Uncertainty in big data applications: lessons from climate simulations

The science part investigated how data of unknown quality can be used to validate and calibrate climate/weather-impact models. It identified the hurdles for such an approach to be implemented in an operational model. The philosophy part developed an uncertainty typology for decision support, to further include uncertainty in Big Data. Argument analysis was applied to the predictive inferences in the scientific part. In this project tools were developed to use low-cost sensor data in climate impact studies. The synthesis part analysed conditions for transferring results to other fields and consequences for the scientific methodology and understanding.

Contrary to popular belief, Big Data science is not free of theory. But philosophical research on how it uses theories is sparse. In climate and weather research, advantages and limitations of process-based vs. statistical approaches have not been explored in detail. So far, Big Data has rarely been used to test models that take into account weather, climate and societal choices.

The goals of the project were to produce:

1. a prototype of a climate-impact model using Big Data approaches to study the potential and limitations of such methods and quantify their uncertainty in current events and trends in extreme weather and impacts;
2. a typology of the uncertainties and underlying arguments;
3. criteria for the transferability of the results to other scientific fields.

There is tremendous economic and societal value in accurate quantifications of weather and climate risks, but damage estimates are often done only in hindsight. Tools and services are missing that are likely technologically feasible and meet the needs of the end-users. This project contributed to better tools and a better conceptual understanding to overcome hurdles towards operational implementation. Ideas and results were regularly exchanged with MeteoSwiss in developing the operational side.

Several pioneering studies in this project have explored both the conceptual and practical opportunities and challenges in using Big Data tools and data of unknown quality in climate modelling and impact studies, paving the way for future applications. With the availability of much more data and computing capacity, this field is growing quickly, but important questions remain on how to combine Big Data with process understanding, and how make progress on interpretable machine learning methods. In cases where repeated verification is not possible, the process of establishing confidence is challenging and often relies on understanding the relevant processes and drivers. Machine learning methods are inherently limited in this respect, yet are powerful in extracting information patterns that otherwise would be inaccessible. The key will be to combine the best of both worlds, and this project has laid important groundwork for that.

Combining theory with Big Data? The case of uncertainty in prediction of trends in extreme weather and impacts