Deforestation and Flooding in the Lower Roanoke River Basin

Abstract

The large natural forest ecosystems in the Lower Roanoke River Basin, in northeast North Carolina, are home to numerous and diverse plant and animal species. However, these unique and precious forest ecosystems have been progressively threatened by deforestation and flooding in recent decades. Logging, agriculture, development, recreational use, and reservoir construction all could cause direct loss of floodplain forests. Changes in landscape, especially deforestation, conducted on the floodplains can cause indirect impacts on the floodplain hydrology. For example, flood events may occur with greater frequency in some areas due to increased upstream impervious surfaces and loss of vegetation buffers. At the same time, dams altered the natural flow, and in particular, have impacted the timing and intensity of overbank flow into the floodplain. This change in hydrology and flooding may lead to consequences for the floodplain plant and animal communities. The objectives of this project are to deepen the understanding of the 2 interacted factors of deforestation and flooding concerning the Roanoke River Basin by 1). investigating the deforestation trends in the past 20 years, 2). analyzing the flood frequencies and duration in recent years, and 3). mapping the flood extents by a remote sensing model. Temporal and spatial trends of deforestation in the Lower Roanoke River were analyzed by the forest loss data from the Global Forest Change database accessed on Google Earth Engine, and the vegetation species of the removed forests were investigated. Over the past 20 years, there were about 1290 km2 of forest loss happened in the Lower Roanoke River Basin, of which 610 km2 in 2001-2010 and 680 km2 in 2011-2020. Over the same 10-year period, forest loss increased by 11.5% after 2010. Spatially, deforestation was mainly distributed on the downstream banks and increased in these areas after 2011. In the past 2 decades, 8.1% or 48 km2 of deforestation occurred in the 100-year floodplain. Similar to the total deforestation, the forest loss areas in floodplains also increased after 2011 but at a higher increase rate of 18.8%. The largest loss of vegetation species in deforestation areas was hardwood. Oak - Sweetgum Floodplain Forest was the most removed vegetation type in both the floodplains and it was also the second-largest vegetation type of the deforestation areas in the whole basin before and after 2010. The flow in the Lower Roanoke River Basin is heavily dominated by 3 upstream dams. Given the dam capacity and empirical observations, a flood event was defined as a continuous period of that discharge of the Roanoke Rapids Dam above 20,000 cfs in this study. All such periods from 2016 to 2021 were screened, and there were 25 flood events in total. During the 6 years, the number of flood events ranged from 1 to 6, showing a seasonal trend of more flood events in winter and spring, and less in summer and autumn. In addition to flood frequency, the inundation time in the floodplain forests was studied by the continuous water level data from 14 monitoring sites along the Roanoke River. For all the flood events, the time required for the forest to dry out varied widely, with an average of 25 days to 40 days. For the monitoring sites, the upstream monitoring ones were underwater for a longer time, the downstream sites needed a medium time, and the sites in the middle basin went back dry the most quickly. Another important finding was the inundation in the forests needed a long time to recede. Even though the dam discharge periods were only about 1-2 weeks, the water remained on the floodplain for up to 57 days. In summary, the floodplain forests were under serious flooding pressure because of the long inundation time, which varied a lot, depending on location, flood events, topography, land cover, and other factors. It is very necessary to understand where the inundated forests are during flood events to study how forest ecosystems respond to flooding stress. A remote sensing model using Sentinel-1 radar data was built to identify the flood extent of a specific flood event by a random forest machine-learning algorithm. The flood extents of 2 flood events in March 2019 and March 2021 were mapped. The resulted flood extent maps had high accuracies. The overall accuracy for March 2019 was 85.6% and that for March 2021 was 89.7%. The most common misclassification was between dry forest and flooded forest due to their similar remote sensing signatures in the predictor composites. Both flood extents overlapped well with the 100-year floodplain in the middle and lower basin, validating the 100-year floodplain was a good predictor of flood extent in this area. But there were areas flooded in both events but not on the floodplain, which needed special attention to flooding. In conclusion, forest loss was accelerating in the Lower Roanoke River Basin, especially on the floodplains. The basin was still at high risk of flooding in winter and spring, and the floodplain forests would be under high flooding pressure because of the long time for water to recede. Remote sensing, in particular with radar data, had been proven as a feasible way to map the flood extent of a specific flood event, which can be a good reference for forest management and dam management. With deforestation and flooding both considered, the 100-year floodplain should be the focus of forest management and conservation work in the Lower Roanoke River Basin. Increased knowledge about shifts in forest practices, water flow responses, and flood extents may inform and benefit future land, forest, and dam management in the Lower Roanoke River Basin.