Review of Environment, Energy and Economics - Re3 Improvements in Flood Risk Assessment: Evidence from Northern Italy
 

 

Print this article Print this article
Jun
11
2015
 
Improvements in Flood Risk Assessment: Evidence from Northern Italy
by Mattia Amadio, Jaroslav Mysiak, Lorenzo Carrera and Elco Koks
Environment - Articles
 

The assessment of potential economic damage caused by floods is commonly done via methodologies based on Stage-Damage Curves (SDC), which provide a relation between the depth of water and the economic damage on a specific land use. SDC are developed for site-specific analysis but seldom calibrated or tested for transferability purpose. In Italy no specific damage functions have been developed so far, despite damage reports being collected after every major flood. Here a refined SDC model is tested against a flood event in Northern Italy. SDCs calibration is underpinned with empirical data from compensation records. Our framework includes the assessment of asset losses, focusing on urban and agricultural land, and production losses. While the first are calculated based on land use, production losses are measured through the spatial distribution of Gross Value Added.

Key-words: Flood Risk, Stage Depth Damage Curves, Italy       
JEL classification: Q54

See other useful links

Suggested citation: Amadio, Mattia, Mysiak, Jaroslav, Carrera, Lorenzo, Koks, Elco,     
Improvements in Flood Risk Assessment: Evidence from Northern Italy
(June 11, 2015). Review of Environment, Energy and Economics (Re3), http://dx.doi.org/10.7711/feemre3.2015.05.005

Introduction
The EU Floods Directive (FD, 2007/60/EC) manifested a shift of emphasis away from structural defence approach to a more holistic risk management, with structural and non-structural measures having the same importance. The FD compels identification of areas exposed to flood hazard and risk, and adoption of measures to moderate flood impacts. A sound, evidence-based risk assessment should underpin public disaster risk reduction and territorial development policies. Stage-damage curves (SDC) are a customary tool for assessing economic structural damage (Genovese 2006; Messner et al. 2007; Thieken et al. 2009; Jongman et al. 2012) as a function of flood attributes (primarily depth, less often speed and persistence) and land use. However, the SDC models have rarely been tested against empirical data. The modelled cost estimates are afflicted by uncertainties stemming from the variability of assets value and the damage sustained, and the quality of the hazard modelling (Messner et al. 2007; Merz et al. 2010; De Moel and Aerts 2011). This uncertainty can be reduced if empirically ascertained damage data can be used to test and calibrate the SDC models, with better results if this empirical damage data reflects the same or similar economic and social circumstances as those for which the future potential damage has to be addressed (Luino et al. 2009; De Moel and Aerts 2011). Another drawback of the method is the implicit assumption that the highest possible damage remains constant throughout the year. This does not hold for sectors whose production is concentrated in some months, such as growing season in agriculture. Moreover, SDC approximates damage only on tangible assets and hence is not able to determine output losses in terms of foregone production. In this paper we explore ways to improve the damage and loss assessments for the sake of a better risk assessment and management.

The development of a sound flood damage model applicable for the Italian economic and social circumstances is essential for well-designed and informed flood risk management policies. Different SDC models have been reported in literature, but most of them have been developed for site-specific context and rarely tested for transferability. SDC based on empirical material from Italy are rare (Molinari et al. 2013; Scorzini and Frank 2015). This is despite the common practice of state compensation for households’ (private) losses for which certified damage reports are collected. In our analysis we first test the applicability and transferability of SDCs based on the declared households damage from the 2014 Secchia flood in Emilia-Romagna administrative region. Successively we describe a crop-specific specific model for agricultural losses, better suitable for compensation claims (Forster et al. 2008; Tapia-Silva et al. 2011; Twining 2014). Ultimately, we explore the use of Gross Value Added (GVA) as an indicator of exposure for production losses (Peduzzi et al. 2009). Similar methods as those explored in this paper have been tested elsewhere at the national (Winsemius et al. 2013) and international scale (Ward et al. 2013).

2. Model performance

2.1 Model comparison for structural damage estimates
We have tested two up-to-date SDC models on example of the Secchia 2014 flood : DamageScanner model (De Moel et al. 2014; Koks et al. 2014), hereafter referred to as SDC-1; and the model developed by the EU Joint Research Centre (Huizinga 2007), referred to as SDC-2. The two models yield damage values that differ by 170 million (one third using the lower estimate as a basis) as shown in Figure 1. Besides, there is a sizeable divergence in distribution of the estimated damage across the land use categories (Figure 1, left). The SDC-1 yields a damage (per hectare) that is four times higher than SDC-2 for industrial land use category. Contrariwise, SDC-1 estimated damage is lower than that produced by SDC-2 by a factor 0.7 for residential land use category and only one fifth for rural category.     

Figure 1 - (left) output of the damage model for the 2014 flood event among aggregated land uses; (right) comparison of SDC models output for urban areas against registered compensation requests from households.

2.2 Model performance against empirical data
Figure 1 (right) shows the comparison of the SDC-1 and SDC-2 damage estimates with the empirical (reported) data on damage sustained. Overall, SDC-1 overestimates declared damage in residential areas by a factor 4.5, but for the urban area outside buildings (which includes shared spaces, squares, streets and parked vehicles) this difference peaks factor 9.2. SDC-2 results are even larger, 13 times greater than those observed. The damage shares between structure, mobile goods and private vehicles simulated by SDC-1 resemble [Note 1]  the ratios of declared damage. In the next step we have chosen SDC-1 for calibration. The SDC-1 was chosen also because it is able to disaggregate structural and content-wise damage in isolated dwellings and built areas. Both estimated and declared damage are geocoded and aggregated in 250m grid. The matching cells between simulated and empirical damage are 61 over 157, corresponding to 83% of simulated damage and 75% of declared damage. The calibration hereafter is carried out only on matching cells using regression analysis under the hypothesis of linear relationship. For each land use category, the maximum damage value is individually adjusted using the B (slope) coefficients as scaling factor. Figure 2 shows the results of linear regression between SDC-1 output and empirical damage before and after calibration for total (A), structural (B) and content (C) damage categories.

Figure 2 - Scatterplot showing empirical damage (X axis) and SDC results (Y axis) per grid cell using original land use values (cross indicator, dotted line) and calibrated ones (circle indicator, black line) for: A) total residential area; B) building structure; C) buildings content.


The pre-calibration output overestimated the total damage in residential areas by a factor 4.5-7 depending on the within-urban land use category. The calibrated damage values are regressed with the observed/reported damage with good results (R2=0.8) for all categories except the public urban areas with parked vehicles. This is probably due to the simplistic assumption that private vehicles are uniformly distributed along the streets. For buildings structure and content the coefficient (B) is close to 1.0, and the final output overestimate recorded residential damage by just 6% (Table 1).

Table 1 - Exposed area, observed and simulated damage inclusive of regression results for each land use category tested against empirical data.


2.3 Agricultural losses
The agricultural damage for major land use categories (i.e. arable crops, vineyards, permanent crops) is modelled as a weighted average of the product of average crop yields and producers’ prices aggregated over Utilized Agricultural Area (UAA) (Altamura et al. 2013). The maximum damage D per crops i at the time t of disaster strike is DtMAX = Σni=1 (GSPi -ΣEndt DCi] x UAAi/UAA). This is the difference between estimated Gross Saleable Product [Note 2]  (GSP) at the time of harvesting production and the not-realised direct production costs [Note 3] (DC) from t to the end of production cycle (0 < t < End). Direct costs reflect labour and machinery costs, distributed over the growing season as follows: 50% sowing period, 20% crop growth period, and 30% final production stage (Figure 3) (UOOML and PSAL 2009). Similar pattern (42/23/35%) was applied for vineyards and other permanent crops.

Figure 3 - Example of cost distribution for the most common cereal crops in the case study area.


The area affected by 2014 Secchia flood comprises predominately rural areas (43 sq.km), with a prevalent share of arable crops (81% of UUA). The typical crops include cereals, in particular soft wheat and maize (40% of arable crops) and forage for livestock breeding (52% of arable crops). Vineyards and other permanent crops cover the remaining 19% of UAA. As shown in Figure 3, in January maize crops are fallow, while wheat is in its vegetative growth stage. This means that around 50% of cereal production is affected. Permanent crops are affected by 20% of annual production value. The maximum damage to cropland estimated by the formula using this shares is 343 €/ha, less than half compared to the damage yielded by SDC-1 (790 €/ha).

2.4 Gross Value Added model for production losses
Spatially distributed (gridded) economic and social variables are employed to improve the assessment of risk with an evaluation of losses caused by business interruption [Note 4] . Gross Value Added (GVA) can be gridded using detailed land use and population data (Peduzzi et al. 2009; Green et al. 2011). We have used the GVA detailed for the local market system [Note 5] (sistemi locali di lavoro, SLL) level for three macro-economic branches: agriculture, industry and services (ISTAT 2013). Each branch’s share has been proportionally spread over land use categories: agricultural and industrial GVA by UAA and industrial area respectively; services related GVA by population density. We assume that services are diffused but the GVA generated is proportional to the number of residents served in each grid cell. Population density grid has been produced based on the 2011 census tracks (ISTAT 2011). Expected losses as a share of GVA per cell are then calculated using a step function based on the literature on flood damage functions (De Moel and Aerts 2011; De Moel et al. 2012; Jongman et al. 2012; Saint-Geours et al. 2014). We assume that the higher the water level, the more persistent is the productivity loss. This assumption is based on three principles: a) higher water depths cause larger asset damage; b) larger asset damage typically requires longer recovery periods; and c) flood water retreat is a function of flood depth. The relation between water depth and persistence of the impact is likely afflicted by large uncertainty.

Impact on GVAS,L = Σ nk=1  FCS,L k  x ck where FC is the flooded cell (250x250m) k, c is the damage factor applied to each FCk based on its water depth and n is the number of cells belonging to the sector S for each system L.

The losses are calculated for each economic sector as a share of total annual production. The largest share of damage come from the industrial sector, affected for 434 million, equivalent to 14% of its annual production (4.2% of total GVA for SLL Modena, see Table 2). The ratio between asset damage and annual GVA sheds light on the equivalence of structural damage and production losses as a function of the flood characteristics (Figure 4). For water depth of ca. 1 meter, asset damage is equivalent to annual production losses.

Table 2 - modelled impact on GVA from the event of Modena 2014.



Figure 4 - Scatterplot of mean water depth (X) and ratio of SDC damage over exposed GVA (Y).

3. Conclusion
Comparisons of SDC-estimated damage with empirical reported damage are critical for quality assurance of risk assessment results. We have shown that the application of the standard SDC method without proper calibration can overestimate the structural damage by a factor 4 to 13, depending on the land use category. The calibrated maximum damage values for residential buildings are respectively a factor 4 and 4.5 smaller than the original values. The calibration for industrial land use categories was not possible because the damage records are not yet available. The model for agricultural losses provides finer interpretation of damage: using local data on production value and a time-dependent evaluation of costs and losses, the maximum crop damage in our specific example is around half of the SDC-1 damage estimate. Estimated production losses of around 600 million Euros correspond to 5.7 per cent of the GVA of Modena SLL. The analysis described in this paper helps improving the reliability of the flood risk assessment. The implementation of additional spatially disaggregated statistical data such as household income or cadastral value of property can further improve the accuracy of this method.

Notes

[Note 1] Simulated damage: 57/33/10%. Declared damage: 60/35/5%.

[Note 2] The average gross income from the sale of the yield expressed in €/ha, not inclusive of direct costs

[Note 3] The sum of technical means and labour, excluding EU supportive funding

[Note 4] Business activities exposed to flood are partially or completely impaired in their production process.

[Note 5] Local labour systems are territorial continuous areas defined by ISTAT in which most of the daily work activity of resident people takes place. Their scale is in-between NUTS3 and municipality level.

On the morning of January 19th 2014 a 80 meters breach opened on the Secchia river levee spilling 200 thousand litres of water per second in the surrounding countryside up to 10 thousand hectares. Up to seven municipalities have been affected, but the small towns of Bastiglia and Bomporto were the most involved and remained flooded for more than 48 hours. The total amount of water pumped out was estimated at more than 20 million cubic meters (Fotia 2014). The max depth reached by flood hazard is depicted by a raster layer with a resolution of 5 meters produced by University of Parma through hydrological modeling (D’Alpaos et al. 2014). The extent of simulated flood is nearly five thousand hectares, with an average depth of 1 meter.

References

Altamura V, Cesena C, Aldo P, Unibo B (2013) Costi di produzione delle principali specie vegetali in Emilia-Romagna 2013.

D’Alpaos L, Brath A, Fioravante V (2014) Relazione tecnico-scientifica sulle cause del collasso dell ’ argine del fiume Secchia avvenuto il giorno 19 gennaio 2014 presso la frazione San Matteo.

De Moel H, Aerts JCJH (2011) Effect of uncertainty in land use, damage models and inundation depth on flood risk estimates. Nat Hazards 58:407–425.

De Moel H, Asselman NEM, Aerts J (2012) Uncertainty and sensitivity analysis of coastal flood damage estimates in the west of the Netherlands. Nat Hazards Earth Syst Sci 12:1045–1058.

De Moel H, Bouwer LM, Aerts JCJH (2014) Uncertainty and sensitivity of flood risk calculations for a dike ring in the south of the Netherlands. Sci Total Environ 473-474:224–234. doi: 10.1016/j.scitotenv.2013.12.015

Forster S, Kuhlmann B, Lindenschmidt K, Bronstert A (2008) Assessing flood risk for a rural detention area.

Fotia F (2014) Esondazione del Secchia. http://www.meteoweb.eu/2014/01/esondazione-del-secchia-evacuati-20-milioni-di-metri-cubi-dacqua/255270/. Accessed 29 Aug 2014

Genovese E (2006) A methodological approach to land use-based flood damage assessment in urban areas: Prague case study. Eur. Communities, DG-JRC, Ispra, EUR

Green PC, Viavattene C, Thompson P, Green C (2011) CONHAZ Report - Guidance for assessing flood losses.

Huizinga HJ (2007) Flood damage functions for EU member states.

ISTAT (2013) Banca dati Italia.

ISTAT (2011) 15° censimento della populazione e delle abitazioni.

Jongman B, Kreibich H, Apel H, et al (2012) Comparative flood damage model assessment: towards a European approach. Nat Hazards Earth Syst Sci 12:3733–3752.

Koks EE, Bockarjova M, De Moel H, Aerts JCJH (2014) Integrated direct and indirect flood risk modeling: development and sensitivity analysis. Risk Anal.

Luino F, Cirio CG, Biddoccu M, et al (2009) Application of a model to the evaluation of flood damage. Geoinformatica 13:339–353. doi: 10.1007/s10707-008-0070-3

Merz B, Kreibich H, Schwarze R, Thieken A (2010) Assessment of economic flood damage. Nat Hazards Earth Syst Sci 10:1697–1724. doi: 10.5194/nhess-10-1697-2010

Messner F, Penning-rowsell E, Green C, et al (2007) Evaluating flood damages: guidance and recommendations on principles and methods principles and methods.

Molinari D, Aronica GT, Ballio F, et al (2013) Le curve di danno quale strumento a supporto della direttiva alluvioni: criticità dei dati italiani. 10–15.

Peduzzi P, Dao H, Herold C, Mouton F (2009) Assessing global exposure and vulnerability towards natural hazards: the Disaster Risk Index. Nat Hazards Earth Syst Sci 9:1149–1159. doi: 10.5194/nhess-9-1149-2009

Saint-Geours N, Bailly J-S, Grelot F, Lavergne C (2014) Multi-scale spatial sensitivity analysis of a model for economic appraisal of flood risk management policies. Environ Model Softw 60:153–166. doi: http://dx.doi.org/10.1016/j.envsoft.2014.06.012

Scorzini A, Frank E (2015) Flood damage curves?: new insights from the 2010 flood in Veneto , Italy.

Tapia-Silva F-O, Itzerott S, Foerster S, et al (2011) Estimation of flood losses to agricultural crops using remote sensing. Phys Chem Earth, Parts A/B/C 36:253–265.

Thieken a. H, Ackermann V, Elmer F, et al (2009) Methods for the evaluation of direct and indirect flood losses. RIMAX Contrib 4th Int Symp Flood Def 1–10.

Twining S (2014) Impact of 2014 Winter Floods on Agriculture in England.

UOOML, PSAL (2009) Linee guida per la sorveglianza sanitaria e la prevenzione dei rischi per la salute e la sicurezza nel settore cerealicolo.

Ward PJ, Jongman B, Weiland FS, et al (2013) Assessing flood risk at the global scale: model setup, results, and sensitivity. Environ Res Lett 8:044019. doi: 10.1088/1748-9326/8/4/044019

Winsemius HC, Van Beek LPH, Jongman B, et al (2013) A framework for global river flood risk assessments. Hydrol Earth Syst Sci 17:1871–1892. doi: 10.5194/hess-17-1871-2013
 



COMMENTS

 

WRITE COMMENTS

 

  Name:

  Surname:

   Email:

  Comment:






Mattia Amadio, Fondazione Eni Enrico Mattei and Euro-Mediterranean Centre on Climate Change, Italy

Jaroslav Mysiak, Fondazione Eni Enrico Mattei and Euro-Mediterranean Centre on Climate Change, Italy

Lorenzo Carrera, Fondazione Eni Enrico Mattei and Euro-Mediterranean Centre on Climate Change, Italy

Elco Koks, Institute for Environmental Studies (IVM), VU University Amsterdam, The Netherlands
 
 
   
 
Top