Author’s Accepted Manuscript Re-thinking socio-economic impact assessments of disasters: the 2015 flood in Rio Branco, Brazilian Amazon Dorien Irene Dolman, Irving Foster Brown, Liana O. Anderson, Jeroen Frank Warner, Victor Marchezini, George Luiz Perreira Santos www.elsevier.com/locate/ijdr PII: DOI: Reference: S2212-4209(18)30525-9 https://doi.org/10.1016/j.ijdrr.2018.04.024 IJDRR880 To appear in: International Journal of Disaster Risk Reduction Received date: 30 November 2017 Revised date: 23 April 2018 Accepted date: 25 April 2018 Cite this article as: Dorien Irene Dolman, Irving Foster Brown, Liana O. Anderson, Jeroen Frank Warner, Victor Marchezini and George Luiz Perreira Santos, Re-thinking socio-economic impact assessments of disasters: the 2015 flood in Rio Branco, Brazilian Amazon, International Journal of Disaster Risk Reduction, https://doi.org/10.1016/j.ijdrr.2018.04.024 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Re-thinking socio-economic impact assessments of disasters: the 2015 flood in Rio Branco, Brazilian Amazon Dorien Irene Dolman1, Irving Foster Brown2,3, Liana O. Anderson4,5, Jeroen Frank Warner , Victor Marchezini5, George Luiz Perreira Santos6 1 Corresponding author: Dorien Dolman, doriendolman@gmail.com, +31 (0)6 52673757 1 Social Sciences Group, Wageningen University and Research, Hollandseweg 1, 6706 KN, Wageningen, the Netherlands 2 Woods Hole Research Center, 149 Woods Hole Road, Falmouth, MA 02540-1644, USA 3 Federal University of Acre –UFAC, Campus Universitário - BR 364, Km 04 - Distrito industrial - Rio Branco – Acre, Brazil, CEP: 69.920-900 4 School of Geography and the Environment, University of Oxford, Oxford, UK 5 National Centre for Monitoring and Early Warning of Natural Disasters - Cemaden São Jose dos Campos, São Paulo, Brazil 6 Municipal Civil Defense of Rio Branco, Rua Rui Barbosa, 285 – Centro, Rio Branco, Acre, Brazil, CEP: 69.900-901 Available ORCID numbers of authors: L.O.A. – ORCID number: 0000-0001-9545-5136 D.I.D – ORCID number : 0000-0002-3362-7163 V.M – ORCID number: 0000-0002-1974-0960 I.F.B. ORCID number: 0000-0003-1877-0866 Abstract The impact of water-related disasters has become more acute in cities of the Amazon Basin. Socio-economic impact assessments have a key role in improving sustainable mitigation projects in order to increase resilience and reduce societal vulnerability. This paper reviews the current state of loss assessments and then explores how to improve estimates for the 2015 flood affecting the city of Rio Branco, Brazil, located on the headwaters of the Amazon Basin. Prevailing models, loss assessments, and databases are not applicable in this Amazonian context due to the lack of detailed cost administration, low levels of human and financial capital, and limited insurance coverage. This paper uses uncertainty ranges of the costs of water-related disasters to provide an assessment of the total impact. Our estimate ranges from 60 to 200 million USD of losses and damage solely due to this flood event, compared to the official estimate of 98 million USD. As floods in Rio Branco are recurrent nearly annually, the cumulative losses over the years may be significantly higher. Our study illustrates the need for improving impact assessments in order to increase the knowledge on the actual costs of flood disasters and avoiding silent impoverishment. Outcomes of impact assessments can show the necessity of mitigation activities which will reduce vulnerability of societies. Keywords: floods, socio-economic impact assessments, uncertainty, Amazon, Brazil 1 1. Introduction Over the past decade more than 1.5 billion people have been affected by major and reported disasters (United Nations, 2015, EM-DAT, 2016). The impact of floods has grown over time: total worldwide economic damage of floods in the last decade is 228 times higher than that reported for 1955-1965 (EM-DAT, 2016). Economic impact consists of losses of and damages to belongings, agriculture, income, services, and buildings. Social impacts include aspects such as increased disease morbidity, violence, traumatic memories, as well as non-monetary recovery efforts. This trend is expected to continue. Climate change, social vulnerability population growth and land-use change are likely to worsen the socioeconomic and environmental impact of natural hazards (Helmer and Hilhorst, 2006; Raddatz, 2009; Poussin et al., 2013; Koks and Thissen, 2015; Raddatz, 2009; De Groeve et al., 2015). Sustainable strategies of disaster mitigation require an in-depth understanding of the dynamics of disasters (Koks and Thissen, 2015; United Nations, 2015). This can be improved by conducting socio-economic impact assessments in the aftermath of disasters (Green et al., 2011). Loss assessments can reveal weaknesses and identify opportunities for reform (De Groeve et al., 2014), especially in developing countries where vulnerabilities are higher and institutional capacities need to be strengthened. Although the DesInventar database offers a wider inclusion criterion for disaster types – such as sub-national or small-scale disasters – and data categories that go beyond mortality and economic losses, as done for El Salvador’s 2001 earthquake (Muñoz et al, 2017), the coverage of losses in diverse sectors still remains patchy and arbitrary in its scope (Zaidi, 2018). This patchiness includes the indirect disaster impacts in the short and long-term recovery periods. In Brazil, for instance, the lack of loss estimation data limits understanding as to how the different sectors are affected during and in the long-term disaster recovery. This article doesn´t address all of these gaps, but provides some insights and findings to improve loss estimation and to reduce institutional vulnerability, taking Rio Branco city in Amazonian Brazil, as a case study Much improvement in the data collection could be made during and after floods. This is especially relevant in contexts of uncertainty, with limited administration by institutions, lack of human and financial resources, and no developed insurance system – in other words, in contexts of institutional vulnerability where formal institutions (constitutions, regulations, bureaucracy, rule of law, codes) and informal institutions (culture, traditions, norms) are too weak to provide disaster risk management measures (Lassa, 2010). To remedy (to a degree) such shortcomings, the present contribution presents a method to calculate the impact of natural hazards which could contribute to an increased understanding of the relevance of data collection in the immediate aftermath of disasters. We developed this method in the context of a flood-prone city in Amazonian Brazil. In Brazil the estimated average of annual Brazilian economic losses due to the four major floods between 2008 and 2011 was 1.4 billion USD (World Bank, 2014). Loss assessments are officially based on rapid assessments during the natural hazard event (World Bank, 2014) as a precursor for financial aid. It is not possible to fully quantify the total socio-economic impacts as these rapid assessments are by their very nature preliminary. This is a limitation that many developing countries face due to the lack of human, financial and infrastructure resources to conduct detailed 2 post-hoc analysis. In addition, other drivers contribute to the increase of the social and economic damages. Although the Brazilian Human Development Index (HDI) is above average at 0.754 (no. 79 in the world1), a high degree of social inequality in vulnerability still exists. The Brazilian Disaster Risk Indicators (DRIB-Index) and Social Vulnerability Index (SOVI) also show that the highest vulnerability urban centers are located in the North (Amazon) and Northeast Regions (Almeida, Welle and Birkmann, 2016; Hummell, Cutter and Emrich, 2016). These vulnerable regions are exposed to recurrent floods, droughts and forest fires, which are predicted to intensify in the future (Malhi et al., 2009). Rio Branco, the capital of the state of Acre in Brazil´s southwestern Amazon (Figure 1) is an example of an area experiencing extreme events with increased frequency (Araujo Lambert, 2015, Espinoza et al., 2014; Ovando et al., 2016). Since 1988, the city has been flooded, operationally defined as river levels exceeding 14.0 meters, in the years 1988, 1997, 2006, 2009, 2010, 2011, 2012, 2013, 2014, and 2015 (Rio Branco Civil Defense, 2016). The 2015 flood was the most severe in the recent history of the city. During 2015, the Acre river, which bisects the city of Rio Branco, was above flood stage for 32 consecutive days and reached a level of 18.4 meters on 4 March 2015 (Civil Defense of Rio Branco, personal communication, June 5th, 2016). On the 5th of March, the National Civil Defense declared a State of Public Calamity. The Civil Defense of Rio Branco estimated that the flood affected 100,000 persons, about a third of the city´s population. In the aftermath of this long-term flood of 32 consecutive days no official socio-economic impact was conducted. Although hazard maps were drawn up in many municipalities after the 2015 flood, there is a dearth of data regarding municipal vulnerability assessment. Moreover, in the aftermath of disasters, the voices of affected people are not heard, and their losses and damages are missing in official loss estimation carried out by national governments (Marchezini, 2015). Without these data is not possible to analyze the spatial patterns of losses in affected areas, as is done in the DesInventar database. Figure 1: Location of the city of Rio Branco Acre State, Brazil within the Amazon Basin. The dark line depicts the Acre river and the shaded area is the Acre River Basin. Courtesy of Sonaira Souza. 1 http://hdr.undp.org/en/countries/profiles/BRA 3 This paper explores challenges to socio-economic impact assessments of long-term floods in developing countries, focusing on the institutional context in Brazil, one of the largest and most diverse developing countries. First, we present the current knowledge on disaster models, assessments, and databases and identify gaps. Second, we review the state of the art of impact assessments and the implications for the case study in Rio Branco. Third, we present the outcomes of the multisector socio-economic impact assessment of the 2015 long-term flood. Finally, we discuss future low-cost improvements to these assessments. 2. Socio-economic impact assessments 2.1 Models, assessments, and databases Decision makers require accurate data on disaster losses of previous events in order to take effective actions about research priorities, planning decisions, disaster assistance, reconstruction activities, and policy evaluations (Downton and Pielke, 2005). We identified three key variables associated with floods and disasters to be improved in order to increase accuracy of impacts and losses quantification: flood models, loss assessments and databases on water-related disasters impacts. Flood damage estimates via mathematical models represent a significant amount of literature (Egorova et al., 2008; FEMA, 2011; Koks and Thissen, 2015). These models are useful for strategic decisions on flood mitigation, as they provide the estimated impacted area, and are aimed at disaster risk reduction (FEMA, 2011). An example of a widely-used flood model is HAZUS-RM developed by the Federal Emergency Management Agency (FEMA), an agency of the United States Department of Homeland Security (FEMA, 2011). This model uses census information to estimate potential damage on public facilities, transportation, and buildings. Current literature analyzes the reasons for inaccuracy (Handmer, 2013), describes the types of costs (Meyer et al., 2013), and provides examples of socio-economic impact assessments of floods (Chatterton et al., 2007). In particular, the inaccuracy of flood loss assessments can be explained by the complexity of loss assessments, variations in approaches, lack of knowledge on data, divergent interests of stakeholders, variation in funds, expertise, and time available (Handmer, 2013). The CONHAZ program (Costs of Natural Hazards, Green et al., 2011) has developed a terminology of cost types: direct, business interruption, indirect, intangible, and risk mitigation. The terminology is useful as each type of cost requires a different assessment. One example of a flood loss assessment, Chatterton et al. (2007), documents well which aspects can be included in loss assessments in other contexts. Over the past decades the number of databases on disaster losses has increased. According to De Groeve et al. (2014), fifteen of the twenty EU member states which participated in their research have a national disaster database. Several international databases on disaster losses have been created such as EM-DAT (Emergency Events Database), NATHAN (Natural Hazards Assessments Network), SHELDUS (Storm Event Database) (Gall et al., 2009) and the DesInventar (Disaster Information Management System). The database which contains information on disasters in Brazil is the Integrated System on Information of Disasters (S2ID) which was created in 2012 (Federal University of Santa Catharina - Brazil, 2012). 4 2.2 Gaps in current literature Significant improvements have been made in disaster models, assessments, and databases but challenges remain (Meyer et al., 2013). We however identified the following aspects as major challenges: (1) model application, (2) assessments made in a context in which detailed administration and financing is lacking, and (3) discrepancies in disaster loss databases. It is important to overcome these barriers in order to improve mitigation projects for future disasters, such as floods (Merz et al., 2010). Many of these factors can be seen in the following analysis of the 2015 flood in Rio Branco, Acre, Brazil. Models such as HAZUS are useful to predict disaster impacts but cannot replace loss assessments conducted in the aftermath of disasters (FEMA, 2011; Koks and Thissen, 2015). In practice size and impact of disasters often differ from model predictions. Flood models face technical implications due to their inflexibility and dependency on high quality data as input, and organizational implications due to difficulties of exchanging results with decision makers (Leskens, 2014). Hence, in order to verify the accuracy of disaster models as a suitable tool for decision makers, it is important to evaluate and calibrate flood models by comparing these with actual impacts calculated with a post-disaster assessment. Loss assessments often focus on contexts which consist of detailed administration of costs by local and national authorities, sufficient human and financial capital, and a developed insurance system, such as Chatterton et al. (2007) observed for Great Britain. Such methods are not applicable in the context of Brazil where the previous mentioned aspects are not available for loss assessments (Szlafsztein, 2012). Researchers emphasize the need for research, procedures, and guidelines which define a systematic and standardized data collection on flood losses. Those will contribute to reduction of inaccuracy in estimations of impacts of water-related disasters (Elber et al., 2010; De Groeve et al., 2014; United Nations, 2015; Gall et al., 2009; Meyer et al., 2012). Databases on disasters are sometimes inconsistent as they lack data, standardized methods and explication of methods for data collection (Gall et al., 2009; Meyer et al., 2009). The accuracy of data of EM-DAT is questionable because the website lacks the following information: aspects included, methods of calculation, and sources. This lack of meta-data makes verification of the data impossible (Gall et al., 2009). The database on all disasters in Brazil on the website “System of Information on Disasters” (S2ID) is based on “Forms of Information of Disasters” (FIDE). FIDE is a tool of the Brazilian Civil Defense through which local branches apply to request funds from the Federal Civil Defense. While it is an appropriate tool for data collection in the midst of disasters, the reliability of this data is not very high as it is collected in a chaotic situation. In addition, the evaluations teams of civil defense are not well trained to apply it without subjectivities and the deadline for data collection is only 10 days after the State of Public Calamity was confirmed (National Civil Defense). Therefore, the data presented in FIDE contains several rough estimations and is incomplete. Hence, the accuracy of the S2ID database is limited (World Bank, 2014). As a consequence, extrapolations of disaster losses have considerable error embedded in them, especially in longterm flood events. 3. Methodology 5 3.1 Data collection Field data collection for this study occurred in Rio Branco between May and July 2016 a year after the flood. In this period 29 semi-structured interviews were conducted with 18 employees of 14 municipal and two state governmental departments and 11 with families that represents different vulnerable neighborhoods affected by the flood of 2015. Listening to the voices of people in the frontline helps us better understand their losses and should be an important step in planning community-based mitigation strategies. The sample of 11 families is small but it provided an estimate of the loss of belongings, given that the types of their different personal and domestic assets (stove, refrigerator) are not measured in FIDE. The sample included respondents of different: backgrounds, financial situations, and of five different neighborhoods of Rio Branco. When available, reports from governmental departments were used (see Supplemental Online Material-SOM, Table S-1). The governmental costs were estimated by circulating a questionnaire to governmental departments of Rio Branco and Acre states. Of the sixteen local authorities involved in the response to the flood nine filled out the questionnaire (SOM). As the response was not complete, some costs were calculated with the method of uncertainty ranges, described below. Based on the costs acquired through the semi-structured interviews and the questionnaire we estimated the total loss. 3.2 Uncertainty analysis The calculation of uncertainty has the potential to generate a range of estimates of costs of disasters. It provides a means for decision-makers to decide whether more resources are merited to reduce the ranges of cost uncertainty. The case study of Rio Branco was carried out in a context of limited administration by institutions, lack of human and financial resources, and no developed insurance system. Therefore, methods such as the one described by CONHAZ (Meyer et al., 2013) and Chatterton et al. (2007) are not applicable. In many calculations of costs, the final value is the product of two or more variables. For example, if the average loss per person is 100 US dollars then for an affected community of 1,000 persons, the cost would be 100,000 USD (100 USD/person x 1,000 persons). But both variables are subject to uncertainty which is propagated to the final product. There are several ways of dealing with this uncertainty, from Monte Carlo simulation to a percent of the estimated value.. In the case of Rio Branco, some estimations (Tables 1 and S.2) were calculated using the Form of Information on Disasters (FIDE). FIDE is based on rapid assessments executed by different municipal departments during the disasters. In the case of Rio Branco experts of the local government carried out the data collection during the 32 days of flooding (Civil Defense of Rio Branco, personal communication, June 5, 2016). As FIDE is a rapid assessment, conducted in a chaotic situation during floods, it generally underestimates the total cost. For example, in the FIDE of the 2015 floods in Rio Branco the number of people affected was estimated to be 71,000 (National Civil Defense, 2015) while in the aftermath this number turned out to be approximately 100,000 according to the Municipal Civil Defense of Rio Branco, a difference of over 30%. Nonetheless, for some variables impacted by the flood FIDE represents the only data available. This resulted in a significant underestimate of the losses and costs of the long-term flood event. 6 The FIDE value served as an ´anchor´ (Morgan and Henrion, 1990) and de facto became the best estimate of the cost. As a result, we were systematically underestimating the damages associated with the flooding initially, creating “silent” impoverishment of the state, as many of the damages are not reported. In order to emphasize the range of possible values, we eschewed the use of a single value and generated maxima and minima, using a simple percentage as a propagation tool. The application of uncertainty ranges is described in literature such as the costs of the 2007 floods in England (Chatterton et al., 2007). Chatteron et al. apply for their estimates a degree of confidence as presented in Penning-Rowsell et al.’s (2005) Multi-Coloured Manual. Penning-Roswell et al. divide uncertainties into the following categories of confidence: best breed, limiting assumptions, gross assumptions, and heroic assumptions (Chatterton et al., 2007; Penning-Rowsell, 2005). “Best breed” are assumptions for which no better data is available and hence an uncertainty of +/- 10% is applied. Limiting assumptions are those which have known deficiencies and are replaced as soon as better data is available. Hence, an uncertainty of +/- 20% is applied. Gross assumptions are data which are not an outcome of the respective researchers but deducted from other sources and therefore a range of 25-35% uncertainty is used (Chatterton et al., 2007; Penning-Rowsell, 2005). For heroic assumptions, Chatterton et al., do not present a specified uncertainty range. Penning-Rowsell describes heroic assumptions as: “No data sources available or yet found; data based on educated guesses” Penning-Rowsell, Table 3.3 on p.18). In the case of Rio Branco uncertainties are mostly considered heroic assumptions (Eq. 1, 2 and 3) for aggregating costs. We developed the following equation to calculate uncertainties. (Eq.1) ( ( ) ) ( ( ) ) (Eq. 2) (Eq. 3) Where is the aggregated cost, is the Minimum aggregated cost, is the Maximum aggregated cost, is the cost per unit and is the total number of units. With a use of two variables and 50% uncertainties, there is a nine-fold range (2.25/0.25) between maximum and minimum estimates, nearly an order of magnitude. 