FIMevaluation

Framework for the evaluation of Flood Inundation Map (FIM) using a remote sensing derived benchmark FIM. The framework includes preprocessing of the benchmark layer and calculation of evaluation metrics.

Two ArcGIS Pro Notebooks (using ArcPy).

Overview:

1. Pre-processing Benchmark FIM (FIMGapFiller_BenchmarkProcessing.ipynb): 1.1 Remove small ‘flooding’ clusters 1.2 Enhancing the Remote Sensing FIM (GapFill/Region Growth) 1.3 Removing Permanent Water Bodies (PWB) 1.4 Matching Benchmark and Model FIMs (extent, resolution, and alignment) 1.5 Prepare the final benchmark layer – classify gap-filled and PWB as NoData

2. Calculating Evaluation Metrics (ModelEval.ipynb): 2.1 Re-classify benchmark [0,2] and predicted [1,2] FIM 2.2 Generate a match classification raster (TP, TN, FP, FN) 2.3 Calculate evaluation metrics (CSI, TPR, FAR, F1, ACC)

Demo Data:

Case study: Neuse River near Goldsboro NC - Hurricane Matthew, October 2016

ArcGIS Geodatabase files:

1. ModelEvalWorkshop_Workspace.gdb.zip - input data as well as satellite imagery of the flood.
2. ModelEvalWorkshop_Workspace_Outputs.gdb.zip - output raster for comparison

Download from: Here.