Antediluvian Climate Data
There’s a legacy problem in the climate space that’s ripe for technology-enable solutions. Along a well-trod Web 2.0 avenue, far from the technology frontier, it’s a data sharing problem. Since we get daily weather forecasts from apps with instant updates, one might assume that atmospheric weather data flows into global climate models in near real-time too. It doesn’t. In fact, the data that informs the scope and scale of measurable climate change currently is more than a decade old.
This was a key takeaway from an article by climatologists Gavin Schmidt and Zeke Hausfather in the NYTimes at the end of last year. As the piece makes clear: “the data that went into the latest round of climate model simulations are based on observations that only run through 2014…Similarly, the forecasts are stuck with scenarios that were common in the early 2000s.” It’s now ten years and counting since 2014, when bitcoin was $310 and all our headphones still had wires. A lot has changed. There’s little analysis of data that’s ten years old through which we could find our 2025 selves, let alone project our future global circumstances. The ten hottest years ever happened in the last decade, along with 204 natural disasters declared in the United States costing $1.4 trillion.* None of this data is in current climate models. The article continues, “to fix this means more comprehensive and faster data gathering from satellites. This needs to be matched by a commitment by the 30 labs worldwide that maintain the earth climate system models to update their simulations to reflect the latest data.”
Sometimes, the existence of dynamic problem-solving technologies are taken for naturally doing so once suitable applications are known. Adoption and saturation are not automatic of course and fixes for climate data sharing may be found via “picks-and-shovels” or “plumbing” tools which successfully aided the proliferation of the SaaS solutions over the past 15 years. These supplemental services streamline interconnection, especially as conduit to data-heavy end products like global climate simulations. Fortunately, there are data infrastructure and ecosystem maps that are overflowing with startup logo solutions focused on transforming unstructured data and disparate data silos into interoperable shared data systems available and accessible to any authorized participant. Standard elements like robust APIs, cloud computing, and distributed data warehouses should all accelerate the pace of climate data retrieval and sharing.
This is not to suggest that delays in the bureaucratic exchange of information across languages, borders, and globally dispersed organizations can be fixed in technology-enabled finger snap. But communication breakdowns, prioritization challenges, and misaligned timeframe expectations are largely human-to-human problems. Problems that can be fixed with existing, out of the box technology products today — and have been in many industries populated by their own opaque, glacial, bureaucratic companies and organizations.
My friend Sanjiv, Co-Founder and CEO of Texture, recently wrote a post calling for collaborative data partnership amongst clean power sources in order to advance and strengthen renewables viability on the grid in aggregate. I believe many of his points regarding standardization and efficiency born out of ecosystem-wide collaboration could be viable and transformative for data distribution and latency status quo that hamstrings global climate analysis today.
One last thing, this gigantic climate data system will absolutely benefit from the AI-ML frontier. It just needs the data first. Without the data — or anywhere near current data — there aren’t worthwhile inputs for agentic AI to analyze or for LLMs to be trained on. With those inputs up to date, I anticipate multimodality classification will improve the accuracy of climate model forecasts, while climate model simulations will include multi-agent systems that capture complex climate interactions, including decision-making scenarios based on economic incentives, policy regulations, and social norms. But that’s for another post.