Capturing drought and post-drought impacts on forest GPP using a range of satellite indices

Europe’s forests are under threat from rising air temperatures and increasingly severe and frequent drought. Yet the carbon storage capacity of these forests is a vital part of the European Green Deal and efforts to mitigate climate change. Due to the multiple demands placed on forests, it is essential to develop accurate methods to monitor their carbon sink strength over large spatial extents, for example by using satellite data. However, capturing drought effects on forest gross primary productivity (GPP) using remote sensing is not straight forward. Drought causes multiple changes to tree physiology and structure over varying timescales that are not always reflected in the optical vegetation indices most commonly used by the remote sensing community.   

As part of the EU Horizon project CLIMB-FOREST, we tested the ability of fifteen indices derived from MODIS to detect the negative impact of severe drought on forest GPP. These included indices that respond to changes in leaf pigments, canopy structure, canopy water content and land surface temperature. The analysis compared GPP during drought and non-drought periods at 14 flux tower sites across Europe between 2003-2023. We found that drought during the mid- to late-growing season led to a decline in GPP compared to non-drought years whereas drought during the early growing season was associated with increased GPP. The only MODIS indices that showed significant changes during drought compared to non-drought conditions were NDVI, CCI, PRI and LST. However, several other indices were significantly lower in the year after a drought event, despite GPP returning to average values, which may be evidence of drought legacy effects on forests. Further work will examine how remote sensing indices are linked to changes in ecosystem functional properties during and after drought. By disentangling the relationship between remote sensing indices and drought-related changes to forest carbon fluxes and function, our findings will help improve the accuracy of remote-sensing based models of forest GPP.