PENTERNATIONAL ATOMIC ENERGY AGENCY | Jc UNITED NATIONS EDUCATIONAL, SCIENTIFIC AND CULTURAL ORGANIZATION plieiiy INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS LCP, P.O. BOX 586, 34100 TRIESTE, ITALY, Casi) CENTRATOM TRIESTE. SMR.780 - 65 FOURTH AUTUMN COURSE ON MATHEMATICAL ECOLOGY (24 October - 11 November 1994) “Systems Analysis in Water Quality Management - A 25 Year Retrospect” Robert V. Thomann Environmental Engineering & Science Program Manhattan College Riverdale, NY 10471 USA. These are preliminary lecture notes, intended only for distribution to participants. Mar Bunome — Sraoe Caras, 11 Ta 22401 Taser 224163 Ta £60392 Micaorrocrmon Log Vin Bem, 31” Ta 224471 Tyarer 224400 Tarr 440302 Goer House Vis Grouse, Ta.224241 Tawar 228531 Taser 460489 far Hinses Vie Baar P Fo 92401 Taras 2240810 Te #40392Reprinted from SYSTEMS ANALYSIS IN WATER QUALITY MANAGEMENT Edited by M. B. BECK PERGAMON PRESS OXFORD - NEW YORK BEIJING - FRANKFURT - SAO PAULO » SYDNEY TOKYO TORONTO 1987SYSTEMS ANALYSIS IN WATER QUALITY MANAGEMENT— A 25 YEAR RETROSPECT Robert V. Thomann Environmental Engineering and Science — Civil Engineering Department, Manhaitan College, Bronx, New York 10471, U.S.A. ABSTRACT. A reflection is presented of the utility, credibility and application of vater quality sys- tens analysis techniques over the past several decades. The enphasis is on predictive water quality modela and the U.S. experience. The complexity of the water quality questions and agsociated aodeling has increased by orders of magnitude. Models of sedimen: interactions and effects of toxic substances are cructal to further development. Four criteria for judging performance and impact are discussed: usefulness, accuracy, serendipity and ownership. Models are widely used in decisions regarding alternative controls specifically co isprove cost effectiveness. The results of aystens techniques naed to be detailed in a wide ranging effort of post audit analysis following Implementation of environmental controls. Legisla~ thon and policies have incorporated, in a general way, the principles of water quality sys- tems analysis, with the notable exceptions of a widespread reliance on technology based and a general disregard of cost trade-offs using principles of opttaiza— tion. Te s concluded that the tapact of systens techniques has been broad and significant. Increased quality assurance of sodel formulation and calculation 1a necessary 0 ensure framevorks thet are rigorous and state of the art. A need exists for upgrading of undec- standing by users of water quality systems techniques and che time has arrived for a major world-wide effort to compile the econonic advantages of using systems techniques for nore tnforaed and efficient decteion making. KEYWORDS Water quality aodels; post audit; systeus analysis; criterie for tapact; mcde! usefulnes: model perforsance: dissolved oxygen; eutrophication. UNTRODUCTION My dictionary (Webster's, 1963) defines a Retrospect #8 "a teview of, or mediration upon ast events” as opposed to a Retrospective which 1s "a generally comprehensive exhibition showing the vork of an artist over a span of years." This paper is reflection on our past, seditative perhaps at tines, but not intended to be a comprehensive exhibition of our col” Tective artistic vorke. That would be a task vell beyond ny own competence. So vhat I have to offer in this retrospect are ay om thoughts, views, comprehensions and understanding of where we have been in the Last 25 years in this (sometimes perceived) arcane practice of systems analysts applied co water quality manegenent. To review our past, to evaluate the impact, or lack thereof, of systems analysis in water quality anagenent {2 essential, if ve are to offer the decision making community a atat of-the-art understanding of contenporary water quality issues. What effect hae our work had fon the decision making process? Does our work really aatter or are we juet talking andRV. THOMANN reporting to ourselves? hac are our "successes"? Our "failures"? Indeed, what are the criteria chat ve can even suggeot as useful for judging the significance of our work on the water pollution community at large? Finally, vige does the future hold and what are the ‘energing {ssues? NATURE OF SYSTEMS ANALYSIS IN RATER QUALITY MANAGEMENT In the present context, ue begin by offering a definition of systens analysis: "The engineering arc of integrating and aynthesizing che physical, chemical, biclogical and mathenaticai sciences wich the gocisl end economic sciences to construct franevorks that elucidate the consequences of alternative water quality and vater use objectives.” The princtpal components of this definition are: 1, Engineering art of integration: implying (a), a focus that is practical tn nature (che ‘engineeting), (>), a certain “flair” that tends ro be personalized and less than totally sctencifically rigorous (the art), and (c) a culling and rebuilding of key elements Of diverse disciplines (che integration). 2. Synthesis of che natural and mathenatical sciences with the social and economic sciences: implying that vhat ve do is more than mathematical modeling of natural systens and incorporates policy, economic, social and cultural issues into the analysis, 3. Elucidation of consequences of alternatives: implying thar water quality systens analysis has auch to say in the process of decision-making including revelation of pre- viously hidden behavior and formulation of new alternatives. Within these broad components, the key steps are: 1. Evaluation of the Problem a. Residuals input detersination bl Mathenatical aodel conetruction 1 Assessment of risk to human and ecosystem population without controls 4. Specification of a range of feastble vater quality/use objectives 2, Evaluation of alternative Controls ‘a, Determination of effectiveness of alternatives b. Optimal cost/benefit anaiysis 3. Dectston end