Water scarcity is largely due to regulatory measures such as drought orders and discharge temperature limits.
In addition to departmental drought orders, a framework decree defines water use for installations classified for environmental protection. In our analysis, these use restrictions apply only to the manufacturing industry, which is not included in the activities covered by departmental orders (Table A C.1).
Use restrictions applied in the study to the various activities discussed
Note: Boxes are greyed out when no restrictions apply. Irrigation rules for specialised crops are applied to fruit and vegetable production.
Source: Authors, based on departmental and interdepartmental drought orders in the region and the framework order on installations classified for environmental protection.
Since restrictions are applied according to the status of each aquifer or river, it is important to know which activities depend on each source (Seine, Marne or groundwater). Data from the national database of quantitative water abstractions are used to determine the quantity of water abstracted annually by each site (over 10 000 m3 per year). As these different users are classified according to their activity, it is possible to determine the quantity of water abstracted by sector and by source. Since, within a given sector, the quantity of water abstracted is proportional to the economic value of the organisation, it is possible to estimate the share of the sector dependent on each of the water resources considered. A restriction applying to a river or aquifer therefore impacts each sector in proportion to its dependence on that source.
Note: The breakdown of abstraction is based on abstraction data for 2018.
Source: Authors, based on 2018 abstraction data from the national database of quantitative water abstractions.
Agriculture
The economic impact of droughts on agriculture results from yield losses caused by two factors: soil dryness and restrictions on irrigation. The agricultural sector is divided into two subsectors:1 field crops – which include cereals (wheat, barley, maize), oilseeds (rapeseed), beet and potatoes – and fruit and vegetables (excluding potatoes). Field crops are affected by soil dryness and restrictions on irrigation. On the other hand, fruit and vegetable production already relies heavily on irrigation (65% in 2020 (DRIAAF, 2022[1])), and climate change will only increase crop dependence on this technology (45% increase in irrigation needs by 2050, according to our socio-economic scenario). It has therefore been assumed that fruit and vegetable production is only impacted by restrictions on irrigation, and that these restrictions impact all production (identical to having 100% of fruit and vegetable crops irrigated by 2050).
The impact of soil dryness on field crop yields has been considered in two stages. Firstly, the potential impact of climate change on agricultural yields has been approached using a CCR model. This model is based on the Drought and Overwhelmed Key Indicator (DOWKI), which estimates yield losses as a function of climate projections (rainfall and temperature). In this way, the model can be used to estimate a distribution of yield losses (or gains) for various climate scenarios through to 2050. The yield loss corresponding to the 90th percentile in an RCP8.5 scenario has been retained. To date, however, this model has been calibrated only for wheat, barley and grass. To estimate yield losses for other crops, historical yields for the region have been used. The aim is to compare the yield losses estimated by the model for barley and wheat with "real" historical yield losses. If the wheat and barley yield losses estimated by the model match the losses observed for these crops in a previous dry year, then the losses incurred that same year are applied to the other crops considered (field crops only).
The historical yield gains or losses have been calculated using the method developed by (Debaeke and Bertrand, 2008[2]). Figure A C.1 shows historical yields for wheat in the region from 1968 to 2022. These yields have been increasing almost linearly over time since the 1970s (this trend is similar for all crops under consideration). It is therefore possible to estimate the theoretical yield for a crop in a given year by using the equation of the straight regression line shown in the graph. The deviation from the theoretical value provides an approximation of the yield gain or loss for the given year. The wheat and barley yield losses estimated by the CCR model match those recorded for 1976, a historically dry year for the region. The yield losses due to soil dryness that have been used for all crops therefore match the losses observed during the 1976 harvest.
Source: Authors, based on data from regional and national agricultural statistics.
The impact of restrictions on irrigation has been calculated using irrigation efficiency (or water use efficiency) data specific to each crop (see Table A C.3). Irrigation efficiency describes the yield gain obtained per additional mm of precipitation (or irrigation). Irrigation efficiency data for field crops (except beet) were provided by Arvalis,2 and values for other crops were estimated as the ratio between average yield per hectare in 2020 (DRIAAF, 2022[1]) and annual water requirements (see source table below). In line with the information supplied by Arvalis, it has been estimated that one round of irrigation corresponds to 30 mm of effective precipitation, and that farmers carry out a maximum of one round of irrigation per week per plot. Thus, a one-week ban on irrigation equates to a yield loss equal to three times the irrigation efficiency divided by the average yield. This absolute yield loss is expressed as a percentage yield loss.
