Extension communication plan

Many extension resources relevant to water recycling are available at individual land grant university websites (Brumfield and DeVincentis, 2014; Johnson et al., 2011; Newman, 2009). Understandably, USDA also funds extension initiatives whose goal is to aggregate preexisting information and provide a “one stop shop” for particular classes of growers. Lea-Cox et al. (2010) describe the creation of a “green industry knowledge center” for nursery and greenhouse operations: Water quality is a major focus of this initiative (Lea-Cox, n.d.). The Clean WateR³ website provides guidance on how to “reduce/remediate/reuse” irrigation water regardless of the commodity grown (White, 2019, 2021). Both of these websites funded by USDA increasingly feature interactive tools that allow growers to input customized information for immediate feedback. The tools currently featured on these sites are more oriented to plant science or irrigation engineering than to economics, so the present tool should be a welcome addition at these locations.

Regulatory risk module

Two common motivations for investing in water recycling are that the operator expects future water shortages or intensified state water quality regulations. Both motivations are prominent in the western United States, whereas the second one dominates in the eastern United States (Fulcher et al., 2016). A unique feature of the online tool is that it provides the user with a “regulatory risk score.” This score is calculated as the sum of a three-part ranking on drought risk and a three-part ranking on pollution problems in the user’s watershed. Both the individual risk components (1 = low risk, 2 = medium risk, 3 = high risk) as well as their sum (2 = lowest risk, 6 = highest risk) are reported to the user.

To create the drought risk score, we used 30-year average annual rainfall data (1981–2010) prepared by the PRISM group at Oregon State University in Corvallis (Oregon State University, 2012). We used geographic information system (GIS) software (ArcGIS; ESRI Corp., Redlands, CA) to calculate zonal statistics and provide the mean average rainfall value within each U.S. zip code. Zip codes that are reported by users of the online tool are then categorized into quantiles that correspond to risk scores 1, 2, and 3.

These rainfall data are now out of date because of climate change. We continue to use them for two reasons: 1) it is important to use a significant span of time to avoid basing the measure on years that are outliers, and 2) in North America, generally there is less rain in places that were already dry, and more rain in places that were already wet. Therefore, the rankings of U.S. zip codes would not change much as a result of climate change, and neither would a quantile-based risk score. It remains possible that this risk score has not captured recent reversals in precipitation trends that have occurred in the Pacific Northwest and other smaller geographies.

Pollution runoff risk was calculated using U.S. Environmental Protection Agency EnviroAtlas modeled estimates for dissolved nitrogen in surface runoff from agricultural fields in tons at the watershed level for data collected in 2002 (U.S. Environmental Protection Agency, 2016). GIS software (ArcGIS) was used to calculate the weighted average of nitrogen runoff with the area of individual watersheds within each zip code serving as the weights. We categorized zip codes into three quantiles that correspond to risk scores 1, 2, and 3.

To maintain confidentiality of user data, we erase all zip codes from the server after the user has received a final financial report. We never solicit names and addresses.

Finance concepts

The program’s algorithms use the partial budget approach (Kay and Edwards, 1994; Rabin et al., 2014) to evaluate the capital investment decision. Any aspect of an operation’s finances that remains unchanged by the investment is ignored. Instead, the goal is to measure only upfront capital costs and changes in annual cash flow that can be directly attributed to the investment. Categories of marginal costs and benefits that one might expect from the decision to recycle water at a nursery are shown in Table 1 (not everything in Table 1 is economic; see Cultice et al., 2016; Lamm et al., 2019; Raudales et al., 2014; and Warner et al., 2018 for fuller treatments of the operator decision-making process). The most important benefit in Table 1 that the online tool does not currently estimate is “benefits to the physical environment.” The existence of such benefits explains why USDA-NRCS provides technical support and grants to help with system installation.

Table 1.

Costs and benefits associated with the installation of a water recycling system at a nursery.

View Table

Two business-related costs associated with a recycling investment are ignored in the online tool. They are 1) regrading land to improve runoff of irrigation water to a tailwater pond; without a site visit, it is nearly impossible to collect the kind of data needed to customize this cost component; and 2) management time and effort. This second factor is an opportunity cost that is difficult to quantify, because return to management varies from producer to producer. Both of these costs could prove to be significant.

