This report was prepared by RFF researcher Henry M. Peskin, with the assistance of Daniel J. Basta and Leonard P. Gianessi. The data bases and models described were developed as part of a larger project to expand the U.S. National Income and Product Accounts sponsored by the National Science Foundation. The water quality transport relationships used in the models were provided by Kenneth Young of GKY and Associates, Inc.
While there is no question that the nation's efforts to reduce water pollution are producing successes, concerns about inflation have raised questions about cost, efficiency, and fairness. If, for example, programs could be directed toward the more polluted water bodies and away from those that already are clean, there may be substantial savings in cost without any reduction in effectiveness. A similar result may be produced by focusing more on industries whose plants tend to be located on those rivers and lakes that have poor assimilative capacity. However, the fact that certain regions and industries may suffer a heavier burden of cost and employment effects cannot be ignored if policies are to be accepted as equitable.
Unfortunately, attempts to find an acceptable balance between considerations of efficiency and equity are hampered by a poor understanding of the likely effects of national policies on particular dischargers and on ambient water quality. Models exist that permit the analysis of the discharge-water quality linkage, but only for a few selected water bodies. It is generally accepted that applying such models to the nation as a whole would be prohibitively expensive since these models tend to be very complex, are difficult to calibrate, and make substantial data demands.
However, by simplifying the structure of the models (with an admitted sacrifice of detail) and by using readily available data sets, researchers at Resources for the Future, supported by a grant from the National Science Foundation, have developed a framework that for the first time links pollution generation activities in each of the nation's counties to a detailed network of 304 rivers, 175 lakes and reservoirs, and 37 bays. Through simple mathematical transport relationships, this framework is able to relate water quality at one location to pollution that may originate several hundred miles upstream. The ability of lakes and reservoirs to trap and otherwise eliminate pollutants also is explicitly accounted for.
River reaches and lake and bay segments have been further delineated by 1,306 nodal points. Depending on drainage patterns, each of the nation's 3,209 counties is assigned to at least one, and sometimes more than one, nodal point. Industrial, municipal, agricultural, and other nonpoint pollution sources are grouped by county. After estimating the amount of pollutants discharged by these sources and how this amount is affected by policy, the discharge—water quality linkage can be completed.
By adopting standard data classification schemes, the system is able to draw on the vast array of industrial, population, agricultural, pollution, and hydrologic data files maintained by such federal agencies as the Department of Commerce, the Department of the Interior, the Department of Agriculture, and the Environmental Protection Agency. The entire process can be grouped into three major steps: determination of residual discharges from both point and nonpoint sources as affected by policy; allocation of these discharges to counties; and insertion of the pollutants into the water network, with computation of resulting ambient concentrations.
For many industries, the data permit detailed plant-by-plant analysis of discharges under such specific control assumptions as Best Practicable Technology (BPT) and Best Available Technology (BAT). However, for other industries, additional data sets such as the Census Public Use Files are used to estimate county-by-count) discharges.
Illustrative results. The ability to link even crudely, polluting activities with the quality of the nation's waters permits many investigations other that the search for more cost-effective and equitable programs alluded to earlier. This flexibility can be demonstrated by the data in the table, showing simulation of the percent of the network's river miles that are polluted with Biochemical Oxygen Demand (BOD) and Total Kjeldahl Nitrogen (TKN). Three situations are depicted: a 1972-world with the current water pollution control policies; a 1972 world with the Best Practicable Technology (BPT) regulations mandated by the 1972 Water Pollution Control Act Amendments; and a 1972 world with the Best Available (Technology (BAT) regulation in place. For purposes of presentation the results are aggregated by water resource region.
These data not only give a rough indication of where pollution problems exist but also, and perhaps more significantly, show that these problem locations critically depend on which residual is found and on hydrologic conditions. Such information has obvious importance for the design of monitoring networks. The data also illustrate the likely differential impacts of policy. For example, in certain regions (for example, the Lower Colorado) a move from BPT to the more stringent BAT has virtually no effect on BOD, and TKN. However, a similar move has a substantial effect in other regions (for example, the Upper Mississippi).
As a final example of the use of the data, simple efficiency analyses can be undertaken. The total incremental cost of the simulated BPT policy is about $14.8 billion annually and for BAT, an additional $9.2 billion. Dividing these numbers into the reduction in polluted river miles under BPT and BAT, respectively, yields ratios of miles of additional improved river per billion dollars of additional expenditure under various flow conditions. It may be efficient to redirect expenditures towards those policies with the higher ratios, assuming, of course, that the objective of policy is simply to reduce polluted river miles (regardless of which pollutant is causing the problem or how serious the pollution is).
Again, the results of this exercise seem to depend on the pollutant in question and on the flow conditions. For example, under average flow conditions the BPT ratio of improved river miles per billion dollars exceeds the BAT ratio for TKN (151 vs. 133), but under low flow conditions the BPT ratio is less than the BAT ratio (367 vs. 383). For BOD the BPT ratio exceeds that for BAT for all flow conditions (107 vs. 59 for average flow and 358 vs. 99 for low flow). Thus, before one concludes that BPT is more or less efficient than BAT, it appears that one must first specify both the pollutant (or pollutants) and the flow conditions that are the targets of the policy.
Of course, none of these analyses are of much value if the estimated pollution concentrations are highly inaccurate. Although there are no standard measures of accuracy for these sorts of estimates, we can get a qualitative "feel" for the performance of the model by comparing "base year" (1972) concentration estimates with recorded observations taken in approximately the same period of time. Fortunately, about 10 percent of the network's nodes correspond to official monitoring stations where actual measurements are made for one or more of the eight pollutants estimated in the system.
While this verification effort has yet to be completed, certain preliminary conclusions can be drawn. There appears to be a tendency for the system to overestimate concentrations, partly because the relatively small number of nodes in the system implies that loadings will be too heavy at single geographical points and partly because certain pollutants (such as dissolved solids) are assumed conserved throughout the network even though in reality they are not totally conserved. However, the trend of concentration point-to-point seems to be estimated fairly well.
A profile of BOD; concentration in the Ohio River illustrates some of these conclusions. There may be a tendency for overestimation in the Pittsburgh area, yet the model seems to predict the general trend in pollution. However, the profile also illustrates certain difficulties in making comparisons.
The major difficulty is that the model predicts best and worst conditions in a particular year while the data describe median conditions over a range of years. As a result, the flow conditions at a particular point may not be comparable between the measured data and the model. Thus, it is quite possible that in 1972 the BOD; reading in Pittsburgh actually approached the predicted 10-12 range on the day with the lowest flow. Unfortunately, for many rivers—the Ohio included—actual daily readings are not available.
Perhaps one result of these verification exercises will be to stimulate more data collection in potentially polluted areas as predicted by the model.
Uses of the framework. At present, the framework is being used for two investigations: a comparison of regional and family differences of the benefit and cost burdens of the Water Pollution Control Act of 1972; and an investigation of those counties that may be in the greatest need of agricultural sediment control programs due to potential agricultural contributions to poor water quality.
Although current applications are directed toward the analysis of conventional water pollutants, the fact that the framework describes industrial activity at the county level of detail enables it to be used as a device for aiding in the analysis of other residuals as well. In particular, the framework can provide information on regional and industrial discharges crucial to the investigation of problems associated with toxic chemicals, solid wastes, and air pollutants. While the framework can never fully substitute for detailed river basin analyses, it does serve to identify potential problem areas. It is a useful instrument for setting priorities for policy planning, management, and the expenditure of funds.
A version of this article appeared in print in the September 1979 issue of Resources magazine.