Promlgation of Contro! Program a. Mater quality standard setting b. Determination of allousble risk €. Opeiaal control strategies 4, Implementation of Control Program ‘a. Haste load allocation bl Negotiation and tssuance of discharge peraic c. Monitoring of Program 5. Post-Audit of Progran a, Attainnane of vater quality standards b! Attainment of water use objectives fc, Evaluation of costs and benefits a. Predictive capability of sodel framework With no apologies for an obvious bias, at the heart of the entire sweep of these components and key steps is the construction of credible, defensible and predictively accurate math ematical aodeling {raneworks. Without such predictive capability, it {s sisply aot possible to develop a firsly based water quality management program. It is for this reason that auch of the effore in the past several decades has been in developing predictive mathenatical nodels of wacer quality at a variety of different levels of complexity. All of these models are aimed first at calculating the expected concentrations of vater quality vartables. ‘These concentrations then forn the basis for risk assessment to the aquatic ecosystem and to the public healeh. Thus there have been Intensive efforts in developing models that can be used with confidence in evaluation of alternative controls, cost/benefit analysis, risk Assesouent and optinal control strat I would Like to focus on this area of water quality models not to the exclusion of the socio-economic models (e.g. optimization of vater quality) but simply to eaphasize the central role chat predictive models play sn the decision-making process.Systems analysis in water quality management Historically, we have developed systens framevorks and nore specifically vater quality aodels for three broad classes of problen contexts: 1) Biochemicel oxygen demand (BOD)/ dissolved oxygen (D0), 2) Aquatic plants and nutrients, and 3) Toxic substances. Wiehin these contexts, attention has been variously placed on steady state and time variable deterministic frameworks to ensure credible inclusion of relevant mechanisns as vell as Ancorporation of uncertainty and probabilistic concepts to insure consideration of stochas- tic elenents in alternative evaluations. Models have grown from the tvo state variable BOD/DO modes to multi-state variable (e.g. 20) aodels of phytoplankton/ mutrtent models. Spatial decail has increased by orders of wagnitude fron simple stream calculations to finite difference models of 500 or more grid points. Tine variable calculations have energed extending from hour to hour calculations to long-term year to year calculations, Hydro~ dynamic circulation models are increasing!y coupled to vater quality models. Reflection indicates some general observations: 1, The aquatic plant/nuteient problens are the most difficult aodels with which ve have worked because of the complexity of the plant biology, the non-linear sater~ actions betveen nutrients and aquatic plants and the interactions of the sediment. 2. The dissolved oxygen problens, connected intimately with primary productivity and sedinent effects, in spite of the long history, tend to be considerably more con plex than generally believed. = Sediment interactions are important to all uater quality problem contexts and apparently credible interactive sediment models are only now appearing. It xtc substances fate sodels, linear in nature, tend to be less complex than gen~ erally believed. Past emphasis vas on nodels of fate (i.e. concentration), future sodels must of necessity include prediction of effects on che aquatic ecosysten and to a degree fon human health; toxic subscances represent the most complex problem context. experienced to date for prediction of effects of exposure concentrations. Wich this background and observations, it is necessary ta inquire to what degree water qual~ ity systems analysis has been “satisfactory” in some sense and the degree of impact on the Larger decision-making process. CRITERIA FOR JUDGING PERFORMANCE AND IMPACT OF SYSTEMS ANALYSIS IN WATER QUALITY NANAGEMENT In this content, "systems techniques” are considered che entire process discussed earlier, within which are esbedded predictive water quality modeling Cranevorks. The following criteria are offered for Judging performance and impact of systems cechniques: 1, The criterion of USEFULNESS, 1.0. the degree of use of the franevorks in decision aking: does it really aatter whether systems techniques are available? 2, The eriterton of ACCURACY, i.e. the comparison of predicted vater quality to actual vater quality after a control program has been implenenced; a post-audit analysis of the problem contexe. ‘The criterion of SERENDIPITY, te. whether systens techniques expose new, pre~ viously hidden interactions that are significant from a decision-making point of view, 4, The criterion of OWNERSHIP, 1.2. che degree co which the community at Large take ownership of our principles cheough legislation, regulations and policies that reflect the insights of systens techniques. Criterion #1 - Usefutee There Is Litele doubt that water qualicy modeling and system techniques are now used quite extensively tn water quality managenent decision contexts. Negotiations for discharge per- nits. evaluation of varying alternatives, support for higher or lower degrees of treatment are all areas chat now make extensive use of vater quality modeling. On the ocher hand, the luge of optimization planning models (e.g. cost minimization models) and optimal implenenta~ tion prograus (e.g. effluent charges) is apparently not as widespread. The relative exten— sive use of vater quality modeling techniques has been justified on economic grounds, t.e. 1 the belief on the pare of regulatory agencies and diechargers chat when properly applied, che application of the principles of predictive modeling is necessary but not sufficient to 2 rational decision. Tienens (1986) reports thar for che USEPA in Washington, D.C., about‘ R. V. THOMANN 100 projects have been reviewed over che past 8 year. In about one-half of these projects, the review, which included application of the principles of systens analysis ia varying degrees, resulted in deferring the decision or a significant change in the proposed environ mental control. The total capital costs of these projects impacted by the review was about SL billion. Thenens estinaced that other smaller projects reviewed elsevhere at the state and regional levels cay be an additional 200-300 in number but with a lesser overall total cost, So our techniques are useful and are being used in a variety of review contexts. A tribucary co che Chesapeake Say ayaten serves as an illustration. For this problem, (Salas and Thomann, 1975), data indicated poten- ‘violation of a 00 standard under lov flow conditions due partly co a large diurnal variation of oxygen. Chlorophyll levels were high (e-g. 300 ug/l) in the vicinity of the input. The question vas whether further renoval of BOD was warranted. A detailed model- ing analysis was conducted evaluating the various alternatives for control of the problem, Figure 1 shows the components of the 00 deficit from chie analysis. The maximum deficit i g 8 gy a 222018 16 141210 8 6 4 2 0 4 2 PHOTOSYNTHESIS LESS AESPRATION fo ZL B a ¢ -4 6 2al 1816 1412 10 8 6 4 2 0 ‘DAM, ‘S1P MILES ABOVE MOUTH OF WICOMICO RIVER Fig. 1. Wicomico River, 00 deficit components from water quality eodel indicating significance of phytoplankton respiration at aile 10 rather than point source carbon effects (Salas and Thomann, 1975). (minimum 00) at site 10 is calculated co be due to phytoplankton respiration exceeding Photosynthesis (together with the sediment oxygen denand) and not dua to the potnt source of oxidizable carson and nitrogen. I was therefore concluded that further reduction in these inputs would be only marginally effective and that emphasis should be placed on reducing the phytoplankton productivity through nutrient control. In the course of the decision aking process, the conclusion was accepted by the regulatory authorities and « recommendation was sade for construction of phosphorus removal facilities rather than additional carbon renoval. Criterion #2 - Accuracy Since a predictive framework employing theoretical principles and past experience ts at che ‘heart of the water quality management system, it is crucial that our models be credible fron an engineering point of view, but, equally important trustworthy and reliable from a management point of view. The forecasting ability of water quality models and uncertainties associated vith predictions have been examined in detail elseuheve (setSystems analysis in water quality management 1983). Here a fev cinple examples are presented for for example, Beck and van Straten, it Ls reasonable to evaluate Allustration. Since DO analyses have such a long history, how accurate our past analyses have been by examining the parfornance of 00 aodels. Post audit is the evaluation of aystem performance folloving Post audit of 0_nodels. ‘Three questions are addressed: ‘actual inplonentation of environmental control factlities. 1, Do the actual 00 data after a treataent upgrade fs tnetalled generally reflece the basic principles of D0 aodels, t.e. does the 00 go up when the BOD goes down? 2. To what degree are the DO models successful in predicting quantitatively the ob- served 0? 3. Dons the accuracy of the 90 models really aaccer in the decision regarding the treataent facilities to be installed? In the work sumarized here, an evaluation was sade of 52 water bodies vhere some data were avatlable on vater quality conditions before and after treatment (lydroQual, 1983). Thirey seven scates, five USEPA regional offices and six regional planning agencies were contacted, but in no case was there a complete compilation of wacer quality, biology, water use, cost of benefit data to perform a detailed post audit analysis. Wowever, 13 Wacer bodies did have some information for a review, The treatsene changes included increases fron primary to secondary and secondary to nitrification and advanced waste Regarding the first post audit question, the dats for the [3 cases indicated that the fnerease in DO normalized by the reduction tn ultimate oxygen denanding (UOD) load was inversely proportional to river flow (Fig. 2). 3 S 8 10 2 z =. 33 S310 x og a8 H a; wf O.t os ge # 2 b = 01 1 10 100 1000 10,000 AVERAGE STREAM FLOW (cts) Fig. 2, Relation batvaen actual 0 increase (ng/t) per 1000 Ibs UOD/day removed and spatial average strean flow (cfs) - [3 wacer bodies. (From HydroQual, 1983) At the very least, this {¢ the simplest confirmation of the classical DO nodel framework: That is, from the baste 00 sag equation we Know chat ADO/AU0D reduccion should be approxinately inverse co the river Clow. Jo firet approximation, then, our basic theory holds cogether and supports a fundamental tenet in D0 systens analysis: the greatest DO improvenent will result from facilities thac provide the largest amount of UOD renoval located on the suallesc water bodies. The dif- Fieulty is that Figure 2 ts a log-log plot, so the first approximation may not be all that satisfactory in a deciston making context. Therefore, ve need to take a closer look at the quaneitative performance of 00 models and address the second question.6 R. V, THOMANN Testing of six river aodels vas performed by setting the conditions (1.e. river flow, temperature and effluent) for the appropriate "after treatment change”. Ail model reaction Fates were identical fo those rates used in the original vaste load allocation analysis. Root mean square (RNS) errors served as one quantitative measure of uodel accuracy in Tepro~ ducing the data collected after a change in treatsent. In post-improvenent testing. &4S errors range from 0.0 ag/3 to about 2.0 ag/1. The