VESIENTUTKIMUSLAITOKSEN JULKAISUJA PUBLICATIONS OF THE WATER RESEARCH INSTITUTE

Erkki Alasaarela & Pertti Heinonen: Alkalinity and chemical oxygen demand in some Finnish rivers during the periods 1911—1931 and 1962—1972 Tiivistelmä: Eräiden Suomen jokien alkaliniteetti- ja CODMfl-arvoja vuosilta 1911—1931 ja 1962—1979 3 Kirsti Haapala, Tuija Ruoho-Airola, Lennart Granat & Olli Järvinen: Comparability of results from deposition samples Tiivistelmä: Laskeumanäytteiden tulosten vertailtavuus 14 Lea Kauppi: Contiribution of agricultural loading to the deterioration of surface waters in Finland Tiivistelmä: Maatalous vesistöjen tilan muuttajana Suomessa 24 Lea Kauppi: Nitrate in runoff and river waters in Finland in the 1960’s and 1970’s Tiivistelmä: Valuma- ja jokivesien nitraattipitoisuudet Suomessa 1960- ja 1970-luvulla 31 Lea Kauppi & Maarit Niemi: The role of runoff water phosphorus in eutrophication Tiivistelmä: Valumavesien fosforin merkitys vesistöjen rehevöitymisessä 41 Reino Laaksonen & Väinö Malm: Changes in water quality in Finnish lakes 1965—1982 Tiivistelmä: Järvien veden laadun muutoksista v. 1965—1982 52 Reino Laaksonen & Väinö Malm: Changes in ionic distributions in Finnish lake water during the period 1968—1983 Tiivistelmä: lonikoostumuksen muutoksista järvissä v. 1968—1983 59 Jarmo J. Meriläinen: Macrophyte vegetation at the River Kyrönjoki estuary in 1982 Tiivistelmä: Kyrönjoen suiston suurkasvillisuus vuonne 1982 61 Jorma Niemi & Juhani Eloranta: Application of the FINNECO model to Lake Pyhäjärvi Tiivistelmä: FINNECO-mallin soveltaminen Tampereen Pyhäjärveen 77

VESIHALLITUS—NATIONAL BOARD OF WATERS, FINLAND Helsinki 1984 Tekijät ovat vastuussa julkaisun sisällöstä, eikä siihen voida vedota vesihallituksen virallisena kannanottona. The authors are responsible for the contents of the publication. lt may not be referred to as the official view or policy of the National Board of Waters.

ISBN 951-46-8081-2 ISSN 0355-0982

Helsinki 1984. Valtion painatuskeskus 77

APPLICATION OF THE FINNECO MODEL TO LAKE PYHÄJÄRVI

Jorma Niemi & Juhani Eloranta

NIEMI, J.S. & ELORANTA, J. 1984 Application of the FINNECO model to Lake Pyhäjärvi. Publications of the Water Research Institute, National Board of Waters, Finland, No. 57. The FINNECO lake model is a one-dimensional water quality simulation model developed in a joint project between the National Board of Waters, Finland and International Business Machines, Finland during the years 1978— 1981. The model has eighteen state variables, includes the effect of wind jo mixing water and simulates the freeze-up and break-up of ice on the basis of energy balance. The model was applied to the eutrophic Lake Pyhäjärvi situ ated in southern Finland. Calibration was carried out by a trial and error method uaing the data of the open water season of 1980. LJsing the same parameter combination and the corresponding data of 1981, the agreement between the model output and observations was päor and the model could not be conaidered as verified. However, reaaonably good agreement was ob tained with this set of data by changing the values of only five parameters. It can be concluded that the model is unable to sirnulate water quality accurately in a lake as eutrophic as Lake Pyhäjärvi. The sensitivity of the model to the input data was also tested. When the input data was given to the model as weekly or monthly averages instead of once every eight hours the values of most state variables, except those of phytoplankton, did not change markedly. This shows that the interval between individual input data can be longer than eight houra, which simpiifies the application of the model and decreases com puter costs. Index words: Mathematical models, water quality modeis, ecological modeis, simulation models, FINNECO model, water quality prediction, Lake Pyhä- järvi

