Rapid Flood Mapping Using Inundation Libraries

Transcription

1 Rapid Flood Mapping Using Inundation Libraries Jude Kastens, Kevin Dobbs, James Halgren, Katherine Balster 2017 ASFPM Conference May 3, mi Kansas River Valley between Manhattan and Topeka jkastens@ku.edu

2 Terrain Processing: DEM (Digital Elevation Model) This DEM was created using LiDAR data. Shown is a portion of the river valley for Mud Creek in Jefferson County, Kansas. DEM (shown in shaded relief) 2

3 Terrain Processing: Flow Direction Each pixel is colored based on its flow direction. Navigating by flow direction, every pixel has a single exit path out of the image. Flow direction map (gradient direction approximation) 3

4 Terrain Processing: Stream Delineation The Mud Creek streamline is identified (shown in blue) using an appropriate flow accumulation threshold. Synthetic Stream Network 4

5 Terrain Processing: Floodplain Mapping The 10-m floodplain was computed for Mud Creek using the FLDPLN model. FLDPLN is a static, 2D hydrologic model that requires only DEM data as input. 10-m Floodplain (DTF Map) Using simple surface flow properties, FLDPLN identifies the depth-varying floodplain in reference to the input stream network (floodwater source). 5

6 Amazon River in Brazil (1700 km). 90-m SRTM DEM data were used. South America Amazon surface elevation drop in study area: 17 m 1 m per 100 km! 6

7 Example: Delaware River Basin above Perry Lake in northeast Kansas

8 Example: Walnut River Basin in southeast Kansas Each colored stream segment has its own inundation library Augusta Merged library

9 The FLDPLN ( Floodplain ) Model There are two ways that point Q can be flooded by water originating from point P: Backfill Flooding Spillover Flooding d water surface Q P P uphill flow (swelling) downhill flow (overland flow) Q } d Water seeks its own level Water flows downhill

10 Backfill Flooding accounts for floodwater expansion due to swelling processes water surface flow divide dry flood depth dry dry ground surface FLOODWATER SOURCE PIXEL OVER HERE flow directions PIXEL ON RIDGELINE 10

11 Spillover Flooding accounts for floodwater rerouting (alternative flow path development) water surface flow divide spillover flood depth flood depth ground surface FLOODWATER SOURCE PIXEL OVER HERE flow directions PIXEL ON RIDGELINE 11

12 PLAN VIEW illustrating backfill and spillover flooding R tributary channel SPILLOVER FLOODING flow divide flow divide flood source point P Q watershed boundary P Q or Depth To Flood (DTF) Contour P Q BACKFILL FLOODING flood source point P Q Spillover flooding meets backfill flooding P Q 12

13 Why Backfill Flooding Is Not Sufficient Here is what a DTF map looks like determined using only backfill flooding. Note the erroneous discontinuities. Flow divides in the flow direction map These are caused by ridgelines in the DEM. 13

14 Why Backfill Flooding Is Not Sufficient By backfill flooding using small flood depth increments, and allowing spillover flooding to occur on the floodplain boundary between iterations, the DTF discontinuity problem is mostly resolved. The 10-m steady state floodplain is shown, computed using the FLDPLN model and 0.5 m increments. 14

15 Longitudinal Floodplain Cross Section FLDPLN Model Solution Profile Overhead view flow Ground Normal Water Level H 2 O Floodwater Source Flood Stage 1 DTF Contour 1 Flood Stage 2 DTF Contour 2 Flood Stage 3 DTF Contour 3 Flood Stage 4 DTF Contour 4 Inflowing Channels Side-channel flow back into the main channel results in depth decay 15

16 Seamless modeling with FLDPLN spillover taper (depth decay) backfill taper (lake effect) maximum value composite* *Works for depth grids. Multi-segment merged DTF maps require minimum value compositing. combined 16

17 zoom area Seamless modeling with FLDPLN

18 Arghandab 5m floodplain

19 East trib 5m floodplain

20 West trib 5m floodplain

21 Combined 5m floodplain

22 Now let s see some actual flood extent mapping

23 Flood Extent Estimation (Example 1) Flooding along the Osage River in Missouri gage July 2007

24 Flood Extent Estimation (Example 2) June 13, 2008 Flooding on the Cedar River crested more than 11 ft above the historic record in Cedar Rapids, Iowa Cedar Rapids, IA

