Hamburg Habitat NDVI Deviations During Compounding Hot and Dry Summers

This data set contains the results of the study: Nadine Kaul, Nikola Lenzewski, Olaf Conrad, Jürgen Böhner, Kai Jensen, Benjamin Poschlod (2026): Remote sensing reveals heterogeneous responses of urban vegetation types to compound heat and drought events in Hamburg, Germany. Urban Forestry & Urban Greening, 2026, 129489, https://doi.org/10.1016/j.ufug.2026.129489.

The associated study analysed short-term drought responses of semi-natural and urban habitats in the city of Hamburg, Germany, to the compound hot and dry summers of 2018, 2020, and 2022, using high-resolution Sentinel-2 NDVI (Normalized Difference Vegetation Index) time series and data of the official habitat mapping survey of the city of Hamburg.

The data set includes:
- an overview of the included habitats and their categorisation into wet and dry sites (file 1)
- the mean NDVI differences between phenological drought curves and reference curves for each analysed habitat (file 2)

The mean NDVI differences were acquired via the following workflow *:
Sentinel-2 Level-1C scenes with low to moderate cloud cover were downloaded for the years 2015-2024, atmospherically corrected, and cleaned from cirrus, cloud, and cloud-shadow contamination. NDVI was derived from bands 4 and 8A at 10 m resolution. A vegetation mask was generated via unsupervised classification of two cloud-free reference Sentinel-2 scenes and clipped to the selected vegetated habitat types (parks, grasslands, heathlands, shrublands, forests, bogs, fens).

For each habitat polygon, pixel-level time series were filtered to remove outliers (NDVI < 0 and dates after which the NDVI increased by more than 0.4 within 20 days). Habitat-level median time series were then constructed for dates where at least half of the habitat’s pixels were available, and only for polygons with ≥10 pixels. These median series were smoothed using an iterative Savitzky–Golay filtering procedure to approximate the upper NDVI envelope.

Drought years (2018, 2020, 2022) were identified using the Standardised Precipitation Evapotranspiration Index (SPEI), while all remaining years were aggregated into a phenological reference curve (median across dates for the years 2019, 2021, 2023, 2024; 2015-2017 were excluded due to to insufficient scene availability). NDVI differences were calculated as the mean deviation between each drought-year curve and the reference curve during the vegetation period (May 1–September 30), with negative values indicating reduced greenness.

* For further details and references, please refer to the associated publication.