Source: Author, based on data from sources listed in the last column.
The total economic cost to agriculture of each scenario is equal to the yield losses caused by each scenario (for field crops and fruit and vegetable production) multiplied by the selling price of each crop. For field crops, total yield losses are equal to the sum of the percentage losses due to soil dryness and the percentage losses due to irrigation restrictions, multiplied by the percentage of the surface area irrigated. Restrictions on the irrigation of field crops correspond to a total ban on surface water abstraction when the crisis threshold is breached, and reductions of 5%, 20% and 40% respectively when the alert, heightened alert and crisis thresholds for groundwater are breached.
Total yield losses for fruit and vegetable production are equal to losses resulting from restrictions on irrigation. These restrictions correspond to a total ban on surface water abstraction for irrigation when the crisis threshold is breached, and a 5% reduction in abstraction when the same threshold is breached for groundwater. The selling price of each crop is the selling price in 2020. This price has been obtained by combining 2022 selling price data for each crop (raw material prices for field crops and Rungis market prices for fruit and vegetables) and the INSEE monthly agricultural producer price indices. Taking contemporary producer prices into account corrects for any bias towards over- or underproduction in 2020.
The total cost to the agricultural sector has been determined to be the total percentage yield loss for each of the two sectors (field crops and fruit and vegetables) multiplied by the value of each of these sectors in 2050. The future value added of each sector has been determined to be the product of the value of production per hectare in 2020 and of the surface area occupied by each of these sectors by 2050. The change in surface area equates to the changes described in the economic scenario.
Source: Author calculation.
Water transport
As indicated earlier, only pleasure boating will be impacted by droughts in 2050 and 2100. Pleasure boating is affected when the crisis threshold of drought orders is crossed, which interrupts filling and thus navigation on the canals. In this study, the stoppage of boating activity is counted in days of restrictions on filling (navigation). The daily cost of stopping boating activities is estimated as being equal to average tourist spending by all pleasure boaters per day of navigation in the region during the summer period, which accounts for over 50% of all boating activity.
Source: Authors.
Energy production
The Paris metropolitan area’s energy production relies on water for hydroelectric power generation using dams, as well as for cooling and steam production (heating network), and for cooling waste-to-energy plants. Hydropower, whose potential seems to have been achieved in the region, and which accounted for only 0.4% of the electricity generated in the region in 2021 (ROSE, 2023[3]), has been excluded from the study. Heating networks have also been excluded from the cost analysis. They generate steam that is used to produce hot water for many hospitals and other critical facilities and are therefore unlikely to be affected by restrictions on abstraction. These networks must also be kept under continuous pressure, or they will cease to function. According to the Compagnie Parisienne de Chauffage Urbain [Parisian Company for Urban Heating], it could take more than six months to restart the system.
The study therefore focuses on waste-to-energy production and cooling. These activities draw most of their operating water from the Seine and are sensitive to the quantity (depth) of water in the Seine to enable pumping. As shown before, the depth of water in the Seine remains unchanged in all scarcity scenarios, so there is no threat to pump submersion. On the other hand, these activities are also subject to regulations concerning the temperature of discharges into the natural environment, and in particular the decree dated 24 August 20173 which sets a temperature limit of 30°C for water discharge. The average rise in water temperature (the difference between the temperature of the water drawn from the environment and that of the water discharged downstream) for these plants is 5°C. To maintain the statutory discharge temperature, the facilities have to reduce their output in proportion to the river temperature. Specifically, when the temperature of the Seine rises above 25°C, energy production declines by 20% for each additional degree, reaching zero above 29°C.
