Semantic segmentation and change detection are two fundamental challenges in remote sensing, requiring models to capture either spatial semantics or temporal differences from satellite imagery. Existing deep learning models often struggle with temporal inconsistencies or in capturing fine-grained spatial structures, require extensive pretraining, and offer limited interpretability—especially in real-world remote sensing scenarios. Recent advances in diffusion models show that Gaussian noise can be systematically leveraged to learn expressive data representations through denoising. Motivated by this, we investigate whether the noise process in diffusion models can be effectively utilized for discriminative tasks. We propose Noise2Map, a unified diffusion-based framework that repurposes the denoising process for fast, end-to-end discriminative learning. Unlike prior work that uses diffusion only for generation or feature extraction, Noise2Map directly predicts semantic or change maps using task-specific noise schedules and timestep conditioning, avoiding the costly sampling procedures of traditional diffusion models. The model is pretrained via self-supervised denoising and fine-tuned with supervision, enabling both interpretability and robustness. Our architecture supports both tasks (SS and CD) through a shared backbone and task-specific noise schedulers. Extensive evaluations on the SpaceNet7, WHU, and xView2 buildings damaged by wildfires datasets demonstrate that Noise2Map ranks on average 1st among seven models on semantic segmentation and 1st on change detection by a cross-dataset rank metric (average F1 primary, IoU tie-break), while being 13× faster and 3× smaller than the generative diffusion baseline (DDPM-CD) due to its single-step discriminative inference. Ablation studies highlight the robustness of our model against different training noise schedulers and timestep control in the diffusion process, as well as the ability of the model to perform multi-task learning.

Accurate and timely burned-area mapping is essen- tial for emergency response, damage assessment, and post-fire recovery. However, strong spatial and temporal domain shifts - driven by variations in fire behavior, land cover, climate, and observation conditions - pose major challenges to generaliza- tion. Consequently, models trained on historical wildfire data often fail to transfer reliably to new regions or future fire seasons. Recently, geospatial foundation models (GFMs) trained on large-scale datasets have emerged as a promising direction for generalization in remote sensing. However, despite their scale and strong in-distribution performance, GFMs can still suffer from performance degradation under domain shifts. To this end, we propose a diffusion decoder that can be integrated with GFM encoders to enhance their cross-domain generalization. The decoder leverages diffusion noise as an implicit regularizer by training over multiple intermediate timesteps along the diffusion trajectory between pre- and post-fire imagery, encouraging the learning of representations that are robust to perturbations and transferable to unseen domains. We design three zero-shot protocols capturing geographic, temporal, and spatiotemporal shifts for wildfire Sentinel-2 imagery across the United States and Canada from 2017–2023. When integrated with three leading GFMs (TerraMind, Clay-v1.5, and Prithvi-v2), the performances improve by up to +4.8 F1 and +6.7 IoU points. Prithvi-v2 achieves the strongest absolute performance, while Clay-v1.5 exhibits the largest relative gains. Overall, our findings demonstrate that our diffusion decoder strengthens the generalization and zero-shot ability of GFMs under diverse environmental conditions.

Wildfire burned-area mapping is essential for damage assessment, emissions modeling, and understanding fire–climate interactions across diverse ecological regions. Recent geospatial foundation models provide strong general-purpose representations for satellite imagery, yet there is still no clear understanding of how to efficiently adapt these models for downstream Earth observation tasks, particularly under geographic and temporal domain shift. This study evaluates three state-of-the-art Geospatial Foundation Models (GFMs) - Terramind, DINOv3, and Prithvi-v2 - for burned-area mapping across the United States and Canada using Sentinel-2 data. Leveraging 3,820 wildfire events from 2017–2023, we conduct spatial and temporal generalization tests across diverse biomes. We systematically compare full fine-tuning, decoder-only fine-tuning, and Low-Rank Adaptation (LoRA) for adapting each model. Across all experiments, LoRA provides the strongest cross-domain generalization while updating less than 1% of parameters, demonstrating a favorable trade-off between accuracy and efficiency. Prithvi-v2 with LoRA achieves the highest overall accuracy and the larger improvement compare to full fine-tuning. These findings indicate that geospatial foundation models, when adapted using lightweight parameter-efficient methods such as LoRA, offer a robust and scalable solution for large-scale burned-area mapping. Code is available at https://github.com/alishibli97/wildfire-lora-gfm.