3.3 Data Analysis 3.3.1 Impact on families affected by the 2015 flood The losses were assessed based on information provided by respondents in the semistructured interviews and by the calculation of uncertainty ranges. Families suffered from the following losses: belongings, agriculture, income, decrease in services and damages to buildings. The methods to estimate these losses are described and summarized in Table S-2. First, we identified the typology of basic belongings that families have and estimated their loss. In FIDE only the damage to housing was included, not the losses of belongings. In England, it is relatively easy to calculate losses of belongings as people claim these losses with insurance companies (Chatterton et al., 2007). In Brazil, few people have home insurance, and even if they have one, it would not cover losses from “natural” disasters as this specific condition is not well developed in the country. Hence, the method proposed by Chatterton et al. (2007) is not applicable. The average loss per family was estimated based on eleven semi-structured 7 interviews in which respondents of different neighborhoods were asked which belongings they lost. This sample was used to identify the typology of basic belongings lost, but we know that point out the diversity of these items are difficult, as well as to quantify how useful they are in the daily activities of each family, especially during disaster recovery phase when other types of social suffering usually arise. Respondents stated they lost micro-waves, tables, wardrobes, clothes, gas stoves, etc. By estimation of the average value of these belongings the loss of each family of the sample was calculated. The average loss per family was multiplied with the estimated amount of 20,000 affected families (Table 1, Table S.2). Second, we explored the agricultural losses as the flood destroyed many crops. In this case, there is no range of uncertainty as the Municipal Department of Agriculture and Forests conducted an extensive survey in the aftermath of the flood. All information on losses of agriculture was collected from the official report published of the survey, no range of uncertainty is applied as it was considered to have 100% coverage.. This respective department executed an extensive recovery program with which they supported the affected the families. Hence, the amount of public funds spent on recovery was deducted from the total agricultural loss. Third, we estimated the income loss as many affected people could not work during the 32 days of overflow of the Rio Acre. These data are not included in FIDE and hence is an innovative contribution of our method. This absence is due to various reasons: inaccessibility of roads, necessity to watch over their houses, flooded workplaces, temporary stay at family distant from work, illness, etc. We considered if one of the models or assessments described above would be applicable to calculate income loss. For example, the HAZUS-MH MR model estimates wage losses based on the Social Accounting Matrix, which represents flows of all economic transactions that take place within an economy (FEMA, 2011). A Social Accounting Matrix for Brazil exists but it does not include incomes of the informal sector. This sector is quite substantial in Brazil; according to Henley et al.(2009), it ranges from 40 to 63%. As there is a large informal sector in Rio Branco, the Social Accounting Matrix is not representative for all incomes and hence HAZUS-RM is not applicable. Our study was based on the assumption that 100,000 people were affected and that approximately one fifth is employed and has an income of 243 USD2 per month. For this article the exchange rate applied is 1 USD equals 3.14 Brazilian Reals. Hence, the average wage was multiplied with one fifth of the affected people (Table 1, Table S.2). Fourth, we estimated the loss of services as this affected families as either consumers or employees. Offices, shops, and warehouses were flooded which influenced the sale of products and services. Meyer et al. (2013) define losses of services as business interruption costs. Methods described in the CONHAZ report to calculate the losses of services are not applicable in Rio Branco due to lack of information on added value of all products, history of product outputs or direct damages of all services affected. Therefore, we estimated the losses of services, with the data in FIDE as a point of departure. In this report, we estimated that the damageto services amounted to 20.8 million USD in case 71,000 persons had been affected which was raised proportionally for the 100,000 persons. 2 According to the 2010 population census the average nominal urban income per capita per month in the state Acre was 765,49 Reals. This equals 243 USD. 8 Fifth, we estimated the damage to buildings as approximately 29,300 buildings were affected by the flood according to the State Civil Defense of Acre (Coronel Batista, personal communication, June 7th, 2016). According to the Civil Municipal Defense of Rio Branco, few houses were destroyed completely but many houses had partial losses and families had high costs of reconstruction. The HAZUS-RM model estimates the losses of housing by the average loss and weighing it with the area of inundation (FEMA, 2011). Unfortunately, such a map is absent in Rio Branco. Therefore, the uncertainty range of damages to buildings was calculated based on the data in FIDE (Table, S.2). In this document the damages to buildings were estimated to be 4.5 million USD if 71,000 persons had been affected. 