Net present value is emphasized as the standard criterion for the investment decision, and the user’s discount rate is solicited for this purpose. Capital costs are modeled as a one-time cash outflow at the start of the period. Real depreciation is assumed to be repaired annually, essentially treating it as an operating cost. With these regular repairs, the recycling system is assumed to be fully operational for at least 50 years, which is the time horizon for the present value calculation. Future inflation is ignored on the assumption that it affects costs and benefits equally, and that it is embedded within the discount rate (Ferraro et al., 2017).

List of inputs and available outputs

Table 2 lists the online tool user-specific data inputs and available outputs. To make the tool manageable for users, only 17 pieces of user-specific information are requested, although some of these items involve sub-questions. For example, after the question “do you plan to drill new wells?” come questions on when, how many, and pumping capacity of each future well.

Table 2.

Water recycling investment online decision tool (Rutgers University, 2021) user data inputs and available data outputs.

View Table

From the 17 pieces of information requested, a site-specific water recycling system can be costed out. This is done using a bottom-up, engineering-style estimation approach. For example, the number of feet of pipe is multiplied by the per-foot cost of a single type of pipe. The electricity required to run pumps is estimated based on the horsepower of each pump-motor combination, which is sized to deal with the expected flow in gallons per minute from a source of standing water, as well as the energy required to cover a distance and overcome the effects of gravity [see Ferraro (2015) for an approach of this type]. The estimation algorithms and numerical constants embedded inside the tool are too numerous and complex to be detailed here: Please contact the authors for more information.

Data sources

A logical first step in a project like this is to aggregate published findings across all recycling operations that have been surveyed, thus creating a larger sample than exists in each article. This data approach is known as a meta study. A meta study sample was created by combining data on individual operations reported in the following published works: DeVincentis et al., 2015; Ferraro et al., 2017; Mangiafico et al., 2008; Pitton et al., 2018; and Raudales et al., 2017. The resulting combined dataset included 39 distinct variables across 38 operations. We inflated or deflated all dollar figures in this dataset to 2016 dollars. We easily estimated some variables not directly reported in a study to achieve a better match across studies. For example, if a study includes annual debt service as its measure of capital cost, then total upfront capital outlay can be backed out, given an interest rate.

We included Ferraro’s (2015) two synthetic nurseries in the meta study sample. Two studies, DeVincentis et al. (2015) and Raudales et al. (2017), included some greenhouses, which are also included among the meta study observations. The primary challenge with the meta study dataset is that each author selected different variables for financial analysis. For most variables, there is too little overlap across studies to generate sample sizes as high as 20, let alone 38. Table 3 shows a set of variables ranked by the number of studies in the literature that include them. The total capital cost for building a water recycling system is widely available across published studies, yet a full 40% of the meta study sample lacks information on the most obvious variable one would want to correlate with this cost measure to build a tool that provides customized estimates. This variable is operation size, whether measured by acreage or by irrigation flow. Turn to individual cost components, like the cost of a filtration system or even annual operating expenses, and the meta study sample size falls into the mid teens.

Table 3.

Number of operations for which data were reported, by variable, in a combined sample of 38 nursery and greenhouse operations drawn from the following studies: DeVincentis et al. (2015), Ferraro et al. (2017), Mangiafico et al. (2008), Pitton et al. (2018), and Raudales et al. (2017).

View Table

We converted the meta study sample data matrix for use in the SAS statistical program (SAS Institute, Cary, NC), although few relationships among its variables are likely to be statistically reliable. The dataset remains essential for model validation, however, because it provides the only real-world costs and benefits to which model outputs can be compared. The sources of data required for the tool’s bottom-up costing approach include individual studies such as Ferraro (2015) and a set of cost multipliers that are used by USDA-NRCS (1997, 2015, 2019).

