A Scale-Adaptive Framework for Joint Spatiotemporal Super-Resolution with Diffusion Models

TL;DR

This paper introduces a scale-adaptive framework for joint spatiotemporal super-resolution using diffusion models, enabling flexible upscaling across various factors with a unified architecture.

Contribution

It proposes a method to reuse the same model architecture for different super-resolution factors by retuning hyperparameters, enhancing transferability and efficiency.

Findings

Demonstrated on precipitation data over France, spanning SR factors from 1 to 25 in space and 1 to 6 in time.

Achieved a reusable architecture and tuning recipe for various joint spatiotemporal super-resolution tasks.

Retuning three hyperparameters allows the model to adapt to different scale factors effectively.

Abstract

Deep-learning video super-resolution has progressed rapidly, but climate applications typically super-resolve (increase resolution) either space or time, and joint spatiotemporal models are often designed for a single pair of super-resolution (SR) factors (upscaling spatial and temporal ratio between the low-resolution sequence and the high-resolution sequence), limiting transfer across spatial resolutions and temporal cadences (frame rates). We present a scale-adaptive framework that reuses the same architecture across factors by decomposing spatiotemporal SR into a deterministic prediction of the conditional mean, with attention, and a residual conditional diffusion model, with an optional mass-conservation (same precipitation amount in inputs and outputs) transform to preserve aggregated totals. Assuming that larger SR factors primarily increase underdetermination (hence required context and residual uncertainty) rather than changing the conditional-mean structure, scale adaptivity is achieved by retuning three factor-dependent hyperparameters before retraining: the diffusion noise schedule amplitude beta (larger for larger factors to increase diversity), the temporal context length L (set to maintain comparable attention horizons across cadences) and optionally a third, the mass-conservation function f (tapered to limit the amplification of extremes for large factors). Demonstrated on reanalysis precipitation over France (Comephore), the same architecture spans super-resolution factors from 1 to 25 in space and 1 to 6 in time, yielding a reusable architecture and tuning recipe for joint spatiotemporal super-resolution across scales.