3.3.2 Governmental costs Municipal departments of Rio Branco and state departments of Acre incurred high costs in providing response to the 2015 flood. Municipal and state governmental departments supported the Civil Defense in evacuation of people, construction of temporary shelters, distribution of food, provision of healthcare, and cleaning of the city. Governmental expenditures are divided by the governmental body paying the expenses: (1) National Civil Defense, (2) federal ministries, and (3) resources of municipal and state departments. The funds liberated by the National Civil Defense were taken from data of the Financial Department of Rio Branco. The funds liberated to the Civil Defense of Acre were collected in a semi-structured interview with their treasurer. These are reported actual costs and therefore no degree of uncertainty was applied. The federal ministries of health and education liberated funds for recovery projects of the municipal and state departments. The funds received by the municipal departments were collected during semi-structured interviews. The funds received by the Educational Department of Acre were estimated with the methods of calculation of uncertainty ranges based on the data of FIDE. In FIDE it was estimated that the recovery of twelvedamaged schools would cost 3.8 million USD. The expenditures which were not reimbursed by any funds were paid with own resources of the respective departments of Rio Branco and Acre. These data are not included in FIDE and hence are an innovative contribution of our method. Although requested multiple times not all governmental departments reported their costs and therefore an uncertainty range was calculated for the lacking information. The expenditures of the nine departments which filled the questionnaire were summed,of this amount the average was calculated. Hence, the average was multiplied by seven to estimate the own resources spend by the departments which did not fill the questionnaire (Table 1, Table S.3). An innovative aspect of this research was the calculation of opportunity costs of the departments of Rio Branco and Acre. These data are included in our method and not in FIDE. Opportunity costs, or alternative choices, are defined as: “The evaluation placed on the most highly valued option of the rejected alternatives or opportunities” (Buchanan, 1991, p.520). The opportunity costs of disasters are numerous as schools are closed, institutions function slower, and employees were diverted from their daily tasks. In prevailing models, assessments, and databases on disasters little attention is devoted to opportunity costs (Chatterton et al., 2007; Meyer et al., 2013). Hence, in this study the opportunity costs of employees of municipal and state governmental departments, who provided relief during the flood, were calculated. The opportunity costs of the nine departments which filled the questionnaire were calculated by 9 multiplying average wage with the average days involved and the number of employees. The opportunity costs of the departments which did not fill the questionnaire were estimated with the averages of wages, number of workers, and days involved of the departments which filled the questionnaire (Table S.3). 4. Results and discussion 4.1 Economic impact on families in Rio Branco due to the 2015 flood The 2015 flood was the most severe in the recorded history of Rio Branco. Local authorities constructed 29 temporary shelters which accommodated 11,000 people for 32 days, which is approximately 11% of the affected population. This is a sub estimate of those displaced as many sought shelter with relatives and friends or remained in their houses in order to protect their belongings from robbery. The flood affected 53 neighborhoods of Rio Branco, 165 kilometers of roads, and 29,300 buildings (SitGeo, 2016; Colonel Batista, Civil Defense of Acre, personal communication, June 8th 2016). Many people were affected as they still resided in zones at risk due to socio-environmental reasons. Of the affected people, three pathways of relocation were identified by the department of Social Assistance of Rio Branco (D. Araujo, personal communication, June 7th 2016): (1) relocation to public shelters, (2) relocation to reside temporarily with family members outside the flood zone, and (3) remaining in their houses in order to protect their belongings from robbery. The most important impacts of the flood on affected families were: losses, diseases, violence, traumatic memories, and difficulties to recover. The losses of families are divided in: belongings, agriculture, income, services, and damages to buildings (Table 1). Families lost belongings during the flood either during the transport from their houses or as these remained in their houses during the flood. Many residents did not leave their houses before these flooded because (1) some did not believe the early warnings, (2) others did not want to leave, or (3) did not receive logistic support. The neighborhoods affected by the flood were among the poorest of Rio Branco and hence the impact of losses was relatively greater. Our results suggest that the average loss per family was 3,800 Reals, approximately 1,172 USD, which revealed part of their social economic vulnerability once their income is 243 USD per month. With the method of calculation of uncertainty ranges we estimate that the uncertainty range of the total losses of belongings of families: between 6 and 56 million USD. The 2015 flood also affected 1,500 families of small farmers. The actual agricultural losses were approximately 9 million USD according to the report provided by the Municipal Department of Agriculture and Forests (SEAPROF, 2015). From these costs the amount of the recovery program was deducted which result in a total agricultural loss of 4.3 million USD. This study estimated that between 1.6 and 14.1 million USD were associated with incomes loss during the flood event. Moreover, our estimates related to the flood impact on the services reveals that between 14.4 and 43 million USD were lost. Finally we estimated that the total cost for families for reconstruction of damages to their homes is between 3.1 and 9.4 USD. The total economic loss of families affected by the 2015 loss is in the range of 29.7 and 127.7 million USD. 10 Table 1: Uncertainty ranges of total losses of families affected by the flood of 2015 in Rio Branco. Losses of families (20,000 families) Belongings Agriculture Income Services Damages to housing Total Minimum loss in million USD 6.0 4.3 1.6 14.4 3.1 29.4 Average loss Maximum loss in million in million USD USD 34.2 56.4 4.3 4.3 7.9 14.1 28.7 43.0 6.3 9.4 78.7 127.7 4.2 Social impacts In addition to the economic losses, floods have social impacts. The most severe social impacts were on health, security, memories, and subsequent well-being. Interviewees reported an increase in illnesses, anxiety about violence and traumatic memories. Also, the recovery after the flood was difficult. Social impacts are considered intangible (Meyer et al., 2013) as no monetary value is attributed to them Health of inhabitants of zones at risk was impacted and many people suffered from diseases. Unfortunately, quantitative data on most health impacts are lacking which limits assessments. Data which are available for Rio Branco are the number of diagnoses done in the temporary shelters. According to O. Almeida (personal communication, June 9th, 2016, department of Health of Rio Branco), more than 3,000 disease diagnoses occurred in the temporary shelters of which 20% were infections of the upper respiratory tracts. Anxiety about violence increased during the flood of Rio Branco. In the semi-structured interviews affected people mentioned they were afraid that their belongings would be robbed during their absence. This fear was based on stories of previous floods of houses being robbed. These robberies occurred during the stay in temporary shelters or with family members. Although there was an increased police surveillance during the flood, some people organized