Master Thesis defense by Emma Johanna Fiedler

Title: Assessing the Influence of Spatial Resolution and PET Methodology on Drought Representation - A Multi-Reanalysis Comparison

Abstract:
Droughts are an increasing risk to water resources and agriculture in Denmark, yet their representation in gridded reanalysis products remains under-characterised. This thesis compares three reanalysis datasets – ERA5, CERRA, and DANRA – in their ability to capture drought conditions in Denmark over the common period 1991–2023 using the Standardised Precipitation Evapotranspiration Index (SPEI) and the Standardised Precipitation Index (SPI). Four Potential Evapotranspiration (PET) methods are additionally evaluated: Thornthwaite, Oudin, Hargreaves, and Modified Makkink. Station observations are used to validate both the reanalysis input variables and the PET estimates. The reanalyses mostly agree on temperature statistics across the year, but diverge more substantially in summer precipitation, where ERA5 and CERRA show large wet biases relative to observations while DANRA agrees more closely. These differences propagate into the drought indices, where inter-reanalysis spread exceeds inter-method spread, identifying reanalysis choice as the dominant source of uncertainty in SPEI estimates over Denmark. The four PET methods themselves produce meaningfully different evapotranspiration values, but these differences are mostly reduced by the standardisation in SPEI. Thornthwaite is the exception, being a systematic outlier particularly in the wet tail of the spring season. In summer, a directionally consistent increase in extreme drought frequency in the modern period (2007–2023) relative to a historical baseline (1991–2007) is identified across all datasets, but cannot be statistically separated from the influence of the exceptional 2018 drought event. Spring shows no consistent signal across analysis periods. The most robust finding across all products and methods is a shift toward wetter conditions, confirmed by bootstrap analysis to be insensitive to individual years and by the longer ERA5 record (1940–2024).

Supervisor: Jens Hesselbjerg Christensen
Censor: Peter Aakjær