Dr Beena Balan Sarojini

Abstract:

Anthropogenic warming has led to an unprecedented year-round reduction in Arctic sea ice extent. This has far-reaching consequences for indigenous and local communities, polar ecosystems, and global climate, motivating the need for accurate seasonal sea ice forecasts. While physics-based dynamical models can successfully forecast sea ice concentration several weeks ahead, they struggle to outperform simple statistical benchmarks at longer lead times. We present a probabilistic, deep learning sea ice forecasting system, IceNet. The system has been trained on climate simulations and observational data to forecast the next 6 months of monthly-averaged sea ice concentration maps. We show that IceNet advances the range of accurate sea ice forecasts, outperforming a state-of-the-art dynamical model in seasonal forecasts of summer sea ice, particularly for extreme sea ice events. This step-change in sea ice forecasting ability brings us closer to conservation tools that mitigate risks associated with rapid sea ice loss.

Abstract:

Up to now, relatively little effort has been dedicated to the quantitative assessment of the differences in spatial patterns of model outputs. In this paper, we employed a variogram-based methodology to quantify the differences in the spatial patterns of root-zone soil moisture, net radiation, and latent and sensible heat fluxes simulated by three land surface models (SURFEX/ISBA, JULES and CHTESSEL) over three European geographic domains – namely, UK, France and Spain. The model output spatial patterns were quantified through two metrics derived from the variogram: i) the variogram sill, which quantifies the degree of spatial variability of the data; and ii) the variogram integral range, which represents the spatial length scale of the data. The higher seasonal variation of the spatial variability of sensible and latent heat fluxes over France and Spain, compared to the UK, is related to a more frequent occurrence of a soil-moisture-limited evapotranspiration regime during summer dry spells in the south of France and Spain. The small differences in spatial variability of net radiation between models indicate that the spatial patterns of net radiation are mostly driven by the climate forcing data set. However, the models exhibit larger differences in latent and sensible heat flux spatial variabilities, which are related to their differences in i) soil and vegetation ancillary datasets and ii) physical process representation. The highest discrepancies in spatial patterns between models are observed for soil moisture, which is mainly related to the type of soil hydraulic function implemented in the models. This work demonstrates the capability of the variogram to enhance our understanding of the spatiotemporal structure of the uncertainties in land surface model outputs. Therefore, we strongly encourage the implementation of the variogram metrics in model intercomparison exercises.