Simultaneous emulation and downscaling of modeled soil state variables with machine learning

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

Not reported

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Science

Operational functions

No function tag matched Derived by keyword rules; see methodology

Problem the AI is intended to solve

We propose a lightweight, computationally efficient machine learning (ML) model capable of emulating the LIS-based soil moisture and soil temperature and downscaling them from a native 10 km resolution to 1 km resolution. Our approach is extendable to other variables as long as a non-linear relationship between meteorological forcing and the variable of interest can be conceptualized as modulated by local conditions (elevation, soil type, land cover, vegetation). Then, a branched neural network (NN) architecture structurally represents this relationship. As a part of the project, different NN architectures and input combinations have been tested and assessed using SHapley Additive exPlanations (SHAP) values and ablation analysis. Currently, the downscaled product is being validated and compared to other high-resolution products.

Expected benefits

Using the proposed method, it is possible to obtain LIS-like quality predictions for soil state variables in seconds, as well as provide unprecedented for LIS downscaled to 1 km data relevant to a wide variety of applications. The low computational cost of the inference and the ability to resolve fine-resolution features expedite obtaining the crucial information for decision-making.

System outputs

The model outputs are emulated LIS-like soil moisture and soil temperature (predictions), as well as downscaled soil moisture and soil temperature.

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

Not reported

Custom-developed code

Not reported

Public source code

Not reported

Sourcing

How it was built

Not reported

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Not reported

Privacy Impact Assessment

Not reported

Authorization to Operate

Not reported

Demographic features

Not reported

Training and evaluation data

Not reported

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported