The food production component of the FEWSVT tool estimates changes in social surplus due to changes in crop yield for a particular commodity. These values are derived using economic surplus methods that, given initial demand and supply, quantify the magnitude of the vertical shift in a commodity’s supply curve due to a change in yield (Alston et al., 1998). This shift in supply, along with the observed change in yield, dictates the magnitude of the change in total surplus for the commodity. In Figure 2, area a-b-c-d represents this monetary measure, due to a vertical supply curve shift (b-a).

This vertical shift in a commodity’s supply curve, herein designated as parameter “K”, can be derived in one of two manners. If there is available price and quantity data for both a baseline scenario and alternative scenario, one can use Eq. 1 from Manning and Ando (2022), where P ₀ and P ₁ are baseline and alternative scenario prices, Q ₀ and Q ₁ are baseline and alternative scenario quantity, and ε S is the price elasticity of supply. K = P 1 − P 0 + P 0 Q 0 ε S Q 0 − Q 1 (1)

However, there may be cases in which one does not have a reasonable expectation as to how prices and quantities will change from its baseline level. Letourneau et al. (2015) developed an alternative calculation for K which is driven by changes in crop yield instead of prices and quantities. Eq. 2 displays this analytically, where Y ₀ and Y ₁ are baseline and alternative scenario yields. Based on the inputs provided, FEWSVT can use either method to calculate K. K = Y 1 − Y 0 Y 0 ε S (2)

Once the shift in the supply curve is quantified, changes in value can be determined. The impacts of this shift depend on change in price and quantity, as well as supply and demand elasticities. In addition, the size of the market in question determines if changes to commodity yields within the region are large enough to influence international market price. If this is the case, trade effects with any region which the study area exports to, herein designated as the “Rest-of-World” (ROW), must be considered to accurately reflect the underlying effects of the changes in yield for the studied commodity. ROW price-elasticities for the commodity must also be considered to capture all market behavior.

Therefore, FEWSVT can model two kinds of cases dependent on spatial resolution. The first case assumes the study area in question does not have a large enough market share of the commodity in which changes to its crop yields would influence international market price for that crop. If yields within the study area changed over time, international market price would remain unchanged, as producers in the study area act as price takers. The underlying equations adapted from Alston et al. (1998) are as follows: ∆ C S = P 0 Q 0 Z 1 + 0.5 Z μ D (3) ∆ P S = P 0 Q 0 K − Z 1 + 0.5 Z μ D (4) ∆ T S = ∆ C S + ∆ P S = P 0 Q 0 K 1 + 0.5 Z μ D (5) Here, Z is the price change relative to its initial value and μ D / ε S are the domestic price-elasticities of demand (absolute value) and supply for the commodity. Since the study area does not influence market price in this case, initial price P ₀ does not change in reaction to alternative-scenario commodity quantities. The Z parameter is derived from the K parameter and price-elasticities of supply and demand. Thus, given baseline values of commodity price and quantity and alternative-scenario quantities, the valuation tool uses Eq. 5 to calculate changes in total surplus.

If the specified study area does have a large enough market share of the commodity to influence market price, trade effects with the ROW must be considered. One example of a study areas that fulfills this case is the Midwest corn market, which accounts for an approximately 50% share of U.S. corn production, with the United States. being the world’s largest corn exporter (Roberts and Schlenker 2013). There are no numerical thresholds within the tool that determine whether inputs fall under this case; rather, spatial location is the sole determinant used by the tool. In these instances, the underlying equations (adapted from Alston et al., 1998) are as follows, for both the study area (subscript “DOM”) and areas who receive exports from it (subscript “ROW”): ∆ C S D O M = P 0 C A , 0 Z 1 + 0.5 Z μ D (6) ∆ P S D O M = P 0 Q A , 0 K − Z 1 + 0.5 Z μ D (7) ∆ T S D O M = P 0 Q 0 K 1 + 0.5 Z μ D (8) ∆ C S ROW = P 0 C B , 0 Z 1 + 0.5 Z μ B (9) ∆ P S ROW = P 0 Q B , 0 Z 1 + 0.5 Z ε B (10) ∆ T S ROW = P 0 Q 0 Z 1 + 0.5 Z μ B (11) Here, C _A,0 is initial domestic (study area) consumption, Q _A,0 is initial domestic production, C _B,0 is initial ROW consumption, Q _B,0 is initial ROW production, and μ/ε are the price-elasticities of demand (absolute value) and supply for the study area (subscript D) and ROW (subscript B). For domestic (study area) producers, we assume price elasticities account for demand and supply for both domestic and ROW consumption. For ROW areas, we assume price elasticities only account for exports received from the study area. The Z parameter is calculated in the same manner as the first case.