average error of 0.9 mg/l was sonewhat larger than the RMS error of 0,7 ng/1 associated with calibration of the six aodels. Fig. 3 shows the correlation of observed co calculated mean 00 concentrations for the cal- ibration and post-luprovenent evalutions. The Figure clearly indicates that we do a good Job in calibration partly because ue have the data in front of us during this nodel cali~ bration phase. On the other hand, the post-improvenent comparisons, vhen we did not have fhe data a priori, indicace that the DO models tend to overestinate actual DO concentra tions at ievels [ess than 7 mg/i- 12 12 3° 10 88 8 i. 6 E a 3 32 2 z 6 ° ° 012345678 910112 01234567 8 9101112 OBSERVED MEAN D.O. (mg/1) OBSERVED MEAN D.O. (mg/1) Fig. 3. Compartson of calculated and observed DO concentrations ~ 13 water bodies (a) from calibration stage of model, (b) from post-audit stage, after treatment upgrade. (From fiydroQual, 1983) Recognizing that wany 00 analyses are conducted without recourse to any data at all, {¢ 4: important to also determin the credibility of the 00 models vhere "simplified desk top" seudies are conducted. Indeed. at least in che U.S., many mote permits are probably Assued by analysts who have never been within 100 miles of the tiver. The acenario theres fore was as follows: an experienced vater quality engineer was asked to analyze the DO in fen streans without looking at the D0 daca and only having available daca on the river characteristics, e.g., flow ané depth. Following the analysis, a comparison could then be wade between the simplified analysis and the actual data quantitatively, the "staplic ftee" models resulted in R45 errors that vere 50% to 200% higher than che 2S ereors developed from sore complex data-avatlable analyses. The average ANS ertor for the toa river analyses was about 2.0 ag/) The answer to the second question above is therefore somewhat sobering. With a detailed godel construction and using reasonably extensive (and expensive) data sets, the RMS error in the actual subsequent comparison to D0 Levels after treatnent upgrade ts about 0.9 mg/l. Sinplified, desk top analyses double that error. With these Kinds of errors, one wonders about all the discussion that sometines ensues in permit negotiations over a few tenths of a mg/1 00. Now, to the third question, t.e. do these errors really make any difference in the dect- Sion-making phase? In the preceding discussion, there are tvo types of errors that way occur in the comparisons: the first error is overestimation of the water quality Laprov.Systems analysis in water quality management 7 mene for ¢ given Level of ereatnenc. Therefore, water quality will be Less than actually thought after treatment upgrade and a water use interference may occur that vas aot aren dicted. “The aecond error is underestioation of the water mislity Loprovement ‘ewlcing ta overdesigned ceeatuent facilities and an overespendicure of funda. “the Firat error eaa be Rermed a water qualicy error (ive. quality (use) will be less then projected). The second error can be thought of as « facilictes error (i.e. the factiity tz overbuilt co, meet target water quality.) Comparisons for 10 rivers were nede between the decisions reached using simplified desk-top Eechniques aa compared to detailed nodeling approaches. Simplified modeling coulé have Potencially resulted in four water quality errors and two facilities errors. In four caseas Ehe decision was identical. The comparison, of course, assumes that che sore complex model analyses with available data results in "correct" decisions, which in fact is not always che case, With respect to an upgrade to nitrification facilities, the comparison indicated that the simplified aodels reached the same decision in nine of the ten cases. This is due Principally co the step increases tn UOD reduction with the instaliseion of nitrification facilities. One concludes from this post audit analysis chat simplified 00 models and to a lesser extent, aore sophisticated 00 nodels do not do very well in predicting actual values of DO after a treatment upgrade. RMS errors of 1-2 ng/i DO are the bad news. The good news {a that from a decision-making poine of view, ic doesn't seen to make ali chat much diffare ence especially for an upgrade to nitrification. The Potomac Estuary Case. This estuary has been the cubject of water quality managenent and sodeling effores for several decades. Frevdberg (1985) has reviewed the history and che implications of the modeling work. In the late 1960’s, extensive algal blooms developed in addition to a depressed oxygen condition in the Washington, D.C. viciaiey, As a result of nodeling efforts by people such a3 Jaworski et al. (19/1), sigaiticant Teductions in incosing carbonaceous and aitrogenous BOD loading vas accomplished along with significant reductions in point source phosphorus loading. Considerable controversy surrounded the phosphorus reduction strategy since it was argued that nitrogen vas the Ltateing nuccient and chet nitrogen should be contralled. Concern waz also expressed over the release of phosphorus from the sedimene. ‘The phosphorus removal seracegy waa founded fon the notion that with suffictent reductions af phosphorus, that chemical could be made the Limiting sucrienc. Since it vas considered cheaper to renove phosphorus then nitrogen, the phosphorus renova! prograz was inscituted. Fig. 4 shows the reductions in phosphorus during the late 1970's and early 1980's. 15,000 i 5000 TOTAL PHOSPHORUS LOAD. POINT SOURCES (bs/d) EUSUTENT INTO ATATA ANNAN PN NLU ASU oO 1910°20 "30 40 ‘50 60 YEAR Fig. 4. Phosphorus loads, Poconac estuary (Javorski eta sbah., 1985; Metro. Nash, Council of Gove's, 1985). a é g 1971; Thomann A sajor algal bloom occurred in 1977 folloving the first stage of the phosphorus reduction Program. An intensive effort was then undertaken to updace the modeling franevork for Eee Nee ee eee eT