Predicting soil organic carbon content
Implementations are described herein for predicting soil organic carbon (“SOC”) content for agricultural fields detected in digital imagery. In various implementations, one or more digital images depicting portion(s) of one or more agricultural fields may be processed. The one or more digital images may have been acquired by a vision sensor carried through the field(s) by a ground-based vehicle. Based on the processing, one or more agricultural inferences indicating agricultural practices or conditions predicted to affect SOC content may be determined. Based on the agricultural inferences, one or more predicted SOC measurements for the field(s) may be determined.
1. A method implemented using one or more processors, comprising:
obtaining a plurality of digital images depicting a field over multiple growing seasons;
applying the plurality of digital images as input to one or more machine learning models to generate first output indicative of two or more inferred agricultural management practices implemented in the field across the multiple growing seasons, wherein the two or more inferred agricultural management practices include:
an inferred tillage practice employed in the field during the multiple growing seasons;
a crop rotation employed in the field during the multiple growing seasons; or
a cover crop grown in the field during the multiple growing seasons; and
applying data indicative of the two or more inferred agricultural management practices as input to one or more additional machine learning models to generate second output, wherein the second output represents a predicted measure of soil organic carbon (SOC) of the field.
2. The method of claim 1 , further comprising obtaining a slope map of the field based on data generated by one or more sensors carried throughout the field by one or more ground-based farm vehicles, wherein the predicted measure of SOC is further predicted based on the slope map.
3. The method of claim 1 , further comprising processing the plurality of images to predict a crop yield of the field during a given growing season of the multiple growing seasons, wherein the measure of SOC is further predicted based on the predicted crop yield.
4. The method of claim 1 , wherein additional data is applied as input to the one or more additional machine learning models to generate the second output.
5. The method of claim 4 , wherein the additional data comprises a ground truth crop yield during the multiple growing seasons.
6. The method of claim 1 , further comprising operating a ground-based robot to carry one or more vision sensors through the field to capture the plurality of digital images.
7. The method of claim 1 , wherein at least some of the plurality of images are obtained from an edge-computing node that is mounted on a boom of a farm vehicle.
8. The method of claim 1 , wherein the plurality of digital images are processed using one or more convolutional machine learning models.
9. The method of claim 1 , further comprising:
receiving indications of one or more tillage practices and rotations of crops likely to be implemented in the field during a future time period; and
predicting a future measure of SOC associated with the field.
10. The method of claim 1 , wherein the plurality of digital images includes high-elevation digital images of the field captured by one or more satellites or unmanned aerial vehicles (“UAVs”) during the multiple growing seasons.
11. A method implemented using one or more processors, comprising:
obtaining a plurality of digital images depicting a field over multiple growing seasons;
obtaining ground truth data indicative of a plurality of ground truth measures of soil organic content (SOC) during the multiple growing seasons;
applying the plurality of digital images as input to one or more machine learning models to generate first output indicative of two or more inferred agricultural management practices implemented in the field across the multiple growing seasons, wherein the two or more inferred agricultural management practices include:
an inferred tillage practice employed in the field during the multiple growing seasons;
a crop rotation employed in the field during the multiple growing seasons; or
a cover crop grown in the field during the multiple growing seasons;
applying data indicative of the two or more inferred agricultural management practices as input to one or more additional machine learning models to generate second output, wherein the second output represents a plurality of predicted measures of SOC of the field;
comparing the plurality of predicted measures of SOC to the corresponding ground truth measures of SOC; and
training one or more of the additional machine learning models based on the comparing.
12. The method of claim 11 , wherein the one or more additional machine learning models comprise a recurrent convolutional neural network.
13. A system comprising one or more processors and memory storing instructions that, in response to execution by the one or more processors, cause the one or more processors to:
obtain a plurality of digital images depicting a field over multiple growing seasons;
apply the plurality of digital images as input to one or more machine learning models to generate first output indicative of two or more inferred agricultural management practices implemented in the field across the multiple growing seasons, wherein the two or more inferred agricultural management practices include:
an inferred tillage practice employed in the field during the multiple growing seasons;
a crop rotation employed in the field during the multiple growing seasons; or
a cover crop grown in the field during the multiple growing seasons; and
apply data indicative of the two or more inferred agricultural management practices as input to one or more additional machine learning models to generate second output, wherein the second output represents a predicted measure of soil organic carbon (SOC) of the field.
14. The system of claim 13 , further comprising instructions to obtain a slope map of the field based on data generated by one or more sensors carried throughout the field by one or more ground-based farm vehicles, wherein the predicted measure of SOC is further predicted based on the slope map.
15. The system of claim 13 , further comprising instructions to process the plurality of images to predict a crop yield of the field during a given growing season of the multiple growing seasons, wherein the measure of SOC is further predicted based on the predicted crop yield.
16. The system of claim 13 , wherein additional data is applied as input to the one or more additional machine learning models to generate the second output.
17. The system of claim 16 , wherein the additional data comprises a ground truth crop yield during the multiple growing seasons.
18. The system of claim 13 , further comprising instructions to deploy a ground-based robot to carry one or more vision sensors through the field to capture the plurality of digital images.
19. The system of claim 13 , wherein at least some of the plurality of images are obtained from an edge-computing node that is mounted on a boom of a farm vehicle.
20. The system of claim 13 , wherein the plurality of digital images are processed using one or more convolutional machine learning models.