Trade effects are only captured by incorporating ROW supply and demand price elasticities for the commodity, and we assume that there are no trade-distorting policies that may otherwise influence market price. There are several reasons why we argue that this assumption will not lead to false estimates of surplus. Policies such as ethanol subsidies mainly impact demand for the commodity (in this case, corn), and since the shock that is fed into the tool is to supply, the effects of these policies are largely not felt for the desired parameters of interest (Babcock and Fabiosa 2011). Furthermore, many of the largest export destinations for United States. corn maintain free trade agreements under which exports enter at low or zero tariffs, which mitigates concerns over the impacts of trade distortions on model estimates (USTR 2021). Thus, given baseline values of commodity price and quantity and alternative-scenario quantities, the valuation tool uses Eq. 8 to calculate changes in total surplus.

In this paper, we input crop yield and harvested area data as described in Section 3.2 into FEWSVT Version 1.0, which assigns elasticity values as specified in Table 1. While the tool contains multiple measures of domestic and ROW elasticities, this study applies domestic elasticities from Roberts & Schlenker (2013) and ROW elasticities from Reimer et al., 2012. ²

The electricity generation component of FEWSVT estimates changes in value of power plant level production due to changes in electricity generation and marginal cost for a given technology. Understanding how measures of value change for certain electricity sectors are useful in examining the impact of various scenarios, such as future climate shock events. Utilizing methods from Logan et al. (2021), the valuation tool takes in spatially explicit user inputs of electricity generation (MWh) and marginal cost ($/MWh), and uses built-in electricity price data to output measures of present value of generation for any given electricity sector at an annual timescale. Given data inputs from both a baseline and alternative scenario, the tool can also calculate the change in value between scenarios.

For any given electrical sector, the valuation tool uses the following underlying equation, adapted from Logan et al. (2021): P V j = G j W − M (12) Here, PV _j is the present monetary value of electricity generation under conditions j, G _j is electricity generation (MWh) under conditions “j,” W is the wholesale electricity price ($/MWh) received for the given sector, and M is the marginal cost of electricity (MWh) for the given sector. The marginal cost term considers the capital (e.g., fuel) and variable operation and maintenance (O&M) costs of producing one additional unit of generation for a given electricity-generating technology.

Conditions j refers to the specific spatial and temporal input received by the tool for the electrical sector of interest. For example, the National Renewable Energy Laboratory’s (NREL) Regional Energy Deployment System (ReEDS) Model produces generation and cost projections by “balancing area” for the contiguous United States, and typically breaks down these parameters into 17 time slices that represent specific seasons and time-of-day throughout the calendar year (Brown et al., 2020).

Besides electricity generation and marginal cost, the user can also choose to specify their own electricity prices to produce monetary estimates per their own preferences. However, the tool does contain default historical wholesale electricity prices obtained from online independent system operator (ISO) databases for the Northeast and Midwest regions of the United States. We input electricity generation and price data as described in Section 3.2 and apply baseline marginal cost values as specified in Chang (2021).

Greenhouse gas emissions are a uniformly mixed pollutant, such that the damages of such pollution do not depend on spatial location. Rather, damages are uniformly distributed across space regardless of the source of pollution and are quantified solely as a simple function of the total net quantity emitted. Examples of such pollutants include carbon dioxide (CO₂) and chlorofluorocarbons (CFCs).

We use this component of the valuation tool to estimate changes in social wellbeing due to changes in net CO₂ flux, where CO₂ is emitted from power plants and absorbed through terrestrial sequestration. The valuation tool uses the following underlying equation: T D = S C C ∗ C O 2 P P − C O 2 S Q (13) Here, TD is the total net damages from carbon emissions, SCC is the social cost of carbon ($/ton), CO2 _PP (tons) is the quantity of carbon dioxide emissions from power plants, and CO2 _SQ (tons) is carbon sequestration levels. It is important to note that carbon sequestration data is typically measured in mass units of carbon. Thus, the valuation tool first converts mass carbon units into its carbon dioxide equivalent by multiplying by the ratio of the molecular weight of carbon dioxide to that of carbon. If the user does not have carbon sequestration data, the CO2 _SQ term is simply dropped from the equation and the resulting monetary estimate is changed to the total gross damages from just carbon dioxide emissions.

The valuation tool contains several default values of the SCC taken from literature that allows a user the flexibility to produce either a low, middle, or high monetary estimate depending on their preferences Hänsel et al. (2020) (Interagency Working Group on Social Cost of Greenhouse Gases, 2016; Nordhaus 2017). As shown in Table 2, all SCC values used by the tool assume a 3% discount rate and are updated to have uniform units of 2021 United States. dollars (USD) per ton of carbon dioxide emitted. The valuation tool also allows a user to input their own estimate which takes precedent over the default values within the tool. ³ This paper uses estimates of net carbon flux from both carbon dioxide emissions from power plants and net carbon sequestration from multiple land cover types for the Midwest and Northeast regions in the United States and applies a SCC value of $51/ton CO₂ as estimated by Interagency Working Group on Social Cost of Greenhouse Gases, 2016.

The water quality component of the valuation tool estimates changes in consumer WTP to use a water body due to changes to its water quality. The valuation tool first takes in water pollutant concentration inputs to produce a “Water Quality Index” (WQI) estimate, and then employs a benefit transfer methodology to calculate monetary values associated with people’s WTP for a water body’s WQI (Johnston et al., 2005; Alvarez et al., 2016; Johnston et al., 2017; Johnston and Bauer 2020).

The WQI is a standardized 100-point index that relates water pollutant concentrations to water body suitability for wildlife and human usage (Johnston and Bauer 2020). Higher WQI values indicate a water body of higher quality, with water pollutant concentrations at a level low enough to support aquatic life, recreational use, or even drinking water consumption. The WQI is calculated using pollutant concentration data inputs, which are translated into subindex values that are combined with an assigned index weight to produce an aggregate WQI estimate. The resulting WQI equation is as follows (adapted from Walsh and Wheeler 2013): W Q I = ∏ i = 1 6 Q i W i (14) where Q _i is the water quality subindex for water pollutant i, and W _i is the assigned index weight of water pollutant i. Table 3 shows all assigned subindex weights for water pollutants in freshwater water bodies. For each water pollutant, shown in Table 4, subindex values are first calculated on a 10–100 scale dependent on pollutant concentration. Once these subindex values are derived, Eq. can be used to generate a WQI estimate for any water body.

A benefit transfer methodology is then applied to estimate the monetary value of changes to the initial WQI estimate. The tool uses a meta-regression model developed by Johnston et al., 2017, using 140 unique observations from 51 stated preference studies published between 1985 and 2013 that estimated WTP for changes to WQI as a function of independent variables (Berrgstrom and Taylor 2006) that characterize features of the study site, water bodies, and nearby affected population from studies in the analysis. The variables used in this model are listed in Table B1 of Chang (2021) and the coefficients on those variables resulting from a random effects regression are summarized in Table B2 of Chang (2021). These coefficients are matched with any geographic features of the water body, as well as the socioeconomic characteristics of those who reguarly utilize it. For example, if the water body in question was located in the United States. Department of Agriculture Northeast region, the user would assign a binary value of “1” to the variable “Northeast_United States.” Once all variable coefficients have been matched with its selected value, household WTP is calculated using standard Consumer Price Index (CPI) adjustment for a specified increase in WQI for the desired water body. Household data is then used to calculate region-wide WTP.

This paper inputs nitrogen concentrations (mg/L) for water bodies located in the Midwest and Northeast United States. For the other five water pollutants that comprise the WQI, the tool will assign a default subindex value of 50 for those pollutants. Thus, the resulting changes in WQI across the climate scenario described in Section 3.2 only reflects changes in nitrogen concentrations.

The NLDAS-2 data was obtained from NASA Goddard Earth Sciences Data and Information Services Center (DES DISC, https://disc.gsfc.nasa.gov/), that covers the CONUS from 1980 to 2019. The historical NLDAS-2 data and intensified climate data were fed into a) the Terrestrial Ecosystem Model (TEM) to produce the carbon sequestration data (Kicklighter et al., 2023 this issue), b) the Integrated Science Assessment Model (ISAM) to produce the food production data (Lin et al., this issue), c) the Water Balance Model (WBM) coupled with Thermoelectric Power & Thermal Pollution Model (TP2M) to produce electricity production data as well as CO₂ emission data (Zhang et al., this issue), and d) the SPAtially Referenced Regression On Watershed attributes (SPARROW) model to produce the water pollution data (Maxfield et al., 2023 this issue). All model simulations were aggregated into each state within the C-FEWS framework defined NE and MW regions (Figure 3, Vorosmarty et al., 2023 this issue). The brief description of the model simulation data that was used in this study are listed in Table 5.

For the historical baseline and heat wave scenarios, we received data from several sources to apply as inputs to the FEWSVT tool. Food production data was obtained from Lin et al. (2023 this issue) for corn and soybeans via the Integrated Science Assessment Model (ISAM), a process-based dynamic crop and vegetation model that simulates the productivity of food and bioenergy crops (Niyogi et al., 2015).

Annual electricity generation (MWh), electricity price ($/MWh) and carbon dioxide emissions (metric tons) data were obtained from the Thermoelectric Power and Thermal Pollution (TP2M) model, which is a simulation model that simultaneously quantifies thermal pollution from rivers and estimates efficiency losses in electricity generation (Zhang et al., 2023 this issue) were obtained from Zhang et al., 2023. Marginal cost of electricity production ($/MWh) data was taken from Chang (2021) and were matched by both time and space to the TP2M generation data.

Annual carbon sequestration (Tg C) data was obtained from Kicklighter et al. (2023 this issue) via the Terrestrial Ecosystem Model (TEM), which is a process-based biogeochemistry model that uses spatially referenced information to estimate fluxes of carbon in multiple land cover types (Felzer et al., 2004). ⁴

Annual water quality data for the baseline scenario are obtained from Maxfield et al. (2023 this issue) via the United States. Geological Survey’s Spatially Referenced Regression on Watershed Attributes (SPARROW) model, which estimates long-term average values of water characteristics by synthesizing monitoring and geographic data (Bergamaschi et al., 2014). Water quality data for the heat wave scenario are calculated based on the following equation, in which TN_river is total nitrogen concentration at a given reach, TN_land is overland nitrogen concentration moving to a river, and T is water temperature in degrees Celsius. T N river = T N land ∗ e − 0.07 * T (15)

For the baseline and heat wave scenarios, we only input annual nitrogen concentrations (mg/L) by county. The other four water pollutants that comprise the WQI estimate (TP, DO, BOD, TSS, FC) are assumed to hold a subindex value of 50 and are unchanged from one scenario to the next. Thus, as detailed in Section 3.1.3, the resulting monetary estimates reflect the change in WTP due to a change in WQI resulting only from a change in nitrogen concentrations. We also received additional information regarding the water body type(s) and geographic location of each region, which dictates the usage of benefit transfer coefficients that are used to calculate monetary values.

All annual inputs for food production, electricity generation, and net CO₂ flux are fed into the FEWSVT tool at the state level for the years 2010–2019. Water quality inputs are provided at the county-level for the aggregated period from 2010–2019. In other words, water pollutant concentrations were provided as the average concentration across 2010 to 2019, as opposed to individual annual values from 2010 to 2019.

The C-FEWS framework contains multiple non-climate and climate scenarios to investigate and quantify the sensitivity of a FEW system (Vorosmarty et al., 2023 this issue). This paper focuses only on the difference between the baseline scenarios and one out of 4 extreme climate scenarios to demonstrate the characteristics of the valuation tool. The baseline scenario is the 40 years historical climate recorded in NLDAS-2 dataset that drove all the C-FEWS models to produce the datasets needed for this study.

The intensified extreme climate scenarios (IECS) that were described in Approach B in Vorosmarty et al. (2023 this issue), where the most climate extreme year in each of four types (heat wave, cold wave, heavy precipitation, and drought) within the last decade (2010–2019) and its 2 sub-sequential years were selected to repeat three times to replace 9 years of climate data from 2010 to 2019. In this study, the focus is the heat wave. The extreme heat wave years 2012–2014 for MW and 2016–2018 for NE were selected to repeat three times and replace the 2010–2018 climate data from the baseline. The intensified extreme climate datasets were then used to drive all the C-FEWS models to generate the data for this study. Figure 4 shows the air temperatures of the baseline scenario and the heat wave IECS from 1980–2019 to show how the latter is constructed. As heat wave has been predicted to be more frequent in the future (Raghavendra et al., 2019; Lorenzo et al., 2021), the scenario chosen in this study is relevant to likely future climate outcomes.

This paper examines the monetary implications of heat wave conditions as specified in Vörösmarty et al., 2023 compared to historically typical (baseline) climate conditions for two defined geographic regions in the United States (Midwest (MW) and Northeast (NE)) ⁵ . The heat wave data received from sources described in Section 3.2.1 reflect one of several climate extreme scenarios described in Vörösmarty et al., 2023. Results from FEWSVT indicate the monetary implications of changing meteorological conditions related to a modeled heat wave scenario from historical conditions as a result of changes in physical flows in food production, electricity generation, carbon dioxide emissions, and water quality (total nitrogen concentrations).

We use valuations of the heat wave scenario results to show which changes in climate change-attributable physical flows are likely to have the largest monetized value. We first use FEWSVT to value the monetary impacts due to changes in physical flows for each FEWS parameter. Then we rank all FEWS parameters by magnitude to examine which elements contain the highest monetary impacts directly attributed to the modelled changes in physical flows.

Researchers outside economics often use simple approaches to estimate the value of changes in sectors of the economy that yield marketable goods. If a scenario change yields a change in the quantity of a good like electricity or crops produced, it is tempting to represent the value of that change as the change in the market revenue from selling that output assuming a fixed price. However, as pointed out by Letourneau et al. (2015), that kind of simple revenue-change calculation can produce highly inaccurate measures of the actual changes in social wellbeing. If output falls because of an increase in production costs, the revenue-change calculation will understate the loss to society because it doesn’t account for increased costs of producing the quantity still produced. On the other hand, the revenue-change calculation can overstate the loss to society because it does not subtract production costs from the social value of the quantity of good no longer produced.

The FEWSVT tool estimates the value to society of changes in production costs and market outcomes for two kinds of goods sold in marketplaces: electricity and crops. In this part of the analysis, we conduct simple revenue-change calculations for each of the changes in outputs associated with the change to the extreme scenario by multiplying the original market price by the change in quantity. We then compare those values to the theoretically sound values produced by the models in the FEWSVT tool.

Managers and policy makers may need to make choices between policies and investments that have trade-offs between physical outcomes. For example, investing further in tile drainage in the Midwest could improve crop production in the face of increasing wet springs, but worsen water pollution because of increased nitrogen runoff. Policy makers could create an index of normalized physical changes (like percentage changes) in qualitatively different physical flows and pick the investments that do best in terms of that index. Alternatively, they could monetize the changes and pick the investment that does best in terms of total monetized value. This section uses data from the baseline and the heat wave scenario in a simple demonstration of how monetization can affect one’s judgment regarding the relative merits of investments that lead to one set of outcomes versus another.