1. INTRODUCTION They are always simplifications of natural ecosystems and therefore should be vigorous In recent years the use of systems analysis in ly tested by applying them to different types assessing environmental problems has in of water bodies in order to determine their creased considerably, in particular the use of suitability for practical water management. mathematical modeis. Water quality modeis, Application of the modeis typically includes which are one group of environmental mod gathering of a large amount of information eis, have typically been used for making pre conceming the inflowing waters, such as quan dictions of water quality and for carrying out tity and quality of tributaries, wastewater research on the basic mechanisms of aquatic treatment plants, diffuse loading, precipita ecosystems. tion and in addition meteorological data and A great number of differet water quality historical data concerning the water quality of modeis have been described in the literature. the case study lake. This information must be 78 organized in the form required by the model the lake basin is divided into hydraulic ele before calibration can be started. ments parallel to the surface of the lake. Ry The calibration process requires thorough draulic elements are considered to be homo knowledge of the structure of the model, con geneous and therefore differences in water siderable computer time, much data and a cer quality are observed only along the vertical tain amount of intuition on the part of the axis of the lake. The model has eighteen state user. Special methods such as automatic cali variabies, namely: temperature, dissolved oxy bration for the estimation of parameters are gen, biological oxygen demand, alkalinity, difficult to use with such a complexmodel as pH, total inorganic carbon, total dissolved FINNECO. The normal method of calibra solids, organic sediment, carbon dioxide, am ting the model is therefore to use trial and monia-, nitrite- and nitrate nitrogen, phos error, which has many disadvantages as the phate phosphorus, phosphorus in inorganic criteria for evaluating the agreement between sediment, hygienic indicator, sodium lignosul the computed and observed results are hard phonate, phytoplankton (10 groups) and zoo to define. Most of the time used in the appii plankton. Driving variables of the model in cation of complex models is spent in gath clude meteorological data as well as data con ering the data and in calibration. cerning the quantity and quality of inflowing The FINNECO model was developed in a waters. The model includes a mechanism for joint project between the National Board of taking into account the effect of wind in mix Waters, Finland and Oy International Busi ing water and it simulates the annual freeze ness Machines Ab, Finland during the years up and break-up of ice cover on the basis of 1978—1981 and it was calibrated and verified energy balance. to Lake Päijänne with the data of four years The structure of the model was presented with good results (Kinnunen et al. 1982). in detail by Kinnunen et al. (1982). The users’ Modeling was carried out with the sets of data manual was prepared by Kauranne (1983). covering the whole year, starting from May and continuing until the end of the next April. In the present application, however, simulation was carried out only with the data of the open water season. Pyhäjärvi, the case study area of this Lake 3. APPLICATION OF THE MODEL work, receives both municipal and industrial effluents. The lake is eutrophic, with high nu 3.1 The case study lake trient concentrations and low concentrations of dissolved oxygen, and has a short detention Lake Pyhäjärvi, situated near the City of time. It differs therefore in many respects Tampere in southern Finland, was selected as from Lake Päijänne, which is eutrophic only the case study lake for the application of the in the northern part. model. The detention time of this lake is rela The objective of this work was to deter tively short and it resembles to some extent a mine whether the FINNECO model could be slowly flowing river. The lake receives waste succesfully applied to this type of lake with waters from factories and from the City of out changing its structure. In addition, the Tampere and its surroundings. The water effect of the interval of submitting the input quality of the lake has been followed over a data on the computed results was tested. It period of several years. These investigations was found that the model was less suitable to have been carried out mainly by the Water Lake Pyhäjärvi than to Lake Päijänne. District Office of Tampere and by the Water Protection Association of the river Kymijoki. In addition, special investigations have been carried out. Kuparinen (1981) investigated the inhibitive effects of various substances on-bacterial activity in the lake. Virtanen et 2. THE -MODEL. al. (1979) investigated the flow velocity of the water in lake. The FINNECO model is a one-dimensional Both municipal and industrial waste waters water quality simulation model that can be are discharged to the lake (Table 1). The wa applied to lakes and reservoirs. In the model ters that are discharged from Lake Näsijärvi 79

Table 1. The amounts of wastewater, BOD7, total ter quality model of this type could be applied phosphorus, total nitrogen and total dissolved solids in to the lake and perhaps later the principal sources of nutrients discharged to Lake Py on be used in hljärvi in 1980. water management.