Model validation approach

A series of simulation runs using the computer tool are presented as follows. One approach to validating the computer model would be to see if these simulated results are similar to those calculated using the meta study data. Although we argued previously that operations profiled in the published literature might be a nonrepresentative sample, this comparison remains a useful exercise. For example, profit is more likely than cost to exhibit bias in the meta study sample, but we do not have usable data on profit (see Table 3). Also, to the extent that the computer model builds its results using standard engineering multipliers rather than experiences reported from the field, one can argue that the model validates the existing literature rather than the other way around. Direct comparisons between meta study results and those drawn from the simulations should therefore be thought of as cross-validation, not as model validation in the traditional sense.

(There is one way in which the simulations presented in the following pages were designed explicitly to match the meta study data to which they are compared: The irrigation intensity of 500,000 gal/acre per year in Supplemental Table 1 was taken from the meta study sample. In the computer model, only green marketing benefits are likely to be affected by this input parameter. There is also something to be said for matching irrigation intensity to the literature to do an “apples to apples” comparison when cross-validating findings on more fundamental relationships, as in Fig. 1. Note that irrigation intensity is always customized when individual users enter their acreage and flow data separately online.)

View Full Size

Water recycling system capital cost for nurseries as a function of annual irrigation flow and pond sizing method. The figure compares simulation results from the Rutgers water recycling investment online decision tool (Rutgers University, 2021) to relationships estimated using a meta study sample created by combining data on individual operations reported in DeVincentis et al. (2015), Ferraro et al. (2017), Mangiafico et al. (2008), Pitton et al. (2018), and Raudales et al. (2017). In the simulations, small pond capacity is calculated as 3-d peak irrigation flow (gal/h × 18 h/d × 3) and large pond capacity is calculated as 30% of annual irrigation flow. Input data for the simulations may be found in the second column of Supplemental Table 1. The parameters of the linear functions for the meta study best-fit lines may be found in Table 4; 1 gal = 3.7854 L.

Citation: HortTechnology 32, 1; 10.21273/HORTTECH04925-21

• Download Figure
• Download figure as PowerPoint slide

A slightly different validation approach would be to use program simulations to confirm or reject specific findings reported in the literature (this, too, is cross-validation, not one-way validation for the reasons stated previously). For example, the literature has shed light on the question of which factors are most important for profitability, and which ones are so minor that they can safely be ignored. The simulations reported in the following pages focus on several operational attributes that have already been found to be important to ROI: 1) the degree of required earth moving—in our case, pond construction (Ferraro et al., 2017), 2) the use of expensive public water before the investment is undertaken (Pitton et al., 2018), and 3) the opportunity to leverage water recycling into higher profits using green marketing (Cao et al., 2017). An important background factor for evaluating these issues, as well as many others, is operation size. Many of the results reported in the following use operation size as a second measure that is varied on the same line graph where factors 1), 2), or 3) are varied.

Programming platforms

We used spreadsheet software (Excel; Microsoft Corp., Redmond, WA) to program a set of algorithms that convert user inputs into financial outputs. We used a diagramming and vector graphic software application (Visio; Microsoft Corp.) to flowchart the user questionnaire, which involves several complicated skip sequences. The online tool programmed in JavaScript is available online (Rutgers University, 2021). JavaScript handles both the communication and calculation aspects of the interactive online tool. We completed multiple runs for simulation and debugging purposes in both JavaScript and Excel. We incorporated feedback from commercial nurseries to improve the model.

Overall model validation, with a focus on pond construction

Because of the data availability problem illustrated in Table 3, only one relationship simulated by the computer tool is compared with the same relationship calculated using the meta study data. This is the relationship between size of operation, measured by total annual irrigation flow, and total capital cost for the construction of a recycling system. Figure 1 shows this comparison between the computer model’s simulation results and the meta study data. The simulations in Fig. 1 also vary assumptions on tailwater pond size, providing additional insights into the importance of earth-moving costs.

