By Susan M. Ernsdorff

Single-Family Collection-Cost Model
Multifamily/Commercial Collection Model
Using the Results
Notes/References

Seattle Public Utilities (SPU) in Seattle, WA, developed two spreadsheet models to estimate collection costs. The models consider variations in collection frequency, truck types, material separation requirements, transfer points, and more. These models were used by SPU as part of an effort to estimate total solid waste system costs, considering collection all the way through disposal or processing, in order to evaluate alternative future visions for Seattle’s solid waste system. They were also used to determine the variables having the most impact on cost and the relative cost impacts of changes in parameters. This paper describes the collection cost–estimation models for both single- and multifamily residential units.

Included are descriptions of the input data used, the model methodologies, the outputs obtained, and a discussion of how the results can be used. Increasingly, municipalities will be called upon to make difficult choices in balancing the level of solid waste services provided and the cost of those services. Models such as these will be helpful in characterizing the options.

The types of collection alternatives modeled by SPU include:

Single-Family

• Co-collection and pod trucks

• Cost impacts of different collection frequencies

• Foodwaste collection alternatives

• Recycling program variations, including mandatory separation

• Alternative transfer locations

Multifamily and Commercial

• Alternative transfer locations

• Recycling program variations, including mandatory separation

The collection cost–estimation spreadsheets developed are also valuable for determining the factors that are most critical to collection efficiency and cost minimization and for exploring the cost impacts of varying specific collection parameters.

Single-Family Collection-Cost Model

Overview of Model

The single-family collection-cost model estimates the cost of collecting materials (recyclables, yardwaste, and garbage) from residential customers who use household containers (cans, bins, or bundles). A wide variety of collection strategies can be hypothesized and modeled. The spreadsheet allows the user to estimate the approximate total annual costs for a particular scenario and to explore the cost impacts of varying individual parameters within that scenario. Barbara Stevens of Ecodata Inc. developed the basic structure of the spreadsheet.

A scenario was defined by the recycling programs it included, by the particular collection methods used, and by the facilities to which materials were sent. For example, the current household scenario for the City of Seattle would include the collection of yardwaste, recyclables, and garbage, with separation of yardwaste being mandatory and the separation of recyclables being voluntary. A dedicated fleet of collection trucks collects each material stream, and materials are directed to two city-owned transfer stations and two private stations. Another scenario might have mandatory separation of recyclables, and yet another might add the collection of foodwaste, with participation being voluntary. Materials may be co-collected, and a different network of transfer stations may be used.

Each "run" of the model requires a complete set of input assumptions, as described in detail in the section below, that characterize a collection scenario. The model allows the user to specify the number of collection fleets and types of trucks, which materials are commingled, the number of compartments on the trucks, the frequency of collection, and the destination of the trucks. The model calculates the number of crews required in each fleet and the total annual costs of each fleet. A fleet of trucks may collect one or multiple material streams—garbage alone, for example, or garbage and recyclables co-collected (see Note 1). The collection cost per ton and per household is calculated for each fleet of trucks and for the collection system as a whole. This overview of the model is shown in Figure 1.

Collection-Model Input Parameters

Below is a list of the parameters used, with indications as to how the appropriate input values were determined for the specific modeling done for Seattle. All of the inputs can be varied, although in practice many of the parameters are constant from run to run.

Material Tonnage, Composition, and Densities. This parameter describes the total tons that will be included in the collection/ transfer/disposal system for a given scenario, broken down into the different streams for collection—garbage, recycling, yardwaste, foodwaste, and/or special collections, as appropriate. Separate waste-generation and recycling forecast models were used to predict the total tons to be recycled and disposed of in Seattle in 2010 (see Note 2). These values were input into the collection model. The forecast models take into account many contributing variables, such as population and household counts, economic activity, and number and size of businesses. Total generation would vary depending on waste-reduction efforts, and how the tonnage was distributed between the garbage and collection streams would vary depending on the recycling collection programs included. For instance, if the scenario being considered includes extensive grasscycling promotion, there will be fewer tons of yardwaste and fewer overall tons. If the scenario being considered proposes new material to be collected in the curbside recycling program, the total tons will be the same, but tons will shift from the garbage stream to the recycling stream. Tonnage and composition data are entered into the model at three levels: the total tonnage generated, a breakdown of the total into percent composition in 20 categories, and the percent recycled of each of those 20 materials. The collection model requires density factors for each of the 20 material categories (lb./yd.3) for both uncompacted material and compacted material. This allows conversion of tons to cubic yards and is a key determinant in calculating truck capacities. Ecodata furnished the input data on material densities.