There are two key organisations that use water from the Seine for cooling systems: SUC and Fraîcheur de Paris (formerly CLIMESPACE). SUC uses only Seine water for cooling, and based on information gathered through interviews, Fraîcheur de Paris will rely on Seine water for 70% of its needs (and on air cooling towers for 30%) by 2050. The amount of cooling delivered by the network is expected to reach 1 000 GWh/year in 2050 according to the master plan for the Paris cooling network, i.e. a 2.5-fold increase in the capacity that is currently delivered (Ville de Paris, 2019[4]). The same applies to SUC. According to information provided by Fraîcheur de Paris during interviews, cooling requirements (and therefore production) are seven times greater in summer than in winter. Production losses due to high temperatures in the Seine are therefore equal to average daily summer production multiplied by the number of days spent at temperatures exceeding 25°C, multiplied by the forced reduction in production (20% per additional degree above 25°C, zero production above 29°C). To calculate the economic cost of this loss of production, a selling price for cooling has been calculated by dividing Fraîcheur de Paris' sales in 2021 by its production in 2021 (the price applied is identical for SUC).
Source: Authors.
The same approach has been used to estimate the impact of drought on energy production (steam and electricity) via waste-to-energy production (Table A C.7). Electricity and heating resale prices have been taken from the AMORCE survey.
Source: Authors.
To estimate the impact of drought on the Paris metropolitan area’s energy sector as a whole, the economic losses described above have been compared with the sector's economic value (Table A C.8).
Source: Authors.
Manufacturing
Some industrial production sites are designated as installations classified for environmental protection. However, if these sites withdraw more than 10 000 m3 of water from the environment each year, they are subject to the water use restrictions set out in the order dated 30 June 2023 (Ministère de la transition écologique et de la cohésion des territoires, 2023[5]). This order provides for a 5% reduction in water abstraction when the alert threshold is crossed, a 10% reduction during an alert and a 25% reduction in the event of a crisis.
The installations classified for environmental protection nomenclature (Ministère de la Transition Ecologique et de la Cohésion des Territoires, 2023[6]) has been used to identify manufacturing sectors with the potential for such installations (INSEE classification). The agri-food (C1), manufacture of coke oven and refined petroleum products (C2) and the manufacture of other industrial products, including textiles, metals, paper and chemicals (C5) sectors are included in the nomenclature.4 The national database of quantitative water abstractions identifies 187 organisations that withdraw more than 10 000 m3 annually, representing 1.1% of industrial sites in these sectors in the Paris metropolitan area. Considering that France's largest industrial companies, representing 2% of the total number of legal entities, account for 42% of the sector's value added (INSEE, 2020[7]), we estimate that installations classified for environmental protection sites withdrawing more than 10 000 m3 account for around 20% of the value added in their respective sectors (C1, C2 and C5), or 13% of the value added of the region's industrial manufacturing sector.
It has been assumed that organisations optimise their withdrawals as far as possible in relation to their production levels, and that they therefore have no choice but to reduce production in proportion to the abstraction reductions set out in drought orders. Thus, the cost to manufacturing is calculated in proportion to the share of its value added affected by the orders. The regional gross domestic product figures published in INSEE in 2020 (INSEE, 2024[8]) have been used to estimate the value added of the manufacturing sector in the Paris metropolitan area, and it has been assumed that this value added will remain relatively unchanged in value and composition by 2050.
Drinking water production
The drought scenarios considered do not pose a risk to the production of drinking water in the Paris metropolitan area. The maintenance of water levels on the Seine and Marne rivers, thanks to the development of reaches and the support of reservoirs, enables water to be abstracted. Furthermore, even if the Oise, which does not benefit from lake support, were to fall to levels that prevented it from being pumped for drinking water production, the other plants in the interconnected network are able to produce drinking water in sufficient quantities to compensate for the shutdown of the Oise facility.
Droughts also threatens water quality. However, drinking water companies seem confident that there will be no significant cost associated with rising temperatures. For the moment, there are no data available to assess the potential costs associated with reduced dilution in the environment.