themselves in groups within their neighborhood to guard their houses. Others decided to remain in their flooded houses in order to prevent robbery. Unfortunately, there is no specific data on the robberies and crime rates in Rio Branco during the flood. Data is available on the number of crimes in the largest temporary shelter, “Parque do Exposição”. The chief of the military police of Rio Branco, (C. Maciel, personal communication, June 20th, 2016), stated that in the largest shelter 96 crimes were registered such as drug trafficking, threats, domestic violence, and robbery. Current data available on violence in the temporary shelters during the 2015 floods does not demonstrate an increase as there is no data available to compare it with. Those crime rates and the anxiety expressed by respondents of the semi-structured interviews does justify the necessity to explore more on this topic. Several studies confirm an increase of child abuse, sexual violence, and other interpersonal violence during disasters (Rezaeian, 2013). Therefore it is important to include the impact of violence in socio-economic impact assessments. 11 People affected by floods also suffer from traumatic memories. Respondents expressed sentiments such as fear and stress resulted from the chaotic disruption of their private and public routines, the loss of home and usual family roles, and the moral dimension of labor. They felt stressed because they faced losses, diseases, and anxiety of violence. Emotions which were expressed by respondents are: desperation, helplessness, fear, and sadness. Moreover, some people described traumatic memories of encounters with wild animals, such as snakes and alligators close to their houses. Traumatic memories have an important impact on lives of affected people and pose several challenges to the “resilience imperative” in contexts of social abandonment (Marchezini, 2015). Finally, we explored how the affected families experienced the recovery after the flood. D. Araujo, department of Social Assistance (personal communication, June 7th 2016), explained how returning to homes after the flood was accompanied by traumatic memories for some residents due to losses of belongings and damages to their houses. The cleaning of houses was laborious and some people stated that a bad smell remained for more than three weeks. According to K.S. Carvalho (personal communication, June 16th, 2016), chief of the waste collection department, in the weeks after the flood 18,000 tons of trash were collected in the impacted neighborhoods, which is equal to 85 days of normal trash collection. About 55% of the respondents stated they felt not being fully recovered more than a year after the flood. Asked what recovery meant to them, they stated that they still did not possess the financial resources to purchase the assets they lost during the flood. The total loss for affected families was between 30 and 128 million USD. 4.3 Governmental costs for the flood of 2015 The total governmental costs are presented in Table 2. The National Civil Defense transferred 3.6 million USD to municipal and state Civil Defense agencies to help the provision of relief during the flood. The funds were mainly used for construction of temporary shelters, provision of meals in the shelters, fuel, chemical toilets, and reconstruction activities. The federal ministries of education and health supported the respective departments in Rio Branco and Acre in the reconstruction of damaged schools and healthcare centers. The Rio Branco Health Department stated they received 1.6 million USD, and the education department 0.3 million USD. The expenditures of the Acre Education Department are estimated as these were not collected during the fieldwork. The total funds liberated by federal ministries are between 3.7 and 7.2 million USD. Therefore, about 5% to 25% of the total governmental costs were covered by federal ministries. Table 2: Total governmental costs to the flood in Rio Branco, 2015 Costs National Civil Defense Federal Ministries Governmental departments of Acre and Rio Branco: own resources Governmental departments of Acre Minimum costs in million USD 3.6 3.4 14.2 4.5 Average costs in million USD 3.6 5.3 21.9 13.8 Maximum costs in million USD 3.6 7.2 29.7 23 12 and Rio Branco: opportunity costs Total 25.7 44.6 63.5 The governmental costs of the flood of 2015 in Rio Branco were between 25.7 and 63.5 million USD. In comparison, the total annual budget of the municipality of Rio Branco in 2015 was 250 million USD and the GDP per capita of Rio Branco in 2015 was 7,125 USD (IBGE, 2015). 5. Discussion: re-thinking socio-economic impact assessments The uncertainty calculation method of disaster impacts shows that the total costs of the 2015 flood in Rio Branco were between 60 to 200 million USD. The estimation of the losses done with the FIDE is within this range; 97.9 million USD – about 40 million USD above our lower estimate and about 100 million USD below our highest estimate. Hence, the estimation of FIDE is not in the middle of the uncertainty range but rather closer to our lower estimate. As the floods in Rio Branco are recurrent the cumulative losses over the years must be even higher. Our method is useful in contexts of absence of: (1) detailed administration of costs by local and national authorities, (2) sufficient human and financial capital, and (3) a developed insurance system. The goal of this method is to raise awareness of local authorities on the need of improvement of socio-economic impact assessments. If authorities would feel that the range is too wide for directing policy, then these ranges serve to motivate local and state governments to collect more accurate data in the future. Moreover presentation of our highest estimates could make governmental actors realize the financial benefits of investing in mitigation projects rather than focusing on disaster relief. . This would benefit the inhabitants of Rio Branco as governmental actors might implement projects to mitigate suffering of the population. Based on the presentation of our preliminary results governmental actors started to discuss what kind of mitigation activities would be most economically viable in order to reduce vulnerability of the population of Rio Branco. Our method of calculation of uncertainty ranges is an appropriate complement to the FIDE method. FIDE is useful to collect data in the midst of an emergency whereas our method is more inclusive. In addition to the costs included in FIDE our method includes: losses of belongings, losses of income, governmental expenditures and opportunity costs. However, several aspects of this method should be considered. First, better estimates would reduce the range of uncertainty. Due to the lack of data in calculations of the costs of the flood in Rio Branco, we applied 50%, which led to a three-fold range for one variable and a nine-fold range for two multiplied variables. Not to use heroic assumptions, as described by Chatterton et al. (2007) would have meant systematically underestimating damages which would lead to unacknowledged “silent” impoverishment. Second, the method of calculation of uncertainty ranges should be used with the aim of increasing