Day 1: Climate Data Science Grant Awardee Presentation - Salman Khan | ADIA Lab Symposium 2025

Day 1: Climate Data Science Grant Awardee Presentation - Salman Khan | ADIA Lab Symposium 2025

🎙 Salman Khan 👥 824 📅 November 5, 2025 ⏱ 24 min 👁 102 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

climate copilotagentic pipelinemultimodal foundation modelwater securityfood security

Summary

Salman Khan presents a research proposal for ‘Climate Copilot’, an AI agentic pipeline aimed at enhancing climate resilience in the MENA region. The project addresses the ‘MENA twin crisis’ of water depletion and crop stress, exemplified by the shrinking Mosul Dam lake. The core idea is to develop a multimodal foundation model that integrates satellite imagery, IoT sensor data, climate models, and textual reports, and then use an agentic layer to decompose complex queries into step-by-step reasoning, leveraging tools and models for simulation and scenario analysis. The project focuses on two applications: water security and food security. It plans to utilize open datasets like GRACE, Sentinel-2, ERA5, CMIP6, and FAO reports, while also engaging with national agencies for proprietary data. The team has prior experience with EarthDial, a conversational EO model, and CopperFM, a multimodal model. Challenges such as hallucination, uncertainty quantification, and tool chain audit are acknowledged. The potential impact includes aiding water authorities, farmers, and policymakers with explainable AI-driven insights. The presentation concludes with Q&A on interpretability, data security, and collaboration with Destination Earth.

176 words

Critical Evaluation

Value of the Information & Strength of the Argument

The presentation offers a clear and well-structured vision for applying agentic AI to climate challenges, building on the presenter’s prior work in multimodal earth observation models. The argumentation is logical, moving from problem definition to proposed solution, and acknowledges limitations such as hallucination and the need for uncertainty quantification. However, it remains at a proposal stage, with no empirical results or comparative analysis to substantiate the claimed advantages over existing siloed models.

81 words

Title / Content Match

The title accurately reflects the content: a grant awardee presentation on climate data science, delivered by Salman Khan at the ADIA Lab Symposium 2025.

Quality & Reliability

7/10

Presentation by a recognized researcher in computer vision, with references to published works (EarthDial, CopperFM) and established datasets (ERA5, CMIP6, GRACE). However, it is a proposal, not peer-reviewed results, and lacks detailed methodological validation.

Key Moments

Cited Sources

  • EarthDial — Mentioned as prior work presented at CVPR, a conversational model for earth observation data.
  • CopperFM — Mentioned as prior work, best paper candidate at ICCV, expanding input modalities.
  • Aurora model — Mentioned as a weather forecasting model with extensive variables.
  • ERA5 — Mentioned as reanalysis data product for historical climate data.
  • CMIP6 — Mentioned as long-term climate projections.
  • GRACE — Mentioned as satellite mission for groundwater estimation.
  • Sentinel-2 — Mentioned as multispectral satellite data source.

Concurring Sources

  • EarthDial — Prior work by the presenter, supporting the feasibility of multimodal EO models.
  • CopperFM — Prior work by the presenter, supporting the feasibility of multimodal EO models.

Contribution & Novelties

The project proposes a novel integration of agentic AI with climate models and earth observation data, aiming to bridge the gap between siloed models and provide explainable, actionable insights. The emphasis on traceability and uncertainty quantification is a valuable contribution. However, the novelty is incremental, building on existing foundation models and agentic frameworks.

Pour aller plus loin :

85 words

Radar Profile

The radar profile shows high scores in technical level and information quality, reflecting the advanced AI concepts and structured presentation. The lower score in reliability is due to the proposal nature and lack of empirical validation. Overall, the profile indicates a technically strong but not yet validated research direction.

Reliability 7/10

💬 Sur les 0 commentaires analysés, aucune tendance n'est disponible.