Built environment
The impact of droughts on the built environment is due to the phenomenon of clay shrinkage and swelling, which mainly affects detached homes built in the past without foundations. The result of alternating clay swelling during damp periods and shrinkage during dry periods, clay shrinkage and swelling is a multifactorial phenomenon, and its link with drought severity is still poorly understood. In the absence of an analysis of the impact of future droughts on the intensity of this phenomenon, the study reports the results of the estimate made by the French Senate (Sénat, 2023[9]), which quantified the cost of clay shrinkage and swelling at the national level. The Paris metropolitan area accounts for 11.5% of homes exposed to this risk (Institut Paris Région, 2022[10]). Assuming that the damage is proportional to the number of homes at risk, the damage suffered by the Paris metropolitan area in 2022 also represents 11.5% of the total damage suffered in France, i.e. EUR 346 million, or 0.02% of the value of region’s real estate assets.
Background and overview
The Adaptive Regional Input-Output (ARIO) macroeconomic model is an input-output (I-O) model designed to calculate the indirect costs of exogenous capital or production shocks. The economy is modelled as a set of economic sectors and a set of regions. In the subsequent text, an "industry" refers to a specific pairing (sector, region). Each economic sector manufactures a generic product and draws its inputs from an inventory. Each sector meets a total demand made up of final demand (household consumption, public spending and private investment) from all regions (local demand and exports) and intermediate demand5 (inventory replenishment). An initial equilibrium state of the economy is built based on multi-regional input-output tables (MRIO tables).
Two types of shock can be used: either at production level (external factors force an industry to produce less) or at capital level (an industry loses some of its production factors due to external factors and is therefore forced to produce less and rebuild its capital stock). The model then describes how external shocks spread through the economy at each time step (one time step corresponds to one day). The direct economic impact is made up of the above-mentioned shocks, while the total economic impact also includes indirect costs. The total economic impact can be measured in two ways: in terms of (i) unsatisfied final demand or (ii) relative production loss.
Detailed description
Initial state
The initial state refers to the economic equilibrium before the shock. The initial values for intermediate orders , final consumption and production are taken from the MRIO tables. An inventory comprises input stocks from which an industry can draw (see below for a detailed description of stocks) and is initialised using the initial intermediate orders : by default, it is assumed that each industry has N days' worth of inputs in advance in its inventory.6
Inventory
In the ARIO model, economic sectors do not directly use inputs from other sectors, but draw on their inventories, which can then be replenished by intermediate orders . An industry's inventory is a vector that specifies the quantity of each product that this industry has in stock. Each sector needs intermediate inputs in proportions set out in the MRIO tables in the initial state. Industries draw inputs from their inventories in an attempt to maintain optimal production levels. However, these inventories cannot be emptied: actual production may therefore be lower than optimal production to ensure that inventories are above a certain threshold (see Production section for details). Inventories have been integrated into the ARIO model to provide a more realistic depiction of supply chain shocks, which can be mitigated by input stocks. The current state of an inventory for a given input is expressed as the number of time steps a sector can produce with this input at the current production level.
Direct shocks
Direct shocks are exogenous and can occur at any time . Conceptually, these shocks are the direct economic consequences of the events modelled. This may involve (i) direct loss of production capacity or (ii) destruction of capital. Capital is the only factor of production, so in any sector, an x% destruction of capital translates into an identical x% loss of production, until the capital is replenished. There are two ways of modelling capital replenishment. The first is internal replenishment demand , which leads to a reduction in real production allocated to final demand or intermediate orders . The second is purely external, with no cost to economic operators: the capital stock in a directly affected industry returns to its initial level over time, without the need to reduce production.
Production
The production module calculates actual production for each industry at each time step .
To do this, it calculates the production capacity , which corresponds to production in the initial state minus the reduction in production capacity (direct exogenous reduction in production capacity and reduction in production due to the destruction of capital not yet replenished). In the ARIO model, capital is the only factor of production.
Optimal production is defined as the minimum between production capacity and total demand: the model is demand-driven, with industries producing no more than total demand.
Finally, actual production is calculated from , taking inventory constraints into account. Actual production is a Leontief function of inputs. Industries produce within the limits of the inputs they have in stock. More precisely, an industry can only produce if it has enough inputs of each type to produce for n consecutive time steps, in order to model the cautious anticipatory behaviour of producers.7 Basically, if the stock of an input is x% less than the quantity needed to produce , actual production is reduced by x% compared to optimal production , so that actual production levels are in line with stock constraints. This constraint must be satisfied for each type of input involved. The inputs required for production are then drawn from the respective stocks.8
Orders and demand
The order/demand module calculates the different types of demand. Total demand is made up of final demand , intermediate orders and replenishment demand .