awareness. Presentation of uncertainty ranges is useful to enhance awareness of the need to improve socio-economic assessments which will contribute to a decrease of uncertainty in the future. Instead of presenting estimates as exact data, as done in FIDE, our method stresses the necessity to have more accurate data by showing an uncertainty range. Our results indicated that local authorities can improve assessments by: (1) carrying out post-disaster surveys, (2) improve data storage and dissemination, and (3) better administration of own resources. Surveys 13 can be conducted during and in the aftermath of a disaster to gather data of the socio-economic impact on affected families (Handmer, 2013). These surveys are useful to obtain information on losses (belongings, agriculture, income, losses of services, and damage to buildings), traumatic memories, and difficulties to recover. Contrary to the current method of FIDE, socio-economic impacts data should be gathered in the aftermath of a disaster and with involvement of the affected people. Once the situation is stable, it is recommended to start the surveys as experiences are still vivid in people’s memories. Considering the lack of human and financial resources we recommend selecting a sample size that reduces uncertainty to an acceptable level for policy decisions. Authorities can collect and share data on intangible, or social, impacts with the local Civil Defense. The health department can gather data of the increase in diseases in the entire city instead of only in the temporary shelters. The security department can collect information on the increased amount of violence as a result of floods. Municipal and state governmental departments can improve the administration of their own resources, both general expenditures and opportunity costs. Execution of surveys, data collection on social impacts, and improved administration would lead to a decreased uncertainty of the socio-economic impact. Third, current literature on impact assessments elaborates on economic losses but rarely discusses valorization of social impacts of disasters of diseases, violence, traumatic memories, and difficulties to recover (Chatterton et al., 2007; Meyer et al., Green et al., 2011; Elmer et al., 2010). It is important to include social impacts during floods as these are often underestimated. In Rio Branco after the 2015 floods, we found these to be serious and durable. Improved knowledge on social impacts could contribute to more adequate projects of local governments to respond to these impacts in the long-term disaster recovery process. Research can contribute to improving socio-economic impact assessments, particularly in a context of a lack of detailed administration, insurance system, and human and financial resources. Furthermore, more research is needed on assessment of impact of intangible costs of floods. Socio-economic impact assessments should include impacts on health, violence, traumatic memories, and difficulties to recover. Scholars should explore how to devote more attention to social impacts in disaster models, assessments, and databases. Perhaps citizen science approaches (Marchezini et al, 2017) can contribute to build a strategy for collecting crowdsourcing data about social impacts, especially during longduration floods. Researchers from different fields of knowledge and policymakers need to discuss together what to do to face the challenges from the frontline of the disasters. Fourth, a form should be designed with the participation of civil society and academia in order to be applied to assess the multidimensional socio-economic impacts in the aftermath of disasters in Brazil. The entire database (S2ID) of disasters in Brazil is based on the FIDE forms which are filled out during the emergency phase, and doesn’t capture the impacts of a disaster. Hence, a form should be developed to assess the impact of disasters in order to improve data collection of disasters in Brazil. One way to facilitate this would be a legal requirement for a post-disaster evaluation for those municipalities and states that received federal funding. Training should be provided to local authorities on how to conduct surveys, which data to collect, and how to administrate their own resources effectively. The development of a standardized form of cost and training on its use will enhance the capacity to execute effective socio-economic impact assessments. 14 6. Conclusion This paper presented a comprehensive assessment of socio-economic costs of the 2015 flood in Rio Branco, in Brazil´s western Amazon. This type of analysis is currently lacking in the literature and high uncertainties of the actual cost of climatic extremes, such as floods, are not fully quantified. The flood of 2015 had a severe impact on affected families in Rio Branco: losses, diseases, anxiety of increased violence, traumatic memories, and difficulties of recovery. The method of calculation of uncertainty ranges of the impact of the floods shows that the total costs of the flood in Rio Branco are between 60 and 200 million USD. In addition, the method of calculation of uncertainty presented this study provides a range of costs that could be used for the local government in order to assess the pros and cons of mitigation strategies for the future. Rio Branco, such as most of South American cities, lacks in detailed administrative information of costs by local and national authorities, due to insufficient human and financial capital, and no developed insurance system. Presentation of these results can lead to the raise of awareness by local authorities for the need to improve assessments by conducting surveys, storage and dissemination, and administration of finances. This could benefit populations living in risk-prone areas as an improved administration of data stresses the need for support. Data on previous floods motivates governmental actors to reconsider their policies and to understand the possible benefits of investing in mitigation rather than relief projects. Hence, we conclude that socio-economic impact assessments can be useful to improve sustainable mitigation projects in order to decrease future impacts of floods. References Almeida, L. Q., Welle, T., and Birkmann, J. (2016). Disaster Risk Indicators In Brazil: A Proposal Based On The World Risk Index. International Journal of Disaster Risk Reduction, 17, 251–272. Aráujo Lambert, M.N.M., (2015). The challenges from extreme climate events for sustainable development in Amazonia: the Acre state experience. AGU Fall Meeting. San Francisco. Buchanan, J. M. (1991). Opportunity cost. The World of Economics (pp. 520-525). Palgrave Macmillan UK. Chatterton, J., Viavettene, C., Morris, J., Penning-Rowsell, E., Tapsell, S. (2007). The costs of the summer 2007 floods in England, Project: SC070039/R1. Delivering benefits through evidence. Downton, M. W., and Pielke Jr., R. A. (2005). How Accurate are Disaster Loss Data? The Case of U.S. Flood Damage. Natural Hazards. 