SMeR. V. THOMANN eutrophication and che Potamac Eutrophication Model (PEM) was constructed (Thomann and Hlespaerick, 1982). This aodel vas calibrated and verified using 7 yeare of data. The PEN “included sedigese interactions aad a fully developed time variable phycoplankton— huttient species franevork. The model was subsequently utilized for analysis of altern tives. ‘then in 1983 with che phosphorus reductions now almost fully in place (2100 Lbs Total Phosphorus (7P)/day versus 24,000 Ibe TP/day in 1970), another aajor bloom occurred of the Dive green alga, Microcystis seruginoga. The enbaynents reached levels of almost 600 te chi/i with the gain channel reaching concentrations of about 300 wg chi/1. With the Geecary a brilliant green, the treatment plants for Vashington, 0.C. removing phosphorus seravdaa of about 0.4 ag TP/1 (compared to the permit levels of 0.2 ag/1), there was sone KSnscernation. ca say the least. The questions abounded. Was this a aajor setback? Was fhe wrong nuteient being renoved? Was the PEX faulty? Would the situation be relieved when Eveatment plane discharge levels actually reached the target effluent concentrations of 0.2 mg TP/i?) What actually wes the cause of the bloom? ‘The Metropolitan Washington Council of Governments, the user agency, applied the unchanged Pox with che 1983 meteorology and hydrology. The model tracked the onset of the bloom up to bout 100 ug/i by the end of July but then failed co reproduce the further intensification Gf che bloom. The principal reason for the failure of PEM to capture the full bloom was that the model ran out of phosphorus (see Fig. 5). Sensitivity anelyses indicated that Chere ‘was ‘au apparent additional phosphorus source of about 4000-8000 Ibs/day not included in PEM, Fig. 3 fron Frevdberg (1985) shows che effect of including this source and it is ‘seen that if partially explains the observed bloom. OBSERVED DATA PEM CALCUTION w/ UNADJUSTED COEFFIC. PEM w/ 4000 Iba/d (MP 25 to 35) © 10 203040 50 60 0 10 20 30 40 50 60 MILES BELOW CHAIN BRIDGE Fig. 5. August 28, 1983 bloom, Potomac estuary (Freudberg, 1985) ‘A series of hypotheses vere investigated (Thomann et al., 1965). It was concluded chat the nose likely mechanism for the phosphorus input was an enhanced aerobic sediment phos- phorus release resulting from the high pH in the overlying water (Di Toro and Fitzpatrick, Toes), Coneributory wechanions included the effect of two-dimensional vertical circula~ tion trapping sucrients and elevating the available phosphorus concentration. tn sumary, the PEM failed to capture the full 1983 bloom because of a combination of a tmigsed wechaniaa” and a simplification of che estuarine transport in the lower reaches of the astuary.. However, one should aot conclude, T believe, that the Potonac situation is an Guanple of uhere water quality modeling is of Lietle value. Rather, without PEM, che Gvaluation of what happened in che 1983 bloom would have deen reduced to qualitative dis~ SGssion of che data. The availability of PEM mediately pinpotnted the principal problea: Gore phosphorus io the estuary than could be accounted for from previous observations. The‘Systems analysis in water quality management Potomac tilustration does indicate clearly the need for extensive monitoring and post~ Gudieing of our work, both to denonstzace credibiliey, or lack thereof, and ¢0 uncover pre~ Viously hidden or ansuspected mechanisas that becove important after Creatnent programs are tn place. Model Quality Assurance, tn recent years, a wide variety of individuals with widely varying educational dnd experience backgrounds are “practicing” systems analysis in vater quality snagenent, These are people representing local, state, or regional governments, or indi- vidual dischargers. They have been placed vith the responsibility of doing calculations to Support the decision-making process. Recently, Gallagher (1986) completed the results of a variety of water quality godeling ana~ lyses prepared £0 support applications for AWT systens. The compilation reveals some sur~ prising insights into where we are in the actual usage of vater quality aodels on the more focal day to day decision making level. A review of Gallagher's coupilation indicates that quality assurance of model formulation and application is essential for the maturation of the disefpline. The following problem areas in model formulation arise: 1, Taput data are not always carefully checked. 2] Software coding errors occur- 3. Variables are incorrectly used in formulae 4) parameters not always checked for reasonablenss of range, Ta the calibration and verification stage, the aost difficult areas appear co be: 1. Parameter estimates may vary significantly enough berveen analysts Co result in significantly different decisions; reaeration rate determination is an example. 2, Sediaent-water column interaction uncertainty persists and often cannot properly address the issue at hand; the calculation of nuerient release, toxic substance Telease and sediment oxygen denand within the model framework are examples. 3. A lack of fuller understanding of the variety of sechanisns that may “explain” data hanpers the credibility of models; improper assignnent of deoxygenation and altrification coefficients sediment interactions, and algal effects on BOD are examples. Im the projection and evaluation of alternative phase, che most difficult areas are: 1. Projection of paraneter changes under different environmental controls, e.g. nicrobial degradation of toxic substances. 