Source Waste water BOD7 Total Total Dissolved of loading discharge phos- nitro- solids phorus gen 103ma1 t a t a t a1 t a1 3.2 Sources of data ‘Viinikan1ahti 20 500 1 300.0 21.40 589.0 484.0 Rahola 4 050 171.0 4.20 136.0 34.1 Tako 4 550 570.0 2.40 10.2 915.0 The data necessary for the use of the model Kyösti 681 14.4 0.92 17.6 10.0 can be divided into initialization and simula Loukonlahti 239 15.8 0.68 6.1 14.5 tion data. Suomen Trikoo 622 32.6 0.30 6.3 19.2 Initialization data includes the data that is 30 642 2 103.8 29.9 765.2 1 477.0 needed to define the condition of the lake at the beginning of simulation and to control the computer program. This data includes mor to Lake Pyhäjärvi through the Tammerkoski phological data of the lake, values of the pa rapids are influenced by wastes of the pulp rameters in the equations and information and paper industry. Because of these heavy controlling the computer program, such as nutrient inputs the water quality of Lake Py the length of the simulation period, the first häjärvi is poor, although during recent years and last day of simulation and the days when slight improvement has been observed. In special printout is required. The initialization particular, the concentrations of coliform bac data is given to the model only once at the teria and nutrients have decreased. For exam beginning of the simulation. ple, phosphorus concentrations in the surface Simulation data includes weather data and water of the lake were in the winters of 1973 data on the quantity and quality of the infio and 1976 200 and 77 pg i—1, respectively, wing waters such as rivers, wastewater treat while in 1979 the corresponding concentration ment plants, precipitation and diffuse loading. was only 25 pg 1. A similar decreasing trend was not observed during the same period in the hypolimnion, because this layer is at times anoxic and nutrients are released from the bottom sediment to the overlying water. The case study area extends from the Tammerkoski rapids to the bridge of Rajasaa ri (Fig. 1) and forms a homogenous entity. According to the investigations of Sarkkula and Forsius (1977) the direction of flow in the lake is in practice solely towards the bridge of Rajasaari. Data of depth and surface area of the case study lake were calculated on the ba sis of the bathymetric map. Surface area, vol ume and average depth of the lake are 22.12 km2, 192.11 10’ m3 and 8.7 m, respectively. The maximum depth is 46 m. The mean dis charge of the Tammerkoski rapids was in Fig. 1. Principal sources of waste water in the case study 1961—1971 66 m3 s—’. According to Sarkkula area, O = Wastewater treatment plant, $ = Factory. (1977) 1 Rahola and Forsius water does not flow equal 2 GAS Oy, Lielahti ly through Lake Pyhäjärvi, but rather along 3 Suomen Trikoo Oy certain defined routes. According to the same 4 Näsijärvi Oy inves-tigation the detention- time of the case 5--YP-Oy, Santalahti study -lake is 27 days. 6 GAS Oy, Tako It was evident that this lake is not a promi 7 Oy Finlayson Ab, Tampere 8 TKS, Naistenlahden voimala sing case study area for any model. However, 9 Viinikanlahti serious water pollution problems exist in the 10 Loukonlahti area and it was decided to test whether a wa 11 Kyösti 80

The simulation data is given to the model treatment plant was given monthly and the once during each simulation time step, in this daily values were interpolated on the basis of application once during every eight hours. this information. The initial values of physical and chemical Precipitation data was obtained from mea state variabies were obtained from the data surements carried out at Tampere airport. bank of the National Board of Waters. The Daily values were calculated on the basis of values of some of these variabies, such as monthly avarage values. The data on the qual BOD, detritus and total dissolved solids, were ity of rainwater was obtained mainly from estimated. Phytoplankton measurements were the