Investigation into the long-term response of vegetation to dam operation in a large river using Landsat time-series data

Xi Shi,Junqiang Xia,Meirong Zhou,Shanshan Deng,Jian Sun

DOI: https://doi.org/10.1016/j.jhydrol.2024.130977

IF: 6.4

2024-03-02

Journal of Hydrology

Abstract:Vegetation is an important component of riparian ecosystems, providing fundamental habitats for numerous species. Over the past few decades, as more and more large river systems have been heavily dammed, the vegetation communities in downstream floodplains have potentially been affected, but the vegetation response to this type of human-mediated disturbance is not well known. In this study, the long-term vegetation dynamics in a large river, the Middle Yangtze River, were investigated using Landsat time-series data, and the focus was placed on the relict mid-channel bars downstream of the Three Gorges Dam (TGD). To reveal the vegetation dynamics, a phenological metric, maturity average vegetation index (MAVI), was defined based on the vegetation index during the maturity phase. Simultaneously, an unsupervised change detection method was developed to identify flood inundation areas by differencing water index value from pre- and post-flood Landsat images. The main findings are as follows: (i) post-TGD flood events exerted a more detrimental impact on the floodplain vegetation compared to pre-TGD events. For some bars, the flood impact on vegetation converted from positive to negative in response to a ∼90% decline in sediment concentration post-TGD, while for some other bars the negative impact was further exacerbated; (ii) The MAVI-cumulative curves for zones under different elevations approached each other in response to a 12-year absence of episodic overbank flooding after the TGD, and for stable bars the curves converged in a time scale of ∼10 years, exhibiting diminished vegetation heterogeneity thereafter; (iii) Of the methods for deriving vegetation dynamics and inundation pattern, the vegetation index, OSAVI, was suitable for characterising the phenology of perennial reed species, and the unsupervised change detection method can generate an accurate flood inundation map for mid-channel bars with limited ground truth data. Overall, the vegetation responses to the TGD operation can help understanding the dam impact on vegetation and facilitate the floodplain restoration projects in the heavily dammed fluvial system.

geosciences, multidisciplinary,water resources,engineering, civil