Figure 1 includes two trend lines from the simulation program, as well as three data series drawn from the meta study data, with operation size varied on the horizontal axis. The two simulated trend lines differ only in their assumptions on tailwater pond sizing. The more steeply sloped line in Fig. 1 is labeled “large pond simulation” because it forces construction of a relatively large tailwater pond. An earlier version of the computer tool estimated pond size by selecting the larger size out of two methods. The first method was based on a regression analysis of reported data on pond size and total acreage in Ferraro (2015). The second was based on a published recommended minimum pond size of 30% of annual irrigation flow (Beardsell and Bankier, 1996). Using an assumption of 500,000 gal/acre per year—a number that is just below the median of all the meta study operations for which we were able to calculate this measure—both methods estimate roughly the same size pond for a given operation size. These “large pond” estimates are, however, generally larger than what is required for recycling.

We ultimately chose an alternative method for use in the online tool. It is similar to what USDA-NRCS uses for many of its own field estimates. The USDA-NRCS method recommends a pond size for recycling that is equal, in gallons, to 2-d peak irrigation flow. The computer tool increases this figure to 3-d peak flow, allowing for greater operational flexibility and providing a slightly better match to the diversity of pond sizes observed in the literature. Pond construction costs based on 3-d peak irrigation flow are shown in Fig. 1 with the label “Small pond simulation.”

Twenty-one operations in the meta study dataset include information on both annual irrigation flow and total capital cost. Two outliers were removed, bringing the sample down to 19. One outlier had annual irrigation flow 3.6 standard deviations above the mean; the remaining observations are within a range of ± 1.7 standard deviations. A second outlier had total recycling capital costs 4.0 standard deviations above the mean, while all remaining observations are ± 1.2 standard deviations. Figure 1 shows the actual datapoints from the meta study data, although the four largest operations of the 19 are omitted for reasons of scale/readability.

All 19 operations were included in a regression analysis of the meta study datapoints. Linear regression results are shown in Table 4; two linear regression lines based on this table are shown in Fig. 1. The upper line in the figure was fit with a nonzero intercept. This is a reasonable way to model those capital costs that are fixed across all operations, like those related to engineering and permitting (e.g., Fig. 4a in Mangiafico, 2008).

Table 4.

Ordinary least squares regression results for the relationship between total irrigation flow and total water recycling system capital cost, for a combined sample of nursery and greenhouse operations drawn from the following studies: DeVincentis et al. (2015), Ferraro et al. (2017), Mangiafico et al. (2008), Pitton et al. (2018), and Raudales et al. (2017). Best-fit lines corresponding to the parameters below are identified in Fig. 1 by the phrase “Metastudy best fit.”

View Table

The resulting intercept of $490,000 in Fig. 1 appears too large to represent only capital costs that do not vary with operation size. Therefore, we ran a second regression on the meta study data with the intercept constrained to zero: the result is shown in Fig. 1. It seems likely that a much larger sample of real-world data on nursery and greenhouse operations would fit a regression line somewhere between the two meta study trendlines shown in Fig. 1.

We consider the magnitude of capital costs that are independent of operation size, shown as intercepts in all of the Fig. 1 trendlines, to be an open question: few studies publish data on engineering and permitting costs. The similarity of the slopes of the meta study trendlines and the small pond simulation in Fig. 1, however, helps validate the latter. According to the best-fit lines, marginal construction cost per unit of irrigation flow in the meta study sample is virtually identical to that calculated in the small pond simulation, but not in the large pond simulation.

Comparing the two simulations with each other, Fig. 1 suggests that alternative assumptions about pond size—all of which are reasonable choices for growers—can have significant effects on overall capital cost, as well as on the marginal financial impact of size (for a 50-acre operation using an irrigation volume of 25 million gal/year, the small pond simulation in Fig. 1 finds that pond construction accounts for more than 90% of total capital costs). Thus, these simulation results confirm the statement of Ferraro et al. (2017) that “the largest cost items for implementing capture and recycling are regrading the production area and digging a recycling pond.” At the same time, Fig. 1 highlights the role played by alternative minimum size standards for recapture ponds. Although that subject is opaque in the case study literature, it is a critical feature of the forward-looking engineering approach taken by USDA-NRCS and by the current online tool.

