Climate change impact on flood hazard over Italy

Aims

• Flood hazard mapping over Italy
• Scientific, reliable approach
• Future projections

Models

• ICTP RegCM Regional Climate Model
• CHyM hydrological model
• CA2D hydraulic model

Project overview

Participants

• Erika Coppola
• Rita Nogherotto
• Filippo Giorgi

• Adriano Fantini
• Francesca Raffaele

1. Can a model chain of climate, hydrological and hydraulic models reproduce flood hazard?

2. How does climate change impact flood hazard over Italy?

3. What is the link between changes in precipitation extremes and changes in flood hazard?

Research questions

1.0 - Flood hazard: overview

RISK

=

HAZARD

×

EXPOSURE

×

COPING FACTOR

Event probability, Return Period

Goods, people and services exposed

Emergency plans, adaptation strategies

1.1 - Flood hazard: overview

What variables are we interested in?

DIRECT

• Flood extent
• Flood water depth
• Flood water speed

• Precipitation extremes
• Discharge extremes

PROXIES

1.2 - Flood hazard: overview

Methods used to obtain hazard maps:

• Historical discharge and flooding records
• Documentary evidence from past events
• Local surveys
• Hydrological + hydraulical modelling

1.3 - Flood hazard: overview

Flood hazard over Italy (PRESENT DAY)

Return Period:

100-200 yrs

CURRENT KNOWLEDGE

ISPRA data obtained from the single regional agencies.

Issues:

• Undisclosed methodologies
• Non-uniform approach
• No future projections

1.4 - Flood hazard: overview

Flood hazard over Italy (FUTURE CHANGE)

• Several studies available
• No study specific to Italy
• Relatively low resolution (>5km)
• No flood extents, only flood proxies (extreme discharge)

Results over Italy:

• General increase in flood proxies by the end of the century
• Especially for Northern Italy
• Low resolution does not resolve smaller basins

2.0 - The GRIPHO dataset

Two main goals:

• Driving the hydrological model with high resolution data
• Validating the regional climate simulations

Raw station data provided by Marco Verdecchia (CETEMPS):

• 2001 - present
• Hourly
• 3712 stations
• No quality check

2.1 - The GRIPHO dataset

Gridding

• 12km Lambert Conformal Conic grid

• Gridding method based on Delaunay polygons using SciPy's interpolate.griddata

• Simple, fast method which minimizes smoothing
• Similar to Norway's KLIMAGRID dataset (Mohr, 2008, 2009)
• NetCDF CF-compliant output format

Mohr M., 2008: New Routines for Gridding of Temperature and Precipitation Observations for seNorge.no

Mohr M., 2009: Comparison of versions 1.1 and 1.0 of gridded temperature and precipitation data for Norway

Velasquez N. et al., 2011: Rainfall distribution based on a Delaunay triangulation method

2.2 - The GRIPHO dataset

Issues:

• Varying station availability and density across the domain
• Inconsistencies, outliers and data errors
• Only ~15 years or data

TIME

FIRST-STAGE FILTERING

FLAGGING

FLAG CHECKING

MANUAL DATA CLEANING

CLEANED DATASET!

2.3 - The GRIPHO dataset

Cleaning procedure

2.4 - The GRIPHO dataset

Data cleaning (R95ptot = %pr above 95th pctl)

2.5 - The GRIPHO dataset

Metrics:

• Mean seasonal
• Extremes (R95ptot, R99ptot)
• Annual cycles
• Probability Density Functions

Validation against:

• E-OBS
• ARCIS
• HMR
• EURO4M-APGD

2.6 - The GRIPHO dataset

2.7 - The GRIPHO dataset

2.8 - The GRIPHO dataset

Conclusions

• Performance in the North similar to other high resolution datasets (EURO4M-APGD, ARCIS)
• Much finer details compared to HMR and E-OBS, especially for extremes
• Only station-based dataset over Italy providing hourly precipitation

Automated filtering

• Removal of extreme values > 200 mm/h
• Removal of isolated > 100 mm/h reports
• Removal of complete months with > 1800 mm (2.5 mm/h)
• Removal of continuous identical values

Rauthe et al. 2013; A Central European precipitation climatology–Part I: Generation and validation of a high-resolution gridded daily data set (HYRAS)

Isotta et al. 2013; The climate of daily precipitation in the Alps: development and analysis of a high‐resolution grid dataset from pan‐Alpine rain‐gauge data

Perry et al., 2009; The generation of daily gridded datasets of temperature and rainfall for the UK

Hiebl et al., 2017; Daily precipitation grids for Austria since 1961—development and evaluation of a spatial dataset for hydroclimatic monitoring and modelling

2.x - The GRIPHO dataset

Flagging

• Total % of valid values
• Total % of valid values ≠ 0
• % of events > mean + {10,15,25} SD
• % of events > median + {10,15,25} IQR
• % of events with 50 < pr < 100mm/h
• % of events with pr > 100mm/h
• % of the top {1,5,10} most common values

2.x - The GRIPHO dataset

Results on warning flags

2.x - The GRIPHO dataset

Data cleaning (mean PR)

Manual filtering on a monthly basis

2.x - The GRIPHO dataset

Outliers

2.x - The GRIPHO dataset

2.x - The GRIPHO dataset

Validation

2.x - The GRIPHO dataset

Validation

2.x - The GRIPHO dataset

Validation (PDF, detail)

2.x - The GRIPHO dataset

2.x - The GRIPHO dataset

3.0 - Regional Climate Model

3.1 - Regional Climate Model

Does the model perform well?