2. Projection of sedinent water column interactions under environmental conditions that have not yet been observed. ‘The difficulty with these concerns is that errors in coding or input, unreasonableness of parameters or failure to recognize aechaniens are not alvays eastly uncovered in the ana— Gyete. Improper overestimation of one parameter nay be compensated for by the analyst by undercstiascing another parameter. In the evaluation of alcernatives, erroneous projections of effectiveness may chen be sade. Criterion #3 - serendipity This criterion addresses che issue of whether systens techniques reveal previously un- detected sechanisns or interactions that have a bearing on the degree and kind of environ~ ‘mental control. The preceding case of che pi sediated release of phosphorus from the sedi~ ‘aents of che Potonac estuary is an exanple. If systems techniques on occasion indicate such Serendipitous results then indeed i 1s worthvhile to use such techniques in che dectston- auking process. cost sininization franevorks often display such behavior where optimin solutions indicate cost trade-offs thac were not previously apparent. The degree to vhich such solutions have been included in actual program implenentation is cot clear. However, in the area of water quality wodeling serendiptious results may be directly incorporated into the decision on environmental control. The Back River ¢: For the Sack River, a tributary of the Chesapeake Bay (Thopann et TSBL); phytoplankton levels were cf the order of 200 vg chi/ and reached naximun Gevels of greater than 500 ug/2. Simple model analyses indicated that a 2 mg/l orchophos~ phorus effluent Liait would not be sufficieat co reduce phytoplankton. Subsequent inter~0 RV. THOMANN est then focused on AWT to total phosphorus levels of 0.2 mg/1 versus an effluent reloca- tion to Baltimore Harbor, More detailed modeling indicated that there vas virtually no aif- ference in che orthophosphorus concentration in the Back River between the two alternaciv. . (see Fig. 6). This result vas not anticipated since relocating the discharge out of Back os: 08 ‘COMBINED 2 @ 2 b ® ore ORTHO = P (mgt) 2 S oO. S14 12 10 8 6 4 2 0 16 i442 10 8 €@ 4 2 0 DISTANCE FROM MOUTH (km) DISTANCE FROM MOUTH (km) ‘ Fig. 6. Calculated phosphorus levels in Back River with (a) AWT and (b) effluent diversion : River was thought to be gore effective. Sut decreasing the input load by treatment was of f= set by diversion because the loss of flow through the Back River enhanced the effects of ' sediment release and allowed increased uptstream aigration of the Bay boundary waters. As @ Pareial consequence of this result, effiueat relocation was aot implemented. Rather, staged program of treatment upgrade to high levels of phosphorus removal hes been. tuple. sented. The first phase is under construction with phosphorus removal scheduled for a later : Phase. { believe chat many more such examples exist, but are often buried in the archives Of individeal studies. Ie is taportant to document such results and espectally to document the econonic consequences of these cypes of unexpected behavior. Ceteerion #4 - Ownership Aside from the gore technical uses of our product, to what degree has our work been incorporated or adopted into legislation and policy related to vater quality management and control? On the U.S. side, it is encouraging that some, but not all of the basic principles of water quality management have been recognized and are reflected in various lays, regulations and policies. It is discouraging that gone principles have been largely ignored. ‘The most fundamental principle on which our entire work is based is that the aquatic envi= 1 ronment has a certain capacity for absorbing and transforming residual inputs without create ing undesirable impacts. Under this fundazental principle, there follow the key steps i outlined above. The entire process is iterative and following an evaluation of the costs, | benefits and risks avoided of a series of alternative objectives, a final determination 1s nade as to the must desirable sec of water uses and vater quality standards. Treatment of residuals is based on che spectfic characteristics of the water body, and various economle, social and policy constraints. Technology-based etEluent Limitsare therefore justified only where the ability to establish receiving water standards does aot exist. If the fundamental principle of environgentsl absorption of wastes vere not recognized in legislacion and policy, then the reason for our work would simply cease. The regulatory Program vould then be an effluent based program with effluents set at the highest cechmoe Logically attainable level. the progran is utter simplicity: either you erest to the highest level possible or else you are in violation. The process ts then secicely a Lege! matter co be argued under adversary proceedings. In the U.S., our national legislation has essentially embraced this latter concept with our zero-discharge goal and to a related degree our program of effluent based requireuents for specific industries through Best Practical Treatment (BPT) and Best Available Treatnent (BAT). On the surface then, one could argue that the most fundanencal principle of systems 7 analysis and vater quality management has been ignored and that we have not only lost the battle, but the war as vell. After all, there 1s really no rational enginsering/ sclentific Justification, in ay opinion, for an across-the-board requirement of secondary tieecaect oFSyscems analysis in water quality management n technology-based industrial chemical control; such effluent requitements to be ‘aposed regardless of the type of water body, of resulting impact on vater quality or aquatic eco= systen, We arrived at that point, at least as far as I can see, from a mistaken perspective that the water quality standard route was aot practicel froma regulatory point of view. The lack of practicality centered around the belfef that our ability to assess water quality Impacts vas taperfect and that it vould aot be possible to evaluate each site specific sit tation. Tm actual implenentacton hovever, the situation is often quite different. Questions abound about the cost of effluent based prograns and about the relative impact of effluent based programs