Add model card for TerraScope

Hi! I'm Niels from the Hugging Face community science team. I've opened this PR to add a model card for TerraScope.

This includes:
- Metadata for better discoverability (pipeline tag, library name, license).
- Links to the paper ([TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation](https://huggingface.co/papers/2603.19039)), the project page, and the code repository.
- A description of the model's key capabilities in geospatial reasoning.

README.md

@@ -0,0 +1,39 @@
+---
+license: apache-2.0
+library_name: transformers
+pipeline_tag: image-text-to-text
+tags:
+- earth-observation
+- geospatial-reasoning
+- multimodal
+- remote-sensing
+---
+
+# TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation
+
+[**TerraScope**](https://shuyansy.github.io/terrascope/) is a unified Vision-Language Model (VLM) specifically designed for Earth Observation (EO). It addresses tasks that require grounding complex spatial reasoning in precise pixel-level visual representations.
+
+- **Paper:** [TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation](https://huggingface.co/papers/2603.19039)
+- **Project Page:** [https://shuyansy.github.io/terrascope/](https://shuyansy.github.io/terrascope/)
+- **Repository:** [https://github.com/shuyansy/Earth-Observation-VLMs](https://github.com/shuyansy/Earth-Observation-VLMs)
+
+## Model Description
+
+TerraScope delivers pixel-grounded geospatial reasoning with two key capabilities:
+1. **Modality-flexible reasoning:** It handles single-modality inputs (optical or SAR) and adaptively fuses different modalities into the reasoning process when both are available.
+2. **Multi-temporal reasoning:** It integrates temporal sequences for change analysis across multiple time points.
+
+The model was trained on **Terra-CoT**, a large-scale dataset containing 1 million samples with pixel-level masks embedded in reasoning chains. It demonstrates significant performance on **TerraScope-Bench**, a benchmark evaluating both answer accuracy and mask quality for geospatial tasks.
+
+## Citation
+
+If you find this work useful, please consider citing:
+
+```bibtex
+@article{shu2026terrascope,
+  title={TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation},
+  author={Shu, Yan and Ren, Bin and Xiong, Zhitong and Zhu, Xiao Xiang and Demir, Beg{\"u}m and Sebe, Nicu and Rota, Paolo},
+  journal={arXiv preprint arXiv:2603.19039},
+  year={2026}
+}
+```