35, 211–228. De Groeve, T., Poljansek, K., Ehrlich, D., Corbane, C. (2014). Current status and best practices for disaster loss data recording in EU member states.: a comprehensive overview of current practice in the EU member states. Report by Joint Research Centre of the European 15 Commission, JRC92290, doi, 10, 18330. Egorova, R., van Noortwijk, J. M., and Holterman, S. R. (2008). Uncertainty in flood damage estimation. International Journal of River Basin Management, 6(2), 139-148. Elmer, F., Kreibich, H., Seifert, I., and Thieken, A. H. (2010) A Delphimethod expert survey to derive standards for flood damage data collection, Risk Annual. 30, 107–124 EM-DAT (2015). Disaster trends. Retrieved on the 10 October 2016, from: http://www.emdat.be/disaster_trends/index.html. Espinoza, J. C., Marengo, J. A., Ronchail, J., Carpio, J. M., Flores, L. N., and Guyot, J. L. (2014). The extreme 2014 flood in south-western Amazon basin: the role of tropical-subtropical South Atlantic SST gradient. Environmental Research Letters, 9(12), 124007. Federal University of Santa Catharina (2012). Capacitação dos gestores de defesa civil uso de Sistema Integrada de Informações sobre Desastres. [Capacitation of civil defense managers on the use of the integrated system on information of disasters]. Floriánopolis. Federal Emergency Management Agency (2011). Hazus®- MH: multi-hazard loss estimation methodology. Washington. Green, C., Viavattene, C., and Thompson, P. (2011). Guidance for assessing flood losses, CONHAZ Report. Flood Hazard Research Centre, Middlesex University, London. Handmer, J. (2013). The chimera of precision: Inherent uncertainties in disaster loss assessment. The Australian Journal of Emergency Management, 18, 88–97. Helmer, M., and Hilhorst, D. (2006). Natural disasters and climate change. Disasters, 30(1), 1-4. Henley, A., Arabsheibani, G. R., & Carneiro, F. G. (2009). On defining and measuring the informal sector: Evidence from Brazil. World Development. 37(5), 992-1003. Hummell, B. M. L., Cutter, S. L., & Emrich, C. T. (2016). Social Vulnerability to Natural Hazards in Brazil. International Journal of Disaster Risk Science, 7 (2): 111-122. IBGE - The Brazilian Institute of Geography and Statistics (2015). Acre, population census. Retrieved on 16 February 2018, from: https://cidades.ibge.gov.br/v4/brasil/ac/riobranco/panorama. Intergovernmental Panel on Climate Change (2012). Managing the risks of extreme events and disasters to advance climate change adaptation: special report of the intergovernmental panel on climate change. Cambridge University Press. Koks, E. E., & Thissen, M. (2014). The economic-wide consequences of natural hazards: an application of a European interregional input-output model. Conf. Pap. 22nd Input Output Conf., Lisboa, Port.,2014. Retrieved on 22 April 2018, from: 16 https://www.iioa.org/conferences/22nd/papers/files/1479_20140509071_Koks_IIOA_ID14 79.pdf Lassa, J.A. (2010). Institutional vulnerability and governance of disaster risk reduction: Macro, meso and micro scale assessment (with case studies from Indonesia). Doctoral dissertation, Bonn University, Germany. Leskens, J. G., Brugnach, M., Hoekstra, A. Y., and Schuurmans, W. (2014). Why are decisions in flood disaster management so poorly supported by information from flood models? Environmental modelling & software, 53, 53-61. Malhi, Y., Aragão, L. E., Galbraith, D., Huntingford, C., Fisher, R., Zelazowski, P., and Meir, P. (2009). Exploring the likelihood and mechanism of a climate-change-induced dieback of the Amazon rainforest. Proceedings of the National Academy of Sciences, 106(49), 2061020615. Marchezini, V. (2015). The Biopolitics of Disaster: Power, Discourses, and Practices, Human Organization, 74 (4), 362–371. Marchezini, V., Trajber, R., Olivato, D., Aguilar, V. M., Pereira, F. de O., & Luz, A. E. O. (2017). Participatory Early Warning Systems: Youth, Citizen Science, and Intergenerational Dialogues on Disaster Risk Reduction in Brazil. International Journal of Disaster Risk Science. http://doi.org/10.1007/s13753-017-0150-9 Marengo, J. A., & Espinoza, J. C. (2015). Extreme seasonal droughts and floods in Amazonia: causes, trends and impacts. International Journal of Climatology. Merz, B., Kreibich, H., Schwarze, R., &Thieken, A. (2010). Review article" Assessment of economic flood damage". Natural Hazards and Earth System Sciences, 10(8), 1697-1724. Meyer, V., Becker, N., Markantonis, V., Schwarze, R., Van Den Bergh, J., Bouwer, L., ...& Hallegate, S. (2013). Review article: Assessing the costs of natural hazards-state of the art and knowledge gaps. Natural Hazards and Earth System Sciences, 13(5), 1351-1373. Morgan, M.G. and M. Henrion. (1990). Uncertainty. A guide to dealing with uncertainty in quantitative risk and policy analysis. Cambridge University Press. New York. 339p. Munoz, V. A.; Marchezini, V; Santos, L. B. L.; Jimenez, N.; Velásquez, A. (2017). DesInventar: ferramenta conceitual e plataforma computacional para sistematização de dados e suporte à pesquisa de risco e desastres. In: Marchezini, V.; Wisner, B.. Londe, L.R.; Saito, S. (eds.), Reduction of vulnerability to disasters: from knowledge to action.1 ed.São Carlos/SP : RiMa, 2017, v.1, p. 311-334. Ovando, A., Tomasella, J., Rodriguez, D. A., Martinez, J. M., Siqueira-Junior, J. L., Pinto, G. L. N., ... and von Randow, C. (2016). Extreme flood events in the Bolivian Amazon wetlands. Journal of Hydrology: Regional Studies, 5, 293-308. Penning-Rowsell, E., Johnson, C., Tunstall, S., Tapsell, S., Morris, J., Chatterton, J., and Green, C. (2005). The benefits of flood and coastal risk management: a handbook of assessment techniques. Middlesex University Press, London. 17 Poussin, J. K., Botzen, W. W., & Aerts, J. C. (2013). Stimulating flood damage mitigation through insurance: an assessment of the French CatNat system. Environmental Hazards, 12(3-4), 258-277. Rezaeian, M. (2013). The association between natural disasters and violence: A systematic review of the literature and a call for more epidemiological studies. Journal of Research in Medical Sciences: The Official Journal of Isfahan University of Medical Sciences, 18(12), 1103–1107. SEAPROF (2015). Enchente 2015: relatório de avaliação de danos e prejuízos na área rural [The 2015 flood: assessment report on damage and loss in rural areas], Rio Branco SitGeo (2016). Sistema de Inteligência Territorial Georefferenciado. Retrieved on 13 September, from: http://sit.riobranco.ac.gov.br/ SwissRe (2011).Acompanhando o risco de inundações no Brasil: prevenção, adaptação e seguro são estandategias sensatas. [Handling flood risk in Brazil: prevention, adaptation and insurance are sensible strategies]. Zurich. Szlafsztein, C. F. (2012). Natural disaster management in the Brazilian Amazon: An analysis of the states of acre, Amazonas and Pará. INTECH Open Access Publisher. United Nations (2015). Sendai framework for Disaster Risk Reduction: 2015 – 2030. Hyogo, Japan World Bank (2014). Lidando com Perdas: Opções de Proteção Financeira contra Desastres no Brasil [Dealing with losses: options of financial protection against disasters in Brazil].World Bank, Washington. Zaidi, R. Z. (2018). Beyond the Sendai indicators: Application of a cascading risk lens for the improvement of loss data indicators for slow-onset hazards and small-scale disasters. International Journal of Disaster Risk Reduction. http://doi.org/10.1016/J.IJDRR.2018.03.022 18