on vater quality and use. At least one result of this questioning has been the Feevaluacion of the requirenent for secondary ereateont in coastal waters with high effluent atlurion, although even here che results have been spotty. Ease of regulation ané political and social constraints appear to have been the major factors for aot incorporating the prin~ ciples of vater quality nenagenent nore fully. on balance, hovever, the previous principles have been incorporated in a variety of ways albeit not totally. fxasples include (1) the requirement by the USEPA that any request for funding for an aWT project costing more than $3 aillion be subject to a rigorous review of costs and effectivencee; (2) the Great Lakea Water Quality Agreenent (Internationa! Joint Commission, 1985) and (3) the water quality-based procedures and policies for texte sub- stances control (USEPA, 1985). For the first example, the Great Lakes Agreement explicitly calls for carget loadings of phosphorus from Cenada and che United States for Lakes Erie and Ontario to meet specified target objectives. ‘The significant economic savings resulting from the second example have been noted above in this paper (see Criterion #1 ~ Usefulness). In the third example, due recognition ts given of dilution effects, domstrean decay ané probabilistic interactions in establishing effluent levels for chemicals or whole affluent toxicity. ‘The Delavare Estuary case. tatensive vater quality sanagenent studies began for this estu- ary in che early 1960's and reauited in waste load allocations for individual discharges to meet water quality standards promulgated by the Delavare River Basin Commission (DRBC) (See ‘Thomann, 1972 for discussion of the full study), Eaphasis was on the DO. Extensive use vas nade of water quality modeling and cost wininization techniques in arriving at trade-offs between various alternatives. Actual prograa implementation incorporated cause-effect rela~ tionships to establish waste load allocations, but did not formally include any optimal con- erol strategy. In the ensuing two decades, this program hao remained largely in place. Significant reductions tn CBOD have been accomplished (about 7OX reduction from 1964 to 4986) (DRBC, 1984, 1986a,8), but sine NBOD was not allocated, NBOD leads have remained virtually constant. The’D0' response is shown in Fig. 7. The trend {n improvement is close fo the projection made 20 years ago. The water quality standard that was adopted as a result of the earlier efforts, has almost been achieved after tv decades, although there 1s Still some variation of DO below the 2% hour miniaum average standard. To what degree did the application of systems techniques in the 1960's sake a difference in che water quality aanagenent program for the Delaware estuary? Since the average percentage BOD renoval is 86-89%, one could argue that such levele vould have been mandated in any event by the across-the-board secondary treatment equivalent of U.S. Clean Water legisla~ tion. On the other hand, the inplenentation of a rationally based program provided a con- tinuous defensible framevork for regulatory purposes. By having an allocation system that included a reserve capacity, provision for reallocation and renegotiation, the Delaware pro- frau could proceed in a reasonably orderly fashion aod aim tovard a water quality objective that was realistic, cost effective and consistent with che goals of the Delavare comunity. Im summary, the record for this criterion of including the basic tenets of our work in legis- lation and’policy is a mixed one at best. The objective seeus to be laudable, i.e. a blend of a program that {s easy to iuplenent with a program that ie scientifically well founded and cost effective. However, the widespread application of techaology based effluent pro- Brame vith Little evaluation of the effectiveness and coate of such programe testifies to the dominance that the eace of regulation can assume vhen the public pressure to get some— ching done is intense. I believe that the continued uncritical use of such effluent-based programs is a mistake and is counter-productive in the long run. A detailed study of this type of policy vould seem to be most warranted. Nevertheless, the ceed for rational deci- sion-making is often so compelling that the techniques of our vark can no longer be avoided and are subsequently included in the decision-making process.R. V. THOMANN (1986) 4 ESTIMATED OBSERVED 3 SUNRAER MEAN 3 3 (1982) OR Averace Sg 2 3 BEo z (1983) (1981) =¢ e 50 3 = iB > 1 = o 964) 4 6 8 10 INPUT LOAD TO ZONE 3 & ZONE 4 10° Ibs. CBOD/ day Fig. 7. Response of Delavare entuary DO to reducttone tn 6800 compared to to projection mado in 1966. (Data from Thonann, 1972 DREC, 1986, 1986a,5). Different years have different fresh ater flow inputa. CONCLUSIONS AND SUGGESTIONS over the past quarter century, ayatems analysis in water quality management has come of ag More specifically, water quality modeling techniques coupled loosely to cost analysis frane- vorke are nov utilized in varying degrees at all levels of decision-making. From early screening of potentially toxie chewicals using generic water bodies to detailed analyses of alternative treatuent strategies, ayetene techniques especially water quality aodeling have been used extensively, Reflection on the preceding criteria and our fev illustrative indicates that indeed systens analysis has had an impact, in fact a decided 4mpact. Houlty 4s that the impact 16 not widely documented or known anong those responsible for key budgetary and regulatory decisions. It {# now considered almost essential that evaluations of cost and effectiveness of alternative actions be nade using state of the arc modeling and systens techniques. Practical decisions are made using these franevorks. The overall supposition is thar such decisions are betcer decisions because they are nore informed, the rationale