Household-Level Specifications. The total number of households to be served must be specified, along with participation rates for whichever collection programs are included in the scenario being modeled. Participation is described at two levels: what percent of households choose to receive the service and, of those participating households, what percent set out material for collection at each opportunity. For modeling Seattle programs, the garbage-collection participation rate is 100% (since everyone by law is required to have curbside collection), but the setout rate is less—about 90%. For Seattle’s current weekly curbside program collecting from bins, the participation rate is 90% and the setout rate is 70%. For new programs being modeled, participation and setout rates are based on results of pilot programs and/or responses to surveys, or on extrapolation from current program participation rates.

Collection-Route Parameters. In order to calculate how many collection crews are required, there are several key collection-route parameters that must be specified:

• Travel time, the time to drive from setout to setout (stop to stop), is impacted by household participation and setout assumptions. The lower the setout rate, the greater the distance between each household needing collection and, therefore, the greater the travel time. The average travel time for Seattle’s collection program ranges from eight to 10 seconds.

• Stop time, the time spent at each stop placing material in the truck and returning containers, is affected by the truck type (especially semiautomated versus manual) and the number of containers. Stop-time estimates for the scenarios modeled by SPU varied greatly, from 14 to 55 seconds.

• Collection frequency, which is how many times a household receives each collection service per year, varies from weekly to every other week to monthly for Seattle residential programs.

For SPU’s modeling, route-time input numbers were derived from actual data collected en route for different types of collection trucks. When actual data were not available, judgement was used to develop a set of stop and travel times that were internally consistent with each other.

Truck Types. There are five truck types specified in the base model. These types are listed in Table 1.

Multifamily/Commercial Collection Model

How the Model Works

The model performs a series of calculations similar to that done in the single-family model. Using annual tons collected, numbers of accounts served, and average collection frequency, the average weight picked up at each account is calculated along with the number of stops made per year. The number of stops needed to fill a truck is determined, as is the time required for a crew to collect one full load. The load time is the sum of the necessary number of individual-account stop times and travel times to give a full load, plus the round-trip cycle time to dump that load at a transfer station. The model then determines how many full and partial loads each crew can complete in their daily available collection time and, from that, the total daily number of accounts that are served by one crew. From this, the model calculates how many crews are needed to collect from all of the accounts.

Collection-Model Input Parameters

The estimated collection cost for a given scenario depends on many of the same basic factors described for the single-family model:

• Tons collected

• Number of participating accounts

• Collection frequency

• Estimated stop, travel, and tip times

• Truck capacity

• Crew cost and time available for collection

Material Tons. This is the total annual tons to be collected by a fleet. The models do not include the specific composition of each of the streams. Because there are fewer opportunities for changes in these collection systems (as compared to the single-family system), less complexity of the model was needed.

Participating Accounts. The base number of accounts in each program is entered, and it is assumed that all accounts receive the collection service at every opportunity. These customers do not set out containers at the curb; the collector goes to the account location and empties whatever is in the container.

One additional factor that was considered here for the commercial sector is that one account often consists of several businesses. Many businesses share the same building or adjacent buildings and therefore share garbage service. For future projections, it was assumed that the current statistic of 3.3 businesses per average garbage account would continue. When modeling commercial foodwaste collection, fewer businesses per account were assumed, because it is not likely that foodwaste-generating businesses will be as clustered as businesses in general.

Collection Frequency. The minimum garbage-collection frequency allowable (by law) is weekly. currently, multifamily and commercial customers have selected a service combination of container size and collection frequency that best suits their needs. Garbage is collected as frequently as daily for large generators with limited area for Dumpsters. Multifamily recycling-collection frequency ranges from weekly to monthly. For this modeling effort, the current average container sizes and service frequencies were used as a starting point (see Note 3). For scenarios that significantly reduced garbage tonnage, the collection frequency was reduced to give the same average setout weight, which implies the same container size. This assumes that accounts will reduce their frequency of service and keep their container size, as opposed to keeping their frequency of service and reducing their container size. For foodwaste collection, twice-a-week collection is assumed because of the extremely putrescible nature of the material. For multifamily recycling, the current standard of every-other-week service is used.

Stop, Travel, and Tip Times. The baseline status quo for drive, stop, and tip times was obtained empirically by timing collection crews in action and from scale-house transaction records. This baseline was modified for each run, depending on how participation, average setout weight, type of container, or number of containers per account differed from the status quo. GIS travel-time estimates, as described above, were used for new scenarios.

Truck Capacity. The current average full-truck weights were used. (This information came from transfer station scale-house transaction records). For garbage and foodwaste, a typical truck holds 8 tons; for recycling, it holds 6 tons.