Final demand for each region is exogenous and set by the MRIO tables.
Intermediate orders are determined on the basis of inventories and break down into two parts: the first is the quantity of inputs used to produce in the current stage, while the second is a fraction of the remaining deviation from the stock target. Stock targets are defined as the stocks needed to produce at .
If capital replenishment is internal, only a fraction 9 of the remaining replenishment demand is ordered at each stage, so that replenishment is not instantaneous but takes a characteristic period of time . Replenishment demand is used to replenish capital stocks, thereby restoring production capacity to pre-shock levels. If capital replenishment is external, there is no replenishment demand.
Distribution
The distribution module distributes actual production among the various types of demand. Distribution follows a proportional rationing scheme: if total demand cannot be met, intermediate orders , final demand and replenishment demand receive a share of actual production proportionate to their share of total demand.
Overproduction.
In the ARIO model, when industries are unable to meet total demand, they can temporarily increase their production capacity from to . This is a gradual process: if necessary, the overproduction factor can be increased from 1 to a base value >1. can then rise to with an exogenous characteristic time period .10 The increase from to also depends on a scarcity index, defined as unmet demand divided by total demand: the higher the scarcity index, the faster the increase. In the current version of the ARIO model, overproduction is free for economic operators: it is limited by the fact that it is not instantaneous and cannot exceed .
Trade in goods
Goods produced in the same sector but in different locations are perfectly substitutable. The trade balance is not modelled, but is taken into account in the initial state.
Changes to suppliers and buyers
Industries can, to a certain extent, change suppliers and buyers, which introduces more substitutability in the ARIO model compared with a "pure" input-output model. When shortages start to emerge, demand is still distributed among suppliers in the same proportions as before the shock. However, as unmet demand increases, suppliers who have not been affected by a direct impact (on capital or production) may overproduce and capture demand from new buyers. Buyers will then marginally shift their orders to producers who can overproduce. This article is inspired by (Guan et al., 2020[11]). In the original version of the model, industries distribute demand for intermediate goods to each supplier in the same proportion as in the original equilibrium state, making supply chains much less flexible. The ARIO model can be run with rigid ("noAlt" parameter) or flexible ("Alt" parameter) supply chains.
Features and limitations of the model
The model has two main limitations: the simplicity of the mechanisms described and the amount of data required to run the model. Although some characteristics of the economy – such as the existence of inventories that mitigate the propagation of shocks, or the introduction of replenishment demand in response to the destruction of capital – have been added to the ARIO model to offer a more detailed approach to the workings of the real economy, these mechanisms are still simple. Unlike computable general equilibrium (CGE) models, the model does not take into account possible price variations or characteristics relating to the general equilibrium. It also uses rigid rules, similar across all economic sectors, to model company behaviour. Based on an input-output modelling framework, the ARIO model is designed for studying the consequences of short-term economic shocks: operators have limited options for substitution, both as suppliers and as buyers.
The second limitation is the large amount of input data required to run the model. In general, data to assess a wide range of parameters (e.g. characteristic times or the overproduction factor ) are not available at the local scale and standard values from the literature must be used. These are often identical for every sector.
The two limitations are linked: the more complex the model, the more data you need to calibrate it. ARIO was therefore chosen as a compromise between these two concerns.
Calibrating ARIO for the Paris metropolitan area
The ARIO model needs to be calibrated on an initial economic equilibrium, determined using the EUREGIO database (European Commission, Joint Research Center (JRC), 2020[12]), which is a set of coherent inter-regional input-output (I-O) tables containing information both on sectoral specialisation, and on the links between different economic sectors within and between different regions (Thissen et al., 2018[13]). These tables are available for each year from 2006 to 2010. The study is based on the most recent version (EUREGIO 2010). The economic sectors detailed in EUREGIO are shown in Table A C.9.
The 24 European countries detailed at the regional level in EUREGIO are divided according to the NUTS 2 convention. The Nomenclature of Territorial Units for Statistics (NUTS) is a hierarchical system for dividing up the economic territory of the European Union and the United Kingdom, and NUTS 2 corresponds to the regional level.
