein the crop type is a summer crop and the winter months comprise at least one month starting with a month following the end of the crop season for the crop type and ending with a month after a beginning of the following crop season for the crop type.