Crew Cost and Time Available for Collection. Ecodata developed the crew-cost estimates used by SPU as part of earlier work done to assess opportunities for cost savings (Stevens, 1994 and 1997). The daily crew time available for collection must take into account time allowed for other routine activities: equipment checks, breaks, wash-up, and the time to drive from and back to the base of operations at the beginning and end of each day.

Collection-Cost Model Output

• Number of crews required: This is the number of crews needed, on average, to provide the indicated collection services to the designated number of accounts.

• Annual collection cost: The annual total cost of providing the specified collection service to all the accounts is equal to the number of crews required multiplied by the annual crew cost.

• Cost per ton: Each spreadsheet calculates this output for its customer/material combination. It is equal to the total annual collection cost divided by the annual tons collected.

Using the Results

Model results can be used in two ways: in an absolute sense, meaning as actual estimates of costs, or in a relative sense, meaning to compare the results of various scenarios with each other.

In order that the model would provide the most accurate cost estimates possible, with the intent of their being valid in the absolute sense, SPU "calibrated" the model inputs for the status quo single-family collection. By calibration, it is meant that initial estimates or approximations of the various input parameters (described above) were incrementally adjusted as necessary until the "crews required" output of the model matched the number of crews actually used. These calibrated inputs were then used as the baseline.

SPU used the collection-cost models to:

• Estimate the total costs of collection-service packages.

• Explore the impacts of isolated changes to a specific collection program, such as adding new materials to the recycling program, changing yardwaste collection frequency, changing container and sorting requirements, or utilizing new tip locations.

• Investigate the impacts of participation rates on cost.

• Estimate costs for completely new programs.

• Explore collection parameters to identify the greatest opportunities for cost savings.

For comparing program costs, it is critical that the inputs for the various scenarios be internally consistent, meaning that they make sense compared to each other. For example, the "drive time" input would necessarily have to be less for a collection scenario with a greater stop density as compared to another scenario, and the "stop time" input would be less for a collection scenario requiring the pickup of one container at each stop as compared to a scenario requiring the pickup of two containers.

For example, one scenario that SPU considered was co-collection of foodwaste and garbage in a dual packer truck. This would entail the collection of a small container of foodwaste along with the standard container of garbage. SPU had conducted a foodwaste-collection pilot and from this had estimates for participation rates and setout weights. However, a measure of the stop time required to empty two containers was not available. An estimate for this was made, based on the status quo time to empty a single garbage container, with a small additional time increment added for the second container.

When exploring new program concepts, there is necessarily some uncertainty in the results. SPU employed sensitivity analyses to determine the extent to which model results would change when considering reasonable ranges of values for the less certain inputs. By sensitivity analysis, it is meant that a single input variable for one scenario is varied while all others are kept constant, and the resulting range of output is recorded. For example, for Seattle, the status quo average household stop time to pick up one container of garbage is 18 seconds. For the scenario requiring pickup of two containers, presented above, a stop time of 22 seconds was estimated and a sensitivity analysis was done for the range of 20 to 25 seconds. The collection-cost estimate of $83.51/ton (for 22 seconds) varied from $80.55 to $91.82 for the range of stop time considered. These numbers are presented to illustrate the use of sensitivity analysis. The base factors will be unique for each jurisdiction, and accurate estimation of these is critical.

SPU considered the range of results from the sensitivity analyses in determining at what level cost differences were significant. The user of the model may also utilize sensitivity analyses to determine which variables are particularly influential on the results and direct additional resources toward improving the quality of data for these inputs.

Notes

1. A "material stream" is used here to identify a grouping of materials disposed of as a single wastestream, of which this modeling work includes four: garbage, recyclables, yardwaste, and foodwaste. Yardwaste, for instance, may include the material yardwaste, as well as some vegetative foodwaste and compostable paper.

2. The year 2010 was selected to allow full implementation of new programs and to include near-term population and business growth when evaluating facility capacities.

3. current multifamily statistics were obtained from the utility’s billing system (number of accounts) and transfer station records (tons). Statistics on commercial customers were obtained from the service companies’ annual filings with the Washington State Utilities and Transportation Commission, as included in Stevens (1997).

References

Seattle Public Utilities, "Seattle’s Solid Waste Plan: On the Path to Sustainability," Draft Final Report, August 1998.

Seattle Public Utilities, "City of Seattle’s Recycling Potential Assessment/System Analysis Model, Basis for Seattle’s Solid Waste Plan: On the Path to Sustainability," 1999.

Stevens, B., "Analysis of Garbage and Yard Waste, Residential Recycling Contracts, Seattle, WA," 1994.

Stevens, B., Memo to Ray Hoffman, Seattle Public Utilities, "Commercial Contract Negotiation," June 16, 1997.

MSW