As the latest available EUREGIO table is for 2010, simulations are carried out using technical coefficients that describe the state of the world economy at that time. This method assumes that the structure of the global economy in 2050 will not differ too much from that in 2010.
Implementing shocks resulting from droughts in ARIO
The direct economic costs of drought events can take two forms.
Asset damage: for example, saltwater intrusion leading to soil degradation, damage to buildings due to clay shrinkage and swelling, or damage to ecosystems due to a lack of water (Freire-González, Decker, and Hall 2017).
Reduced production capacity: for example, rain-fed and irrigated agriculture, difficulty in supplying drinking water, low flow rates reducing hydroelectric generation, high water temperatures limiting nuclear power generation (World Bank 2019).
Damage can be implemented in the ARIO model in the form of capital destruction or reduced production capacity. The magnitude and duration of these direct economic impacts need to be known and distributed across the EUREGIO economic sectors. Damage to assets has not been included in the calculation of indirect costs, because it mainly consists of damage to homes caused by clay shrinkage and swelling, and because taking these losses into account is complex. Firstly, it is difficult to ascertain whether or not the property is available during the renovation work: if it is, it would make no sense to reduce the value added of the "rental" (real or imputed). Secondly, imputed rents are not included in EUREGIO (Thissen et al., 2018[13]). Finally, the indirect costs associated with the reduction in rental value added are not qualitatively clear: since the reduction in inventory occurs in private homes and not in commercial buildings, it is mainly a final consumption loss for households that does not spread further along the supply chain.
Sensitivity analysis
As described above, company behaviour is modelled using various parameters describing overproduction processes and inventories. These parameters are based on empirical values drawn from the scientific literature. However, the overproduction mechanisms led to unrealistic modelling results in the first simulations: for example, the agriculture sector in the affected region begins to overproduce before the drought is over, which is inconceivable on drought-affected land. Overproduction was originally implemented in ARIO to model the increase in production to fuel the replenishment of destroyed capital after the event was over. This is why overproduction is not allowed in our main parameter set, by choosing =1 instead of =1.25 (a value taken from the literature). The sensitivity of the results to this assumption is examined at the end of this note.
Direct cost scenarios
Only six EUREGIO sectors are affected by direct costs: 1) Agriculture, hunting, forestry, 2) Transport, storage and communication, 3) Mining and energy supply (scenarios 2 and 3 only), and manufacturing, 4) Agricultural and food industry, 5) Textile, leather and other industry and 6) Other manufacturing industries.
Direct costs, calculated in terms of monthly value added lost, by sector for each of the three scarcity scenarios, are used to model direct costs. These costs are expressed in 2020 euros and correspond to 2020 production levels for the region. Direct costs are then converted to values consistent with the EUREGIO 2010 input-output tables, by transforming 2020 euros into 2010 euros and adjusting these costs to 2010 value added. According to INSEE data, between 2010 and 2020, inflation was 11% and growth in real value added in the region was 14.61%.
Indirect cost simulations
The direct costs presented above are used as production stoppage scenarios in the ARIO model. The model is run with the set of parameters described above for the duration of the direct shocks (from January of year 1 to September of year 2). Most of the analyses that follow correspond to scenario 1. The next section details the timescales in which direct and indirect costs emerge, then explores the sectoral breakdown of indirect costs. The remainder of the note examines the geographical spread of indirect costs across France and the EU, as well as the effect of each sectoral impact separately, before presenting the results of the sensitivity analysis. Finally, as previous results were only presented for scenario 1, the final section compares results between the three scenarios.
Unless otherwise indicated, all cost values shown are value-added losses and not gross production losses.
Timescales for direct and indirect costs
The drought scenario considers two timescales:
Some impacts only occur in the summer of year 1 ("Mining and energy supply" and "Transport and communication").
Impacts on agriculture and manufacturing last a full year, but do not start at the same time (September of year 1 for agriculture, January of year 1 for manufacturing, Figure A C.2 by comparing panels [a] and [b]).11
Total impacts (blue line) can be much more significant than direct impacts (orange line) (Figure A C.2 [c] and [d]). For example, in graph [c], direct summer impacts are low compared with total impacts, which are mainly due to declining demand for the mining and energy supply sector from manufacturing industries.
Value-added loss for the region in the "Agriculture, hunting, forestry" (a), "Other manufacturing" (b), "Mining and energy supply" (c), and "Transport, storage and communication" (d) sectors
Source: Authors.
Breakdown of indirect costs by sector
All sectors of the French economy are affected by the drought-related production stoppages in the region. Some sectors, such as "Other manufacturing industries", are mainly affected by direct production stoppages (Figure A C.3: the orange bar representing total losses is almost at the same level as the blue bar representing direct losses). Others, such as distribution or real estate, rental and trading activities, are only affected by input-output links. The sectors experiencing the highest indirect losses in absolute terms are the "Real estate" and "Distribution" service sectors (Table A C.10). In the simulations, the propagation of losses is mainly explained by the fact that industries stop production and therefore do not buy intermediate inputs from their usual suppliers. There is no forward propagation (from producers to buyers) because production stoppages are not long enough or strong enough to offset the reduction in inventory.
Value-added losses expressed as a percentage of the sector's value added on a national scale
Source: Authors.
Source: Authors.
Geographical propagation of indirect costs
The loss of value-added spreads beyond the Paris metropolitan area to other French regions and even around the world, but not to the same extent. The loss of GDP relative to regional GDP for other French regions is between 1% and 10% of the relative loss for the Paris metropolitan area (Figure A C.5). Where other European regions are affected, their losses amount to less than 1% of the Paris metropolitan area’s relative losses. Figure A C.4 shows the French regions and European countries most affected in absolute terms.
Source: Authors.
Loss of value added for European regions (NUTS 2 level), as a share of GDP lost relative to the share of GDP lost by the Paris metropolitan area [left], and as a share of GDP [right]
Note: Left: The various French regions experience production losses of between 1% and 10% of the Paris metropolitan area’s losses as a proportion of their regional value added [red zone], while regions in neighbouring countries tend to experience losses of less than 1% of the Paris metropolitan area’s losses [orange zone]. Right: Losses are expressed here relative to production in the region studied. In the Paris metropolitan area, the loss of value added amounts to 0.12% of regional production over the duration of the simulation (612 days).
Source: Authors.
Effect of various sectoral impacts
The extent to which each sectoral shock spreads to the rest of the French economy is studied in Table A C.11. To isolate the significance of the shocks experienced by each sector, simulations are carried out using only the cost incurred by one sector. Indirect costs are then assessed and analysed by sector. It appears that shocks experienced by the manufacturing sector result in higher indirect costs per euro of direct shock (higher amplification ratios), particularly for the "Agricultural and food industries" sector. This is mainly due to the fact that manufacturing sectors have low ratios of value added to production, which means they use more intermediate consumer goods. As the propagation of costs is primarily backwards in the current simulations, these sectors cause higher intermediate consumption losses when they cease production.
Source: Authors.
Sensitivity analysis
Two assumptions were subjected to a sensitivity analysis for each scenario. Since replenishment requirements are zero, and the possibility of overproduction on such a short timescale seems unrealistic, α_max=1 (no overproduction) was chosen as the default parameter. A sensitivity analysis with a more conventional parameter value (α_max=1.25) was carried out. Greater rigidity in supply chain adjustment was also tested ("noAlt" parameter).
Amplification ratios (indirect/direct losses) are higher when there is no overproduction and when supply chains are more rigid (Table A C.12). The main configurations (first column) thus lead to higher total costs than other configurations where overproduction is allowed, but lower costs than when there is no overproduction and supply chains are more rigid.
Source: Authors.
Comparison of costs caused by the three scarcity scenarios
There is not much difference in the overall results for Scenario 1 and Scenario 2 (Table A C.13), as direct costs are almost identical and the sectoral breakdown is very similar. The costs caused by scenario 3 are significantly higher, which is intuitive since direct costs are higher.
Source: Authors.
[2] Debaeke, P. and M. Bertrand (2008), “Évaluation des impacts de la sécheresse sur le rendement des grandes cultures en France”, Cahiers Agricultures, Vol. 17/5, pp. 437-443, https://doi.org/10.1684/agr.2008.0230.