Taken in its entirety, Fig. 1 shows that the online tool generates capital costs as a function of operation size that are in the same range as the best available real-world data. The small pond assumption tends to generate costs near the lower end of the meta study range. It could be that operations sampled in the existing literature have done significant regrading, which the current tool ignores, or they have opted for tailwater ponds significantly larger than the USDA-NRCS standard. If that is true, then widespread use of this tool could help steer growers toward more economically sized ponds and basins, which can be expected to increase ROI, other things equal.

However, there is more to consider in the selection of pond size than ROI alone. Growers and managers need to consider the potential for very dynamic and extreme changes in precipitation due to climate change, which could demand larger ponds holding volumes beyond 3 d of irrigation water, to account for longer regional or seasonal droughts, and retain larger stormwater runoff during periods of high or prolonged precipitation. Larger ponds could also be used to help reduce (dilute) pollutant concentrations in regions with regulatory oversight, more so if stormwater runoff volumes are channeled to larger ponds. Growers and managers need to carefully consider their local/regional environmental regulations and any available research-based recommendations for pond size in their locations.

The remaining simulations used to cross-validate the model report net present value (NPV), a direct measure of profitability. Factors that the published literature has found to be important to profitability are systematically varied in the accompanying figures.

Public water rate

Figure 2 reports simulated NPV results for operations of four different sizes: 1 acre and 500,000 gal/year, 10 acres and 5 million gal/year, 20 acres and 10 million gal/year, 50 acres and 25 million gal/year. All scenarios assume a chlorine disinfection system and have enough vacant land on which to construct a tailwater pond, so no profits are lost due to land taken out of production. Before the recycling system is installed, each operation draws 100% of its irrigation water from a municipal water system. A maximum of 50% of the public irrigation water can be replaced by recycled water after the recycling capital equipment is installed. This is a slightly optimistic ceiling based on Table 7–1 in USDA-NRCS (1997).

View Full Size

Net present value of a nursery water recycling investment as a function of the price of public water and operation size, simulated using the Rutgers water recycling investment online decision tool (Rutgers University, 2021). Input data for the simulations may be found in the third column of Supplemental Table 1. Annual irrigation flow is standardized at 500,000 gal/acre (4677.0 m³·ha⁻¹) × total production acres; 1 acre = 0.4047 ha, $1.00/1000 gal = $0.2642/m³.

Citation: HortTechnology 32, 1; 10.21273/HORTTECH04925-21

• Download Figure
• Download figure as PowerPoint slide

Figure 2 indicates that NPV of the recycling investment is negative at a public water price of zero, which is equivalent to no public water being used. It is difficult for a recycling investment to be profitable in partial budget terms if irrigation water is currently available at minimal cost, such as pumping from a well. The recycling of fertilizer is virtually the only operational saving that offsets the significant upfront capital cost. However, for a 20-acre operation, a public water price of $0.98 per 1000 gal makes the recycling investment break-even, whereas higher public water prices make the investment even more profitable. For a 10-acre operation, the break-even public water price is $1.41 per 1000 gal. For a 1-acre operation, the break-even water price is more than $9.00, which is unusually high. A possible option for small producers in close proximity to each other is a cooperatively run recycling system.

Several published studies report a break-even point for recycled vs. public water, although other attributes of the profiled operations are not necessarily comparable to those underlying the simulations in Fig. 2 (see the list of model inputs in Supplemental Table 1). The study by Pitton et al. (2018) of a 420-acre operation in California, which uses a mix of public and well water, found that recycling was easily profitable at municipal water rates between $2.00 and $3.00 per 1000 gal. The break-even analysis of Ferraro et al. (2017) takes one 27-acre nursery for which municipal water is currently the cheapest option and estimates that doubling its cost of public water would make a recycling investment profitable. This increase in public water cost is described as “substantial but not inconceivable.” Using recycling cost data from a small sample of operations that vary in size, DeVincentis et al. (2015) report that recycling is not currently the most profitable choice, as public water prices are between $.80 and $1.60 per 1000 gal. Yet these same authors felt that upward trends in water prices could easily tip the balance in the near future.

