Vertical SOC Estimation Using Hybrid ML and Process Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for estimating Soil Organic Carbon (SOC) are limited by their inability to provide precise, scalable, and spatially variable assessments, particularly in heterogeneous landscapes, and existing machine learning models fail to accurately predict SOC beyond the surface layer due to limited satellite penetration and seasonal variability.
Innovation Solution
A domain-aware machine learning model integrated with a process-based model, utilizing spectral and temporal embeddings, attention mechanisms, and correction factors to estimate vertical SOC distribution, incorporating satellite data, management proxies, and soil spectral libraries, along with process-based models to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning models are used for SOC estimation, then surface SOC can be predicted, but the model fails to predict vertical SOC distribution beyond surface layer due to limited satellite penetration
Solution Approach 1:
The patent introduces process-based models as intermediary components that bridge the gap between satellite-observed surface SOC and deeper vertical SOC distribution. These models use environmental covariates and biogeochemical processes to simulate carbon dynamics at different depths, enabling indirect estimation of vertical SOC where direct satellite observation is limited
Solution Approach 2:
The patent transitions from two-dimensional surface SOC estimation to three-dimensional vertical SOC distribution by incorporating depth as an additional dimension. This is achieved by integrating process-based models that simulate carbon dynamics across multiple soil layers, transforming the prediction problem from surface-only to volumetric SOC distribution
2Productivity
If conventional ML models are used, then estimation can be performed, but the models fail to account for seasonal and regional variability
Solution Approach 1:
The patent transforms static SOC estimation into a dynamic system that adapts to seasonal and regional variations. This is achieved by incorporating time-varying environmental covariates, using recurrent neural networks to capture temporal dynamics, and allowing model parameters to adjust based on regional characteristics and seasonal conditions
Solution Approach 2:
The patent employs parameter changes to enhance model adaptability by incorporating environmental covariates such as temperature, precipitation, and vegetation indices that vary seasonally and regionally. These parameters are integrated into the machine learning models to capture spatiotemporal variability in SOC dynamics across different landscapes and time periods
3Measurement precision
If laboratory-based methods are used for SOC estimation, then precise measurements can be obtained, but the methods are time-consuming, cumbersome and expensive, making them less suitable for large-scale spatial assessments
Solution Approach 1:
The patent replaces mechanical laboratory-based measurement systems with remote sensing and computational modeling systems. Satellite-based spectral data and process-based models substitute for physical soil sampling and laboratory analysis, enabling large-scale spatial assessment without the time and cost constraints of traditional methods while maintaining acceptable accuracy through model calibration
4Ease of operation
If geostatistical methods are used, then spatial interpolation can be performed, but the methods assume spatial variability remains constant (stationarity) which may not hold in highly heterogeneous landscapes
Solution Approach 1:
The patent applies local quality by allowing model parameters and relationships to vary spatially across different landscapes rather than assuming uniform stationarity. This is achieved by incorporating region-specific environmental covariates, using geographically weighted regression techniques, and allowing the model to adapt to local soil, climate, and management conditions in heterogeneous landscapes
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This disclosure relates generally to a method and system for dynamic estimation of vertical soil organic carbon (SOC) at a region of interest (ROI). State-of-the-art methods are greatly dependent on satellite data. The satellite measurements are reliable for estimating the surface SOC, but it lacks accuracy when it comes to making estimations at sub-surface layers. The present disclosure addresses these problems through a method of dynamic estimation of vertical SOC by combining a domain-aware machine learning (ML) model and a process-based model. The domain-aware ML model incorporates spectral data, management practice data, soil spectral library (SSL) data, and a correction from the process-based model a corrected surface SOC. The process-based model further estimate depth-specific SOC fractions, along with fractions derived from global soil datasets and the SSL. By combining the domain-aware ML model and the depth-specific SOC fractions vertical SOC is estimated.