Fig 1.
Execution of probHAND model includes four steps: 1) create HAND map from DEM, 2) extract hydraulic geometry from HAND map for each reach, for a range of water stages which correspond to HAND elevations, 3) build synthetic rating curves to relate stage to discharge, and 4) translate rating curve to probabilistic floodplain map using flood frequencies, considering uncertainties in input parameters used to develop synthetic rating curves and identify flood frequencies. Input datasets and variables are highlighted by blue boxes. Spatial datasets in figure are from the University of Vermont’s Spatial Analysis Lab.
Table 1.
General inputs for probHAND model, specifying inputs for the Lake Champlain Application.
Fig 2.
Area of probHAND model application; Lake Champlain Basin, Vermont, USA.
We executed the model for 2,083 river-km within 93 different HUC12 units. Model integrity was evaluated from a 1D hydraulic model of three HUC12’s in the Mad River Valley and from 42 observed high water marks within the Winooski River basin. Spatial datasets in figure are from the Vermont Center for Geographic Information and available for open access download at https://geodata.vermont.gov/.
Table 2.
Description of PDF’s specified for Monte Carlo uncertainty analysis in Lake Champlain Basin application.
Fig 3.
Evaluation of integrity of probabilistic floodplain maps from probHAND model (“HAND”) compared to 1D hydraulic model (“HEC”) output for recurrence interval floods. (A) We found that floods with higher recurrence intervals had higher F-statistics (measure of model agreement where 1.0 is perfect agreement) and that 50% probability maps best mimicked hydraulic model output. (B) Example comparison of HAND and HEC models for 100-year flood. Spatial datasets in figure are from the University of Vermont’s Spatial Analysis Lab.
Fig 4.
Evaluation of performance of probHAND probabilistic floodplain maps relative to surveyed high water marks. (A) First, we translated high water marks (HWM) to edge of water (EOW) locations. (B) We then identified the highest probability map, for the approximate recurrence interval flood, which encompassed the EOW point. (C) We found that floodplain map probabilities well-represented the distribution of observed flood extents. Cumulative plot of HWM observations that increasingly fall within lower probability (higher percentile) maps. The band represents uncertainty in the location of the HWM observations. Spatial datasets in figure are from the University of Vermont’s Spatial Analysis Lab.
Table 3.
Average parameter values for valley setting groups, classified by reach slope.
Fig 5.
Summarized by recurrence interval flood (A&C) and valley setting, for the 100-yr flood (B&D) in the Lake Champlain Basin, Vermont. Floodplain width is calculated as the total mapped area divided by the total stream length (A&B), and greatly differs among valley setting groups (B). The proportional increase of the total potential width (i.e., the 95th percentile map) does not differ substantially for floods of varying magnitude (C) but does differ based on valley setting (D). Uncertainty in input parameters translates to larger differences in floodplain width in settings with high channel slope that have narrower valleys, than for floodplains in settings with either low or moderate channel slopes.
Fig 6.
The proportion of variance in model output (predicted stage for a given recurrence interval flood) described by four input parameters, based on the original distribution (right) and a narrower subset (left). (A) Although the proportion contributed by geom decreases with increasing flood magnitude and the proportion of variance contributed by QRI increases, the contribution of each parameter remains relatively similar (e.g., geom is greatest and S is least). Narrowing the distribution also reduces the variability contributed by geom while increasing QRI. (B) Valley characteristics mediate the contribution to variance, which is also influenced by the width of the probability distribution.