DeepExtremeCubes: Integrating Earth system spatio-temporal data for impact assessment of climate extremes

TL;DR

The DeepExtremeCubes database is introduced, tailored to map around these extremes, focusing on persistent natural vegetation, to streamline data accessibility, structuring, pre-processing, and enhance scientific reproducibility, and facilitate biosphere dynamics forecasting in response to compound extremes.

Abstract

With climate extremes' rising frequency and intensity, robust analytical tools are crucial to predict their impacts on terrestrial ecosystems. Machine learning techniques show promise but require well-structured, high-quality, and curated analysis-ready datasets. Earth observation datasets comprehensively monitor ecosystem dynamics and responses to climatic extremes, yet the data complexity can challenge the effectiveness of machine learning models. Despite recent progress in deep learning to ecosystem monitoring, there is a need for datasets specifically designed to analyse compound heatwave and drought extreme impact. Here, we introduce the DeepExtremeCubes database, tailored to map around these extremes, focusing on persistent natural vegetation. It comprises over 40,000 spatially sampled small data cubes (i.e. minicubes) globally, with a spatial coverage of 2.5 by 2.5 km. Each minicube includes (i) Sentinel-2 L2A images, (ii) ERA5-Land variables and generated extreme event cube covering 2016 to 2022, and (iii) ancillary land cover and topography maps. The paper aims to (1) streamline data accessibility, structuring, pre-processing, and enhance scientific reproducibility, and (2) facilitate biosphere dynamics forecasting in response to compound extremes.

Figures (6)

• Figure 1: An example of CHD detection in the Dheed database showing the evolution across Europe of the maximum daily temperature (Tmax, top row), the Precipitation - Evapotranspiration balance averaged over the previous 30 days (PE30, second row), along with the threshold of 0.01 on the ranked values used to detect extremes (Rank, first and second rows), the synthesis of the four indicators (Event-Cube, third row) and the labelled CHD events lasting at leat three days (Label-Cube, fourth row) from 2019-06-27 to 2019-07-01.
• Figure 2: A schematic diagram illustrating the DeepExtremeCubes database development and validation workflow. It includes: (a) The Dheed dataset presenting an example of a CHD event in this data. (b) The CHD event days map showing locations that experienced 10 or more event days according to the Dheed event detection dataset. (c) Minicubes that experienced 10 or more CHD days ("extreme" minicubes) and those that did not experience any CHD days in the Dheed dataset ("non-extreme" minicubes). (d) A demonstration of a minicube, which includes remote sensing images, climatic and meteorological variables, Digital Elevation Model (DEM), land cover, etc. (e) The representativeness and spatial autocorrelation of the DeepExtremeCubes dataset for minicubes with different land cover types. (f) The spatial data split of DeepExtremeCubes for potential users to train their forecasting models.
• Figure 3: The distribution of all sampled minicube locations. (a) Minicubes that experienced 10 or more CHD days ("extreme" minicubes) and (b) minicubes that did not experience CHD events during 2016 and 2021 ("non-extreme" minicubes). The land cover types depicted represent either the pure land cover type for the pure land cover minicubes or the dominant land cover for the minicubes covered by mixed land covers.
• Figure 4: Land cover representation in the DeepExtremeCubes dataset. The two plotted datasets use different reference points. The red bars consider the global areas of selected land cover as 100%, while the light and heavy orange bars treat the total sum of land cover types across all minicubes as 100%.
• Figure 5: The proportion within a certain distance where an "extreme" minicube can find a "non-extreme" minicube covered by the same land cover.