models: a geospatial investigation Mark Bernhofen PhD Student - - - PowerPoint PPT Presentation

River size representation in global flood models: a geospatial investigation

Mark Bernhofen

PhD Student - University of Leeds cn13mvb@leeds.ac.uk

GFP Conference, Guangzhou. Session 3 – 12 June, 2019

Overview

Slide 1/12

How do the Global Flood Model river size thresholds affect exposure estimates?

• Follows on from the global flood model intercomparison work (Trigg et al, 2016).
• Study found 30-40% agreement between 6 global flood models in Africa.
• Difficult to quantify the significance of river size representation on model

disagreement using different global flood models.

• Need some independent (from GFMs) measure of flood extent.

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC Slide 2/12

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC 5000 km2 threshold Slide 2/12

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC 1000 km2 threshold Slide 2/12

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC 500 km2 threshold Slide 2/12

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC 50 km2 threshold Slide 2/12

Global Model Thresholds: Nigeria

Drainage Thresholds Global Model 50 km2 Fathom 500 km2 U-Tokyo, ECMWF 1000 km2 CIMA-UNEP 5000 km2 JRC 5 km2 threshold Slide 2/12

• Terrain model developed by Nobre et al. (2011)
• Needs only terrain derived datasets as inputs:
• Digital Elevation Model (DEM)
• Drainage directions
• River upstream drainage area
• Has been successfully used in studies as a quasi
• r (0 dimensional) flood model.
• Quick to run at large-scales.
• Easily produce an auxiliary dataset of river size

(upstream drainage area)

Height Above Nearest Drainage (H (HAND)

Slide 3/12

from Nobre et al. (2011)

HAND Model Setup

Slide 4/12 Strahler Stream Order HAND height 1 2 m 2 2 m 3 3 m 4 4 m 5 5 m 6 6 m 7 7 m 8 8 m 9 10 m 10 12 m 11 12 m

• Using MERIT Hydro data
• HAND heights were “calibrated” using 100y flood

maps in Nigeria and the UK.

• For rivers with Strahler Stream Orders >=6, JRC’s

100 year flood map was used in Nigeria.

• JRC’s model was the best of the GFMs in a previous

validation study in Nigeria (Bernhofen et al, 2018)

• For smaller rivers with stream orders <=7 the

Environment Agency’s 100 year flood zone map for the UK was used.

HAND Model Quality

Slide 5/12 HAND Model Aggregated GFMs (from Trigg et al, 2015)

Models in Agreement

1m 12m HAND Height

HAND Model Quality

Slide 5/12 Good Fairly good Fair Fairly poor Poor HAND Model Performance Koppen-Geiger Climate Classifications

Exposure Data

Slide 6/12

• Using Facebook and Columbia University’s

High Resolution Settlement Layer (HRSL) (30m resolution) in the 22 countries it is available.

• Use WorldPop data (100m resolution) in all

countries.

• Can compare results to see how population

data affects the exposure estimates.

HRSL availability

Flood Exposure - Afr frica

Slide 7/12 “%” relates to total population

19% 15% 11% 9% 7% 50 100 150 200 250 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2)

Population Exposed to Flooding (WorldPop)

Flood Exposure - Afr frica

Slide 7/12 “%” relates to total population

12% 9% 6% 5% 4% 20 40 60 80 100 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2)

Population Exposed to Flooding (WorldPop)

HRSL vs. . WorldPop

Slide 8/12

• 28%
• 29%
• 31%
• 35%
• 12%
• 40%
• 35%
• 30%
• 25%
• 20%
• 15%
• 10%
• 5%

0% 5-50 50-500 500-1000 1000-5000 5000< Percentage Change River Drainage Area Thresholds (km2)

Percentage change in exposure estimates when using HRSL vs. WorldPop Results correspond with Smith et al (2019) Nature Comms.

Flood Exposure – HRSL

Slide 9/12

11% 8% 6% 5% 4% 10 20 30 40 50 60 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2)

Population Exposed to Flooding (HRSL)

“%” relates to total population

Slide 10/12

In Interesting Countries

11% 8% 5% 4% 1%

0.5 1 1.5 2 2.5 3 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2) Population Exposed to Flooding (HRSL)

Madagascar

12% 10% 9% 9% 3%

0.2 0.4 0.6 0.8 1 1.2 1.4 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2) Population Exposed to Flooding (HRSL)

In Interesting Countries

Benin

Slide 10/12

8% 7% 6% 6% 6%

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 5< 50< 500< 1000< 5000<

Population

Millions

Minimum Drainage Area Threshold (km2) Population Exposed to Flooding (HRSL)

In Interesting Countries

Ivory Coast

Slide 10/12

• Different GFM minimum threshold river sizes result in significantly different

exposure estimates (on average, up to a factor of 2.0).

• Certain country’s exposure estimates are more sensitive to river size

representation than others.

• Improving river representation beyond current GFM limits could

incorporate an additional 30% of potential exposed population into the modelling frameworks

• WorldPop population data overestimates exposure compared to Facebook

High Resolution Settlement Layer, especially on smaller rivers.

Conclusions

Slide 11/12

• Fine tune the HAND model calibration to include different climate zones.
• Expand study globally
• Global picture of flood exposure from different size rivers.
• Can inform global flood modelling community about where it’s important

to consider river size representation and which countries would benefit from improved representation

Next xt Steps

Slide 12/12

Thanks for listening.