Yang et al. (2026) Phenology constrains the predictability of soil respiration in cropping systems revealed through interpretable machine learning

Identification

• Journal: Agricultural and Forest Meteorology
• Year: 2026
• Date: 2026-08-30
• Authors: Jing Hu, Dana M. Miles, Frances A. Podrebarac, Daniel McCraine, Joby M. Prince Czarnecki, John P. Brooks
• DOI: 10.1016/j.agrformet.2026.111443

Research Groups

• Geosystems Research Institute, Mississippi State University
• Department of Civil, Environmental and Construction Engineering, University of Central Florida
• USDA Agricultural Research Service

Short Summary

This study examines the predictability of soil respiration in cropping systems using stage-explicit modeling and interpretable machine learning. The results show that phenology constrains the predictability of soil respiration, with predictive skill confined to specific phenological stages.

Objective

• Investigate how the importance and response patterns of environmental and canopy variables vary across phenological stages.
• Examine whether commonly used predictors adequately represent underlying biological processes across development.

Study Configuration

• Spatial Scale: Field-scale (56 experimental plots) in a no-till dryland system.
• Temporal Scale: Four years, with measurements taken on 38 sampling dates for cotton and 28 sampling dates for corn.

Methodology and Data

• Models used: Random Forests (RF)
• Data sources: Soil respiration measurements, soil temperature, soil moisture, UAV imagery, and vegetation indices

Main Results

• The predictability of soil respiration varies systematically across phenological stages.
• Vegetation indices provide informative signals for soil respiration prediction during select phenological periods.
• Cover crop residual effects primarily function as a magnitude filter, modulating the baseline intensity of soil respiration without fundamentally altering its sensitivity to environmental drivers.

Contributions

• This study provides empirical insights into the stage-dependent limitations of commonly used predictors and informs future feature selection and data collection strategies in soil respiration modeling.
• The results demonstrate that phenology constrains the predictability of soil respiration, highlighting the need for stage-explicit modeling approaches.

Funding

• USDA Agricultural Research Service (ARS) project number [insert project number]
• Mississippi State University's Geosystems Research Institute (GRI)
• University of Central Florida's Department of Civil, Environmental and Construction Engineering

Citation

@article{Yang2026Phenology,
  author = {Yang, Jing and Hu, Jing and Miles, Dana M. and Podrebarac, Frances A. and McCraine, Daniel and Czarnecki, Joby M. Prince and Brooks, John P.},
  title = {Phenology constrains the predictability of soil respiration in cropping systems revealed through interpretable machine learning},
  journal = {Agricultural and Forest Meteorology},
  year = {2026},
  doi = {10.1016/j.agrformet.2026.111443},
  url = {https://doi.org/10.1016/j.agrformet.2026.111443}
}