Granted Patent B2

US 12,393,995 · App. 17/805,926 · Granted Aug 19, 2025

Inventors: Jagabondhu Hazra (HBR Layout, IN); Manikandan Padmanaban (Chennai, IN); Isaac Waweru Wambugu (Nairobi, KE); Lloyd A Treinish (Cortlandt Manor, NY); Ivan Kayongo (Nairobi, KE); Kumar Saurav (Bhojpur, IN); Ranjini Bangalore Guruprasad (Bangalore, IN)

Assignee: International Business Machines Corporation

G06Q50/02G06Q10/063

Abstract

Claims (50)

1. A method for environmental impact estimations, the method comprising:

obtaining data pertaining to an agricultural area identified by a user in an environmental impact user interface, wherein the agricultural area is comprised of a plurality pixels corresponding to a ground area, wherein the data includes coarse resolution greenhouse gas emissions data and fine resolution explanatory variable data;

deriving one or more features from the data pertaining to the agricultural area;

displaying one or more stubble burning areas to the user in the environmental impact user interface, wherein the one or more stubble burning areas are identified within the agricultural area using a trained machine learning model, wherein the trained machine learning model utilizes the one or more derived features as input for a crop agnostic binary classification model;

determining greenhouse gas emission estimates for the one or more stubble burning areas using a stacked super resolution convolutional neural network by performing a layer by layer enhancement based on the coarse resolution greenhouse gas emissions data and the fine resolution explanatory variable data, wherein the convolutional neural network learns a functional mapping between the fine resolution and coarse resolution data; and

determining an environmental impact for each of the one or more stubble burning areas, wherein the environmental impact includes the greenhouse gas emission estimates for each pixel of the one or more stubble burnings areas within the plurality of pixels corresponding to the ground area based on the functional mapping of the one or more stubble burning areas.

2. The method of claim 1 , wherein the one or more derived features from the data pertaining to the agricultural area include a SAR power spectrum, a SAR decomposition, and a signal superposition.

3. The method of claim 2 , wherein the one or more stubble burning areas are comprised of one or more burnt pixels, the one or more burnt pixels being classified using the crop agnostic binary classification model, wherein the trained machine learning model is trained utilizing data from a verified fire information database, wherein at least the SAR power spectrum and the signal superposition are normalized using cohort analytics to build the crop agnostic binary classification model.

4. The method of claim 1 , wherein determining the environmental impact for each of the one or more stubble burning areas further comprises:

utilizing one or more explanatory variables and greenhouse gas emissions data as input for a deep learning model, wherein the deep learning model is the convolutional neural network; and

upscaling the one or more explanatory variables using one or more standard interpolation techniques.

5. The method of claim 4 , further comprising:

determining the greenhouse gas emissions for each of one or more burnt pixels classified by the crop agnostic binary classification model using the deep learning model; and

displaying the greenhouse gas emissions for each of the one or more burnt pixels to the user within the environmental impact user interface, wherein the greenhouse gas emissions data is presented separately for each of a plurality of greenhouse gases.

6. The method of claim 1 , wherein determining the environmental impact for each of the one or more stubble burning areas further comprises:

allocating coarse greenhouse gas emissions data to a finer spatial resolution using a statistical model, wherein the statistical model utilizes a FRK technique in enhancing the coarse resolution greenhouse gas emissions data based on one or more fine resolution covariates.

7. The method of claim 1 , wherein the data pertaining to the agricultural area includes contextual information, wherein the contextual information is utilized in deriving cohort analytics to normalize one or more time series features of the one or more derived features from the data pertaining to the agricultural area.

8. The method of claim 7 , further comprising:

identifying one or more cohort groups within the agricultural area based on a similarity score generated for each of a plurality of agricultural fields within the agricultural area, wherein the similarity score is generated by a multi-class classification model utilizing the cohort analytics as input, wherein the contextual information is weighted according to preferences specified by the user in the environmental impact user interface.

9. The method of claim 8 , further comprising:

utilizing the one or more cohort groups identified to refine the input for the crop agnostic binary classification model.

10. The method of claim 1 , further comprising:

displaying the one or more stubble burning areas identified to the user in the environmental impact user interface using one or more visual representations and a classification for each of a plurality of pixels comprising the agricultural area.

11. The method of claim 1 , wherein the greenhouse gas emissions estimates are an output of a deep learning model utilizing a stacked super resolution convolutional neural network technique to enhance a spatial resolution of greenhouse gas emissions data by utilizing both coarse resolution greenhouse gas emissions data and fine resolution explanatory variable data as input.

