Floods, Floodplains, and Fluvial Studies

Type: Paper Session

Time: 11/7/2020 03:30 PM (Eastern Time (US & Canada)) to 4:45 PM

Session Description and Agenda

Special Paper Session: Floods, Floodplains, and Fluvial Studies

1) Getting your feet wet: Strategies for implementing field-based stream research in large undergraduate classes

2) Variations in the Water Quality of Tributary Streams within the Cache River Watershed in Eastern Arkansas

3) Drivers for vegetation recovery on point bars – modeling and mapping vegetation regrowth in the mid-Apalachicola River, Florida

4) Origin Hypothesis for North Carolina Outer Banks Barrier Islands, Capes, and Coastal Plain Surface Hydrology Patterns

5) Insights from Paired Profiles of a Disturbed Coastal Plain River

Drivers for vegetation recovery on point bars – modeling and mapping vegetation regrowth in the mid-Apalachicola River, Florida

Session: Floods, Floodplains, and Fluvial Studies
Type: Paper Session

Abstract

Anthropogenic activities alter river-floodplain systems and leave landform legacies. One of the examples is the mid-Apalachicola River, where intensive dredging activities occurred from 1950s to early 2000s. The direct dredging impacts included channel widening, stage decreases, and dredged disposal features, which led to ~263% enlargement of point bars from 1941 to 2004. A decade after dredging curtailed, the point bars shrunk ~15% by regaining vegetation. This study utilized logistic regression (LR) and random forest (RF) models to determine the drivers behind vegetation regrowth on point bars. We applied 1999 and 2019 aerial photos to define the recovery and non-recovery zones, and randomly sampled 3,200 points from 16 selected point bars. Eight vegetative, hydrologic, and topographic variables were chosen for modeling. The preliminary results from two models indicated that the distance from 1999 tree line was the most important variable to both models, which contributed 100 to the RF model. Roughness, slope, and flooding depth were significant based on LR, and had an importance measure in the 30s for the RF model. Velocity, stream power, and shear stress were not significant with importance values of 10 or less. According to the 10-fold cross-validation, the accuracy of LR and RF were 0.86 and 0.92. These results indicated that tree seeding distance was the most important driver to recovery. The roughness and slope variables denoted that the micro topographic relief helped to trap organic matter and shelter young growth. Our findings can be applied to active restoration.

Authors