Challenges in climate analytics validation

Can climate models reproduce observed trends? The answer can be challenging. Our new review paper in Science Advances led by Isla Simpson and Tiffany Shaw discusses challenges and ways forward in confronting climate models and observations. It's tricky. Climate models and observations may disagree (1) by chance, due to unforced internal variability, (2) due to error in the model response, (3) due to inaccurate prescribed external forcings, (4) due to incomplete or uncertain observations or (5) due to inappropriate comparison methods. The paper discusses ways forward in disentangling the reasons for potential mismatches between observed and simulated trends. It provides a long catalogue of examples of success, discrepancies and unclear situations that require further attention. https://lnkd.in/dHrEJfDh Let by Isla Simpson and Tiffany Shaw with Paulo Ceppi, Amy Clement, Erich Fischer, Kevin Grise, Angeline Pendergrass, James Screen, Robert Jinglin Wills, Tim Woollings, Russell Blackport, Joonsuk Kang, and Stephen Po-Chedley supported by US CLIVAR

Our new article will appear on IEEE Geoscience and Remote Sensing Magazine soon “From Data to Policy: Strengthening Essential Climate Variable Monitoring with Deep Learning Algorithms and Data Quality Standards” Jean-Philippe Montillet, PhD ; Gael Kermarrec; Ehsan Forootan; C K Shum; Shaoxing Mo; W. Finsterle, K.S. Samse This review emphasizes the importance of ensuringdata quality, traceability, and consistency to derive reliable features from ECV datasets, addressing challenges such astemporal and spatial coverage gaps, calibration discrepancies, and harmonization across diverse sources. A preprint version can be found here: https://lnkd.in/ezzTx7ha

New perspective out in Earth System Dynamics! Climate extremes are increasing, but understanding their societal impacts remains difficult. In this new perspective led by Gabriele Messori, we discuss three key challenges facing the field: • limited and inconsistent impact data • difficulties understanding the processes leading to impacts • limitations of future projections The paper also highlights exciting opportunities, from LLMs extracting impact data, to interdisciplinary approaches, and storylines for exploring plausible future scenarios. A call for stronger collaboration across disciplines and sectors to better understand, and ultimately reduce, the impacts of climate extremes. Great to contribute alongside Emily Boyd, Joakim Nivre and Climes - Swedish centre for impacts of climate extremes 🔗 https://lnkd.in/exWtdG_5

  “Uncertainty in climate modeling and impact assessment”   Model-based climate change impact assessments, which rely on interconnected systems of models, play a key role in disaster risk management and climate change adaptation.   Traditional climate-induced flood risk analyses follow a modular top-down approach aligned with the now well-established 1994 Technical Guidelines from the Intergovernmental Panel on Climate Change (IPCC). A brief overview of the process follows below:   This approach starts with greenhouse gas (GHG) emission scenarios that “force” global climate models (GCMs) or Earth system models (ESMs).  The GCMs and ESMs models generally yield climate change projections at a coarse resolution compared to the needs of most impact modelers.   The coarse-resolution projections are then refined to regional or local scales using statistical and/or dynamical methods via regional climate models (RCMs).  The climate variables are often bias adjusted against observations, whereafter they can serve as more reliable input to hydrologic or hydraulic models to simulate climate impacts on water systems on either regional or local scales.   The simulated impacts can further feed into a variety of risk models, such as economic damage cost models, that translate hazards, exposure and vulnerability into risks of economic losses or the potential negative impacts on health and safety.   The combined model chains inform cost–benefit or cost-efficiency analyses and support a wide range of decision-making in water management.  Running models in a sequence, however, systematically PROPAGATES uncertainties from one model to the next, leading to CASCADING uncertainties. Understanding this uncertainty cascade is vital for asserting CONFIDENCE in models and conducting robust disaster risk assessment and climate change impact adaptation.   While detailed modeling frameworks and uncertainty analyses have become increasingly common due in part to improved data and computing power, the challenge of UNCERTAINTY PROPAGATION has been named one of the 23 unsolved problems in hydrology; it remains so today.     Despite its recognized importance, relatively FEW studies have systematically analyzed how uncertainty is sampled and propagated, and accumulated across models.    See Mik-Meyer et al. (2026) in WIRES Climate Change, “Uncertainty Representation and Propagation in Flood Risk Modeling Under Climate Change: A Systematic Review”