Environmental drivers of space-time dynamics of vegetation in Bardia National Park, Nepal

Abstract

In the Terai Arc Landscape a number of endangered faunal species, including the elusive Bengal Tiger (Panthera tigris tigris), depend on sub-tropical tall grassland habitats. In Bardia National Park, Nepal, these grasslands are predominantly located close to the Geruwa river, the eastern branch of the Lower Karnali system. The presence of these grasslands is associated with disturbances, such as fluvial processes, forest fires and anthropogenic cutting and burning. In the last decades, these grassland habitats are observed to be vulnerable to encroachment of woody species. In 2009, a change in the hydrology occurred: the dominant discharge branch of the Karnali river, bifurcating at Chisapani, shifted from the eastern branch (Geruwa river) to the western branch (Kauriala river), with reduced discharges along the western border of Bardia National Park as a consequence. This study mapped the development of the vegetation pattern during the last decades in the western part of Bardia National Park and related environmental drivers to the observed vegetation dynamics to gain insight in underlying processes and drivers of change. The environmental drivers considered are precipitation, discharge, flood extent and forest fires. Land cover dynamics derived from annual land cover maps, for which remotely sensed imagery (Landsat) together with data from the field is combined in a supervised Random Forest classification model. In the annual land cover maps, two levels of detail are considered: the level 1 classification, from 1993 to 2019, has 4 classes and the level 2 classification, from 2013 to 2019, has 8 classes. Accuracies obtained for the level 1 and level 2 classification are +- 85 % and 72%, respectively. Landscape fragmentation metrics are used to quantify the development of grassland patches. These metrics indicate that the grassland patches decreased in number and perimeter length, indicating a decrease of heterogeneity of grassland in the landscape. In relation with environmental drivers, peak discharges coincide with successional resets before the shift in river course in 2009. After 2009, this was not observed, because the discharge through the Geruwa river has decreased. Since then coverage of vegetation close to the Geruwa river increased, mainly involving the expansion of alluvial tall grasslands (Saccharum spontaneum dominant). For precipitation, only extreme precipitation events have a signal of successional resets in the classifications. We could not show direct relations between the used forest fire dataset and changes in land cover. For grassland (phantas) of which is known that yearly burning takes place, succession to later successional stages is prevented and they have been present for a longer time than other grassland patches.