Compared with the published case studies, the simulations in Fig. 2 have the considerable advantage of holding other things equal while varying only two key variables. They show that the greater the amount of public water currently used, the easier it is for savings from recycled water to offset the fixed upfront investment that is required; profitability becomes a matter of volume (Pitton et al., 2018). We conclude that all but the smallest operations, if they purchase a significant portion of their irrigation water from a public system and have no opportunity to dig new wells, should consider installing a water recycling system. Such operations could enter their current public water price into the online tool for a customized estimate of the recycling system ROI.

Profitability from green marketing

The online computer tool includes an option to simulate the existence of a green marketing program. If the user selects this option, a profit of one penny for each plant sold is added to the partial budget’s bottom line. This figure is net of the marketing program’s fixed costs, which could be considerable. Note that $0.01 of profit per plant is ≈1/100th of the willingness-to-pay premium reported in Hartter (2012), so it is a conservative estimate. Assuming 40,000 plants/acre per year, Fig. 3 shows the NPV of the recycling investment, both with and without green marketing. As in Fig. 1, size of operation increases from left to right.

View Full Size

Net present value of a nursery water recycling investment as a function of the existence of a green marketing program and operation size, simulated using the Rutgers simulation model (Rutgers University, 2021). Green marketing is assumed to generate a net profit of $.01/plant under the assumption that the operation has a density of 40,000 plants/acre (98,842.2 plants/ha). Input data for the simulations may be found in the fourth column of Supplemental Table 1. Annual irrigation flow is standardized at 500,000 gal/acre (4677.0 m³·ha⁻¹) × total production acres; 1 acre = 0.4047 ha.

Citation: HortTechnology 32, 1; 10.21273/HORTTECH04925-21

• Download Figure
• Download figure as PowerPoint slide

The result from this simulation is similar to the problem of using public water. Green marketing is a volume game: the extra profit is earned on each plant, whereas many of the costs are fixed. (Due to fixed costs, the marginal profit of green marketing and labeling the 10,000th plant each year would be larger than that for the first plant, which would necessarily incur a loss. For simplicity, the online tool assumes a fixed but small marketing-related profit on each plant.)

The Fig. 3 result aligns with the finding of Cao et al. (2017) that “in almost all cases for which at least a portion of a retail consumer premium was returned to growers, the premium was adequate to compensate for recycling investment costs.” In other words, assuming that the price premium identified by Hartter (2012) is accurate and sustainable, the decision to engage in green marketing can be the difference between a profitable recycling investment and an unprofitable one. Because it depends on consumers’ willingness-to-pay for a product they perceive as special, the expected price premium could diminish if sustainability practices become more widespread in the industry. Issues of industry-wide certification and consumer trust also must be addressed (Cao et al., 2017).

Figure 3 also shows that in the absence of volume-related cost savings or revenues, increased nursery size can increase the size of private losses attributable to a recycling investment. In such cases, the upfront investment cost is proportional to operation size (Fig. 1), whereas net operating benefits are modest or even negative (e.g., maintenance of the recycling system; need to purchase disinfectant).

Additional simulations

Two additional simulation runs are reported here without graphic illustration. One set of simulations found that avoided well-drilling costs (DeVincentis et al., 2015) are not likely to be a huge factor in profitability. It takes a relatively large quantity of recycled water to displace an average-sized future well. A discount factor is also applied to this avoided capital expenditure, which is modeled as a one-time cash outflow of $30,000.

A second analysis explored the possibility that a grower would need to take land out of production to construct a tailwater pond, an issue that is also addressed in Ferraro et al. (2017). In a set of simulations that estimated negative NPV for a 100-acre operation, an assumption of $1000 profit per acre for land taken out of production increased the NPV loss by about 8%. An assumption of $2500 profit per acre increased the NPV loss by 20%. These results confirm the conclusion of Ferraro et al. (2017) that land taken out of production can be a “principal cost factor.” Both the earlier study’s results and our own emphasize the importance of minimizing the size of the recapture basin if at all possible, and this is even more important if profitable land must be taken out of production to construct a pond. Margins per acre can vary widely across the country. The model allows producers to input their individual information to give results for their operation.