VegeDiff: Latent Diffusion Model for Geospatial Vegetation Forecasting

TL;DR

VegeDiff introduces a probabilistic, latent-diffusion framework for geospatial vegetation forecasting under climate-driven uncertainty. It combines a vegetation autoencoder to learn a latent representation of RGBN imagery with VegeNet, a DiT-based network that decouples the global impact of dynamic meteorological variables from the local influence of static environmental variables, forecasting future vegetation states in latent space and decoding back to RGBN imagery. Compared to deterministic baselines on EarthNet2021X, VegeDiff achieves lower RMSE and higher SSIM, particularly at longer lead times, and enables multi-perspective vegetation forecasting via indices such as NDVI, EVI, and SIPI. This approach provides a flexible probabilistic baseline for practical applications in agriculture management, disaster warning, and environmental monitoring, while highlighting opportunities to improve efficiency and transfer to other tasks.

Abstract

In the context of global climate change and frequent extreme weather events, forecasting future geospatial vegetation states under these conditions is of significant importance. The vegetation change process is influenced by the complex interplay between dynamic meteorological variables and static environmental variables, leading to high levels of uncertainty. Existing deterministic methods are inadequate in addressing this uncertainty and fail to accurately model the impact of these variables on vegetation, resulting in blurry and inaccurate forecasting results. To address these issues, we propose VegeDiff for the geospatial vegetation forecasting task. To our best knowledge, VegeDiff is the first to employ a diffusion model to probabilistically capture the uncertainties in vegetation change processes, enabling the generation of clear and accurate future vegetation states. VegeDiff also separately models the global impact of dynamic meteorological variables and the local effects of static environmental variables, thus accurately modeling the impact of these variables. Extensive experiments on geospatial vegetation forecasting tasks demonstrate the effectiveness of VegeDiff. By capturing the uncertainties in vegetation changes and modeling the complex influence of relevant variables, VegeDiff outperforms existing deterministic methods, providing clear and accurate forecasting results of future vegetation states. Interestingly, we demonstrate the potential of VegeDiff in applications of forecasting future vegetation states from multiple aspects and exploring the impact of meteorological variables on vegetation dynamics. The code of this work will be available at https://github.com/walking-shadow/ Official_VegeDiff.

Figures (8)

• Figure 1: Illustration of the geospatial vegetation forecasting task and models, where V indicates geospatial vegetation states, M indicates dynamic meteorological variables, and E indicates static environmental variables. (a) The overview of the geospatial vegetation forecasting task. (b) The overview of deterministic models performing the geospatial vegetation forecasting task. (c) The overview of VegeDiff performing the geospatial vegetation forecasting task.
• Figure 2: The overall structure of VegeDiff. VegeDiff models the high uncertainty of geospatial vegetation change process with diffusion model.
• Figure 3: The overall structure of VegeNet.
• Figure 4: NDVI images of sample inference results of ConvLSTM, Earthformer and VegeDiff on the EarthNet2021X test dataset.
• Figure 5: ARVI images of sample inference results of ConvLSTM, Earthformer and VegeDiff on the EarthNet2021X test dataset.