Pollen-based normalized difference vegetation index reconstruction and its application to two Holocene sites of the Tibetan Plateau

Abstract

The Tibetan Plateau (TP) vegetation exhibits high sensitivity to climate fluctuations and disturbances, making it a focal area for studying alpine ecosystem dynamics. Normalized difference vegetation index (NDVI), as an indicator of vegetation growth and health status, is one of the most effective satellite-based vegetation parameters that are used to monitor vegetation dynamics. However, long-term NDVI data are not available due to the short temporal span of instrumental observations. Quantitative reconstructions of past vegetation based on pollen data can provide crucial evidence for understanding long-term vegetation evolution and validating the vegetation/climate model to enhance the accuracy of future projection. On the basis of the relationships between modern pollen data and remote sensing NDVI, we apply the modern analogue technique (MAT) to reconstruct vegetation NDVI from pollen data of the TP. The accuracy of the reconstruction was evaluated by comparing the reconstructions with remote sensing observations. Results showed that a search window of 10 km radius around sampling sites combined with the inverse-distance weighting method were the most suitable for extracting vegetation parameters to establish the pollen-NDVI reconstruction model. Under the above parameter settings, the pollen-based modern NDVI reconstructions were in good agreement with the observed ones (R² = 0.79). The pollen-based reconstructions accurately reflect the spatial pattern of NDVI observed by remote sensing on the TP, although NDVI values were overestimated in sparsely vegetated areas and underestimated in dense forest areas in a few cases. The Holocene NDVI changes of the Zoige Basin and Hurleg Lake were then reconstructed based on their fossil pollen records by applying the pollen-NDVI reconstruction model. The reconstructed paleo-NDVI changes at the two sites were highly consistent with their Holocene regional vegetation and climate patterns, which further confirmed the reliability of recovering vegetation parameters from pollen data on the TP.