Validation for precipitation + temperature:

• Mean seasonal
• Extremes (R95ptot, R99ptot)
• Annual cycles
• Probability Density Functions

3.2 - Regional Climate Model

3.3 - Regional Climate Model

3.4 - Regional Climate Model

3.5 - Regional Climate Model

3.6 - Regional Climate Model

RegCM Conclusions

• Model precipitation and temperature generally in line with observations
• Increased projected average precipitation by the end of the century in winter in the north; decrease in the south and isles in summer            dipole
• Precipitation extremes projected to strongly increase
• The most extreme events increase more

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

TAS

PR

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

3.x - Regional Climate Model

4.0 - Cetemps Hydrological Model

CHyM Is a distributed (gridded) hydrological model. Peculiarities:

• Can build DEM from various sources, smoothing by cellula automata algorithms
• Can use several kind of inputs, such as station observations, gridded model data, etc.
• Designed to work on any domain
• Used daily at CETEMPS for operational forecasts
• Hourly NetCDF output

4.1 - Cetemps Hydrological Model

• 9 simulated domains
• Tested several Digital Elevation Models, chose HydroSHEDS
• Specific tuning for each region
• 300-900m resolution
• Argo and Marconi clusters
• 3000 runtime hours
• 100k core-hours
• ~35TB

Three PR drivers:

• GRIPHO (MM5) 2001-2016
• RegCM-ERA       1980-2016
• RegCM-HAD      1972-2099

4.2 - Cetemps Hydrological Model

Does the model perform well?

Validation only possible against a few discharge stations:

• Average discharge
• Yearly maximum discharge
• Projected Q100 discharge
• Standard hydrological model metrics (NSE, KGE, correlation, index of agreement, ...)

4.3 - Cetemps Hydrological Model

CHyM (GRIPHO)

4.4 - Cetemps Hydrological Model

4.5 - Cetemps Hydrological Model

4.6 - Cetemps Hydrological Model

CHyM Conclusions

• Model discharges in line with observations, for the few available stations
• Average discharge projected to slightly increase in the north; mixed changes elsewhere
• Extreme discharges projected to strongly increase by the end of the century (often >1.5x)
• Extreme discharge changes (Q100) do not match extreme precipitation changes            hydrological model adds useful information!

4.x - Cetemps Hydrological Model

Q100: 100 YEAR ESTIMATED DISCHARGE

GRIPHO

HadGEM

present

HadGEM

2020-2049

HadGEM

2070-2099

100-year discharges estimated following:

Maione et al., 2003: Regional estimation of synthetic design hydrographs

Thanks to Francesca Raffaele for the research work!

4.x - Cetemps Hydrological Model

96 river network reconstruction tests for each region

(manual and automatic, with distance measures)

4.x - Cetemps Hydrological Model

CHyM-OP reproduced domains:

4.x - Cetemps Hydrological Model

CHyM-OP reproduced domains:

4.x - Cetemps Hydrological Model

4.x - Cetemps Hydrological Model

So far:

• Performed ~2000 model simulations to find the best configuration for the river network reconstruction
• Identify and compare the reconstruction of the Po river with different metrics: mean distance, basin area, distance Q95...

4.x - Cetemps Hydrological Model

5.0 - CA2D hydraulic model

2D flood inundation model from Dottori and Todini, 2010, 2011, modified by Rita Nogherotto to run in parallel

5.1 - CA2D hydraulic model

• 90m resolution
• HydroSHEDS DEM and river channels
• Still ongoing, currently completed only for CHyM (GRIPHO)
• 5528 "virtual stations" cover all rivers every 5-10km
• Two case studies in North-Western Italy

5.2 - CA2D hydraulic model

RP = ~200yr

5.3 - CA2D hydraulic model

AdBPo

GRIPHO/ CHYM / CA2D

ISPRA

5.4 - CA2D hydraulic model

CA2D Conclusions

• Encouraging initial results
• Good results in the case study
• General agreement with ISPRA maps

5.x - CA2D hydraulic model

5.x - CA2D hydraulic model

"Virtual stations"

6.0 - Summary

New data:

• New state of the art GRIPHO precipitation dataset
• Two new RCM EURO-CORDEX simulations
• Three new high resolution CHyM hydrological simulations

6.1 - Research... answers

1. Can a model chain of climate, hydrological and hydraulic models reproduce flood hazard?

Yes, it can!

2. How does climate change impact flood hazard over Italy?

Increase in all flood proxies, sometimes > 150%

3. What is the link between changes in precipitation extremes and changes in flood hazard?

The two are often linked, but not always!

6.x - Visualization

An R/Leaflet tool for flood, river, DEM, basin and station visualization

6.x - Visualization

Click_edit: an R/Shiny tool for WYSISYG editing of NetCDF files