12. A computer system for environmental impact estimations, comprising:

one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is capable of performing a method comprising:

obtaining data pertaining to an agricultural area identified by a user in an environmental impact user interface, wherein the agricultural area is comprised of a plurality pixels corresponding to a ground area, wherein the data includes coarse resolution greenhouse gas emissions data and fine resolution explanatory variable data;

deriving one or more features from the data pertaining to the agricultural area;

displaying one or more stubble burning areas to the user in the environmental impact user interface, wherein the one or more stubble burning areas are identified within the agricultural area using a trained machine learning model, wherein the trained machine learning model utilizes the one or more derived features as input for a crop agnostic binary classification model;

determining greenhouse gas emission estimates for the one or more stubble burning areas using a stacked super resolution convolutional neural network by performing a layer by layer enhancement based on the coarse resolution greenhouse gas emissions data and the fine resolution explanatory variable data, wherein the convolutional neural network learns a functional mapping between the fine resolution and coarse resolution data; and

determining an environmental impact for each of the one or more stubble burning areas, wherein the environmental impact includes the greenhouse gas emission estimated for each pixel of the one or more stubble burning areas within the plurality of pixels corresponding to the ground area based on the functional mapping of the one or more stubble burning areas.

13. The computer system of claim 12 , wherein the one or more derived features from the data pertaining to the agricultural area include a SAR power spectrum, a SAR decomposition, and a signal superposition.

14. The computer system of claim 13 , wherein the one or more stubble burning areas are comprised of one or more burnt pixels, the one or more burnt pixels being classified using the crop agnostic binary classification model, wherein the trained machine learning model is trained utilizing data from a verified fire information database, wherein at least the SAR power spectrum and the signal superposition are normalized using cohort analytics to build the crop agnostic binary classification model.

15. The computer system of claim 12 , wherein determining the environmental impact for each of the one or more stubble burning areas further comprises:

utilizing one or more explanatory variables and greenhouse gas emissions data as input for a deep learning model, wherein the deep learning model is the convolutional neural network; and

upscaling the one or more explanatory variables using one or more standard interpolation techniques.

16. The computer system of claim 12 , wherein determining the environmental impact for each of the one or more stubble burning areas further comprises:

allocating coarse greenhouse gas emissions data to a finer spatial resolution using a statistical model, wherein the statistical model utilizes a FRK technique in enhancing the coarse resolution greenhouse gas emissions data based on one or more fine resolution covariates.

17. A computer program product for environmental impact estimations, comprising:

one or more non-transitory computer-readable storage media and program instructions stored on at least one of the one or more tangible storage media, the program instructions executable by a processor to cause the processor to perform a method comprising:

obtaining data pertaining to an agricultural area identified by a user in an environmental impact user interface, wherein the agricultural area is comprised of a plurality pixels corresponding to a ground area, wherein the data includes coarse resolution greenhouse gas emissions data and fine resolution explanatory variable data;

deriving one or more features from the data pertaining to the agricultural area;

displaying one or more stubble burning areas to the user in the environmental impact user interface, wherein the one or more stubble burning areas are identified within the agricultural area using a trained machine learning model, wherein the trained machine learning model utilizes the one or more derived features as input for a crop agnostic binary classification model;

determining greenhouse gas emission estimates for the one or more stubble burning areas using a stacked super resolution convolutional neural network by performing a layer by layer enhancement based on the coarse resolution greenhouse gas emissions data and the fine resolution explanatory variable data, wherein the convolutional neural network learns a functional mapping between the fine resolution and coarse resolution data; and

determining an environmental impact for each of the one or more stubble burning areas, wherein the environmental impact includes the greenhouse gas emission estimated for each pixel of the one or more stubble burning areas within the plurality of pixels corresponding to the ground area based on the functional mapping of the one or more stubble burning areas.

18. The computer program product of claim 17 , wherein the one or more derived features from the data pertaining to the agricultural area include a SAR power spectrum, a SAR decomposition, and a signal superposition.

19. The computer program product of claim 18 , wherein the one or more stubble burning areas are comprised of one or more burnt pixels, the one or more burnt pixels being classified using the crop agnostic binary classification model, wherein the trained machine learning model is trained utilizing data from a verified fire information database, wherein at least the SAR power spectrum and the signal superposition are normalized using cohort analytics to build the crop agnostic binary classification model.

20. The computer program product of claim 17 , wherein determining the environmental impact for each of the one or more stubble burning areas further comprises:

utilizing one or more explanatory variables and greenhouse gas emissions data as input for a deep learning model, wherein the deep learning model is the convolutional neural network; and

upscaling the one or more explanatory variables using one or more standard interpolation techniques.

Assignments (1)

Continuity (1)

References Cited (107)