[1] DRIAAF (2022), Agreste Île-de-France - Mémento 2022.
[12] European Commission, Joint Research Center (JRC) (2020), Regional Input-Output Data for Europe, http://data.europa.eu/89h/84356c3b-104d-4860-8ce3-075d2eab37ab.
[11] Guan, D. et al. (2020), “Global supply-chain effects of COVID-19 control measures”, Nature Human Behaviour, Vol. 4/6, pp. 577-587, https://doi.org/10.1038/s41562-020-0896-8.
[14] Hallegatte, S. (2013), “Modeling the Role of Inventories and Heterogeneity in the Assessment of the Economic Costs of Natural Disasters”, Risk Analysis, Vol. 34/1, pp. 152-167, https://doi.org/10.1111/risa.12090.
[8] INSEE (2024), Produits intérieurs bruts régionaux et valeurs ajoutées régionales de 1990 à 2022, https://www.insee.fr/fr/statistiques/5020211.
[7] INSEE (2020), Les entreprises en France, Édition 2020, https://www.insee.fr/fr/statistiques/4986825?sommaire=4987235#tableau-figure7_radio1.
[10] Institut Paris Région (2022), Vulnérabilité de la région Ile-de-France aux effets du changement climatique.
[6] Ministère de la Transition Ecologique et de la Cohésion des Territoires (2023), NOMENCLATURE DES INSTALLATIONS CLASSÉES, https://aida.ineris.fr/sites/aida/files/2023-10/BrochureNom_v54public.pdf.
[5] Ministère de la transition écologique et de la cohésion des territoires (2023), Arrêté du 30 juin 2023 relatif aux mesures de restriction, en période de sécheresse, portant sur le prélèvement d’eau et la consommation d’eau des installations classées pour la protection de l’environnement, https://www.legifrance.gouv.fr/jorf/id/JORFTEXT000047784127.
[3] ROSE (2023), TABLEAU DE BORD, https://www.roseidf.org/panorama-regional-1-1-1/.
[9] Sénat (2023), Rapport d’information sur le financement du risque de retrait gonflement des argiles et ses conséquences sur le bâti.
[13] Thissen, M. et al. (2018), “EUREGIO: The Construction of a Global IO Database With Regional Detail for Europe for 2000–2010”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3285818.
[4] Ville de Paris (2019), Schéma directeur du réseau de froid.
← 1. As livestock production accounts for just 6% of the value of the region's agricultural production (DRIAAF, 2022[1]), and as the impact of a water shortage on production in this sector is very difficult to assess, the sector has been excluded from the analysis. Similarly, forests are highly sensitive to multi-year drought phenomena, but not very sensitive to extreme drought events (excluding forest fires), and the impact of scarcity on timber production is not considered in this analysis.
← 2. Only crops currently irrigated in the region have been used to calculate the impact of irrigation restrictions.
← 3. Decree dated 24/08/17 modifying, through a series of ministerial decrees, the provisions relating to the discharge of hazardous substances into water from installations classified for the protection of the environment.
← 4. Electrical and computer equipment manufacturing; manufacture of machinery (C3) and manufacture of transport equipment (C4) are not included in the installations classified for environmental protection nomenclature. They are therefore excluded from the potential impacts of drought orders on such installations.
← 5. The terms "intermediate demand" and "intermediate orders" are used synonymously throughout the rest of the note.
← 6. In the current version, N is based on a literature review and is generally equal to 90 days. See (Hallegatte, 2013[14]).
← 7. For this analysis, is defined on the basis of a literature review and is equal to 72 days (Hallegatte, 2013[14]).
← 8. There are certain exceptions to this general rule for products that are not considered essential to production and therefore cannot limit it. In the model, only two EUREGIO sectors have been qualified as non-essential: "Accommodation and catering" and "Financial activities".
← 9. The characteristics of the time variables and are defined as in days (Hallegatte, 2013[14]).
← 10. is generally 125% and is equal to 365 days. See days (Hallegatte, 2013[14]).
← 11. Production stoppages for the manufacturing sectors start earlier, as groundwater levels are low even before summer begins in the drought scenarios considered.