Predictability of weather and climate

Abstract:

Posit numbers, a recently proposed alternative to floating-point numbers, claim to have smaller arithmetic rounding errors in many applications. By studying weather and climate models of low and medium complexity (the Lorenz system and a shallow water model) we present benefits of posits compared to floats at 16 bit. As a standardised posit processor does not exist yet, we emulate posit arithmetic on a conventional CPU. Using a shallow water model, forecasts based on 16-bit posits with 1 or 2 exponent bits are clearly more accurate than half precision floats. We therefore propose 16 bit with 2 exponent bits as a standard posit format, as its wide dynamic range of 32 orders of magnitude provides a great potential for many weather and climate models. Although the focus is on geophysical fluid simulations, the results are also meaningful and promising for reduced precision posit arithmetic in the wider field of computational fluid dynamics.

Abstract:

Cities worldwide are increasingly facing the challenges of heat stress, a problem expected to worsen with ongoing climate change. The lack of detailed, city-specific data hinders effective response measures and limits the adaptive capacity of urban populations. In this data descriptor, we introduce a comprehensive database providing climate and heat stress information for 142 cities globally, covering the present and extending projections up to 2100 across three distinct climate scenarios, including two overshoot scenarios. This dataset includes 34 heat stress indicators at a spatial resolution of 100 meters, offering a unique database to identify vulnerable areas and deepen the understanding of urban heat risks. The data is presented through an accessible, user-friendly dashboard, enabling policymakers, researchers, and city planners, as well as non-experts, to easily visualise and interpret the findings, supporting more informed decision-making and urban adaptation strategies.

Abstract:

Plain Language Summary: Climate model emulators are approximations of climate models that provide a quick and low‐cost alternative to exploring future climate scenarios. Traditional emulators generate large amounts of data covering the whole world, which still need to be condensed when exploring local and regional impacts. In this paper, we propose a new emulator based off image compression techniques. The setup allows one to “zoom” in and out from global to regional to local levels, providing user‐relevant information across scales. It furthermore conserves both large‐scale and local features and can be run in minutes. Given its versatile framework, the approach is easily extendable to new variables, and in this paper we demonstrate its ability to jointly capture temperature and relative humidity.

Abstract:

Over the past 20 years, extreme event attribution has developed rapidly, providing a wide range of methods to attribute weather events - from unconditioned probabilistic to strongly conditioned storyline approaches. Advancing the field now requires combining results from multiple methods, allowing more robust conclusions drawing from various lines of evidence. Yet, doing so remains challenging. We call for closer interaction within the attribution field to develop approaches with method comparison in mind. We highlight the need to explicitly define the research questions answerable by specific methods, and to clearly outline the limitations of each method.