for the decision is more structured and has & firmer engineer ing/ectentific foundation. As a reault, there 1a the hope that decisions in vhich syatene analysis played a role will be sore universally acceptable. That this universal acceptance Ls not always achieved 19 obvious and the endless, sometimes contentious, erguaeats over wacer quality modeling give cestimony to this lack of universal acceptance. Although there {s se{ll a lingering group who see systens techniques and attempts at a priort alternative assesonent as modified "witchcraft", the voices are getting dimer. There are exceptions, Of course, where the clanor for maximum technological reuoval of vastes sometines drowns out a plea for a more balanced approach. The present directions of control of chemical dis- charges may fall toto this category unless there {s a continual campaign for the use of systema techniques for sore inforned decision-makingSystems analysis in water qual ‘management We can properly take eredit at least to sone degree for # general incorporation of sone of the baste principles into policy and legislation. The use of effluent discharge permits, seasonal treataent and associated seasonal permits and writing the requirement for quanti- tative assessment directly into policy are examples. On the other hand, many Key ideas of optimization apparently still lie dormant: cost sinimtzation, effluent charges, optioun progras implenentation are examples. Where we need co do more work is obvious. As noted previously, there is a need to document our work more extensively, especially those specific exanples where decisions have been impacted, costs have been reduced, benefite realized and new interactions exposed. It is suggested chen that ¢ world-wide effort be devoted to compiling such vork with the expressed purpose of displaying the considerable econoaic advantages from using systems techniques in water quality sonagenent against a strictly effluent based approach. Technically, we lack defensible cested aodels of sediment interactions, effects of chemicals on aquatic ecosystens and stochastic frameworks that have been calibrated and verified. Enereased quality assurance of model formation and calculation 1s secessary and -indicates a need for upgrading the training of operational uses of vater quality ayscems techniques. State of the art computer technology, e.g. graphics, and use of expert systens should be Fully exploited to meet thie critical need for properly trained analysts We have filled an essential role admizebly during this past quarter century. Our work has been good; ost importantly, it has been truly beneficial in fully respecting the intricate balance that must be struck betveen a healthy eavironment and a healthy econoay. ACKNOWLEDGEMENTS ‘The thoughtful diseuestons and input of Thomas Gallagher of HydroQual, Inc., Charles App and Myron Themena of the USEPA, and the assistance of Bruce Lin and Dennis Scannell, graduate students at Manhattan College, are gratefully recognized. The readiness of the Delaware River Basin Commisaion to be of help and provide data is also appreciated. Special thanks are due to Mrs. Eileen Lutouski for her patient typing and retyping of the aanuseripe. REFERENCES: Beck, M.B. and van Straten, C., Eds. (1983). Uncertainty and forecasting of water quality. Springer-Verlag, .¥., 386 pp. Delavare River Basin Comission. (1984). Water quality inventory for the Delavare River. Trenton, NeJ.» 49 pp. Delaware River Basin Commission. (1986a). Delavare River vater quality assessment, 1984-85 305(b) Report. Trenton, N.J., 53 pp. Delaware River Basin Commission. (1986). Water quality and efficient load deta, Private communication. Di Toro, D.M. and Fitzpatrick, J.P. (1984). A hypothesis to account for the 1983 algae bloom. Tech. Meno, HydroQual, inc., 28 pp. Froudberg, S.A. (1985). A post-audit of the Potomac Eutrophication Model. Met. Wash. Councii of Goves. Paper presented at Auer. Soc. Civil Engr. Water Res. and Planning Div. Conf., Buffalo, N.Y., 12 pp. Gallagher, T."(1986).," Compilation of AWT fact sheets. HydroQual, Ine., Mahwah, N.J. HydroQual, Tac. (1983). Before and after comparisons of vacer quality following municipal Treatment plant iaprovesents. USEPA, Office of Water Prog. Operations, Facility Requirements Div., Wash. D.C., 7 Chapt. + Append. International Joint Commission (1985). Great Lakes vater quality agreement of 1978, 52 pp. Javoraki, NA., Clark, L.J., and Feigner, K.D. (1971). A water resources-wacer supply study ‘of the Potonac estuary. Tech. Rept. No. 35, EPA Water Quality Office, Middle Atl. Reg., Aped] 1971 Metropolitan Washington Council of Governments (1985). Water quality of the Potomac River. 1988. W.S. and Thomann, R.V- (1975). The Chesapeake Bay waste load allocation study. Prep. by Hydeoacience, inc. for che Md. Dept. of Nat. Res. Admin., Annapolis, MD, 287 pp. ‘Thomann, R-V. (1972). Systems analysis and vater quality managenent, McGraw-Hill Book Co., N.Y. Reprinted by J, Williams Book Go., Oklahoma City, OK, 279 pp ‘Thomana, R-V., Gallagher, T.W. and Pacquin, P.R. (1981). ‘Review of the water qualiey ‘analyses of the Back River. Report to USEPA, HydroQual, Inc-. Mahwah, N.J., 29 pp. salry R. V. THOMANN Thomann, R.V. and Fitzpatrick, J.J. (1982), Calibration and verification of a model of the Potomac estuary, HydroQual, D.C., 500 pp. * Append. Tne., Final Report to D.C. Dept. of Env. Sciences, Wash., Thonana, R.¥-5 Javorskt, WJ.) Hixon, S.W., Paerl, H.W. and Taft, J. (1985). The 1983 algal Bloom in the Potoaac estuary. Wash. D.C., 7 chapters. ‘Themen: U.S. Env. Pro. Agency (1985). control. Uffice of Water, Wa: Prepared for Potomac Strategy State/EPA Manageaent Comm. y u, (1986). Privace coumuateation. Technical support document for water quality-based toxice ssh. D.C.» 74 pp. + 5 Append. Webster's “Seventh Nay Collegiate Dictionary (1963). G, & C. Merriam Co. Publishers, ‘Springfield, MA, 1222 pp-