Does enforcing physics improve evapotranspiration downscaling? A controlled comparison of mass-conserving and direct deep-learning models for WaPOR
Abstract
Abstract Satellite evapotranspiration products underpin agricultural water accounting. Yet the field-scale (20 m) data irrigation needs exist for only a handful of sites, whereas coarse (≈ 300 m) products span the entire African continent. This motivates deep-learning downscaling of the continental product to field scale. Physics-informed networks that enforce hard conservation constraints are widely advocated, on the premise that they improve accuracy and transfer. Yet this premise is rarely tested in a controlled way. To our knowledge, this is the first controlled test of it for evapotranspiration downscaling. It is also the first to adapt image-restoration and geospatial foundation models to the task, and to transfer across contrasting African agro-climates. We downscale WaPOR actual evapotranspiration (ETa) from Level-1 (≈ 300 m) to Level-3 (20 m). Inputs, training, sites, and protocol are held fixed across ten African irrigation pilots. We compare three direct super-resolution models (a compact SwinIR transformer, a Prithvi Vision Transformer, and their ensemble) against a physics-compliant network (PRSN). Direct models explain 78–91% of L3 ETa variance and transfer zero-shot to unseen agro-climates. They roughly halve naive-upsampling error and beat an operational per-pixel baseline. A compact 1.8 M-parameter SwinIR matches the 300 M-parameter Prithvi foundation model. The physics-compliant models keep their guarantees but are less accurate at every site, by ≈ 6% to ≈…
Read the paper
Related studies
- Preliminary Toxicity Analysis of 3DCRT versus IMRT on the High Dose Arm of the RTOG 0126 Prostate Cancer Trial
- Adaptive Radiotherapy for Head and Neck Cancer: Initial Clinical Outcomes from a Prospective Trial
- American Association of Physicists in Medicine Task Group 263: Standardizing Nomenclatures in Radiation Oncology
- Motion Management Strategies and Technical Issues Associated with Stereotactic Body Radiotherapy of Thoracic and Upper Abdominal Tumors: A Review from NRG Oncology
- High-dose-rate interstitial brachytherapy as a monotherapy for localized prostate cancer: treatment description and preliminary results of a phase I/II clinical trial
- Highly efficient training, refinement, and validation of a knowledge-based plan quality control system for radiotherapy clinical trials
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.