25 Example 3: 1938 Texas Flood Simulation adapted from Burnett (2008) Brady San Saba Menard 25

26 1938 Flood Depth Grid This flood depth grid was determined using Intermap elevation data. 26

27 Example 3 Verification Intermap N Oblique aerial photo* over San Saba, Texas, during a record flood that occurred in July FLDPLN floodwater surface estimates using different elevation datasets NED High water marks collected by the USACE in 1938 were used to model this event. *Burnett, J. (2008). Flash Floods In Texas. College Station, TX: Texas A&M University Press

28 Example 3 (continued) Intermap Oblique aerial photo* of San Saba during the 1938 flood (not necessarily at crest). Note the locations of the water tower & the courthouse (green dots). N NED N Reports and pictures in the Dallas Morning News, The Saba News and Star, and the Wichita Falls Record News show that in the City of San Saba, flood waters from the river spread through a great part of the business district and around the courthouse and spread over more than one-third of the City. -- excerpt from san-saba-texas/san-saba-texas.php *Burnett, J. (2008). Flash Floods In Texas. College Station, TX: Texas A&M University Press N 28

29 Example 4: Reconstructing the 1993 Missouri River Flood in Kansas* *KDEM request for 2011 floods 29

30 Rulo St. Joseph Atchison Leavenworth Sibley Kansas City 30

31 Rulo St. Joseph Atchison Leavenworth Sibley Kansas City 31

32 Rulo St. Joseph Atchison Leavenworth Sibley Kansas City 32

33 Rulo St. Joseph Atchison Leavenworth Sibley Kansas City 33

34 Example 5: Susquehanna River Tropical Storm Lee (Sep 2011)

35 Susquehanna River Water Surface Elevations CKLN6 BNGN6 Gage heights represent 9/8/11 flood crest VSTN6 From existing hydraulic model 500-yr 100-yr 50-yr 10-yr Filled DEM

36

37 zoom area

38 Flood Depth Grid estimate for September 2011 flood event

39 Kansas Coverage (5-m LiDAR) 39

40 Conceptual Framework Data Prep Database / Server Implementation DEM input NED LiDAR InterMap SRTM other DEM Conditioning (ex. NLD, NID) Arc Hydro Tools & Stream Segmentation FLDPLN Model (MATLAB) SLIE Database Segmented Library of Inundation Extents GIS Server Custom Extent Map / Depth Grid SLIE Selectors Observed (point) Gauge Data HWM Library Ground Observer Satellite (raster) GFDS (low res) DFO (mod res) Other Modeled HEC-RAS HAZUS Other Client Applications 40

41 41

42 Recent Research 42

43 Flood Boundary Points as Gage Proxies Lower Kavango (Africa)

44 Flood Boundary Points as Gage Proxies Southeast Kansas Study Using HEC-RAS 2D 44

45 HEC-RAS 2D Breaklines / Depth Grids 1K cms HEC-RAS 4.8K cms HEC-RAS 45

46 HEC-RAS Water Surface Elevation 46

47 FLDPLN Inundation Library Extent 47

48 5m FLDPLN DTF 48

49 Flood Boundary Points as Gage Proxies Southeast Kansas Implement sampling strategy for simulation analysis 5 points was smallest number examined 49

50 Flood Boundary Points as Gage Proxies Southeast Kansas Using selected points: - Identify flood source pixel for each boundary point - Build DTF profile for stream pixels 50

51 Flood Boundary Points as Gage Proxies Southeast Kansas Use DTF profile to create library-based flood extent Compare to HEC- RAS 2D simulated flood extent Compute F-Statistic: F = 100 * ( A op /(A o +A p -A op )) A o : observed area of inundation (HEC- RAS 2D) A p : predicted area of inundation (FLDPLN) A op : area that is both observed and predicted as inundated 51

52 Agreement between simulated and predicted flood extents μ = (for 4.8K) = (for 1K) SBP = simulated boundary point 52

53 Other Applications for the FLDPLN Model 53

54 River valley boundary delineation masking for identification of floodplain wetlands 54

55 DTF maps provide a useful guide when specifying cross sections for hydraulic modeling. 55

56 Identifying High-Impact Riparian Areas Fourmile Creek, Morris County, KS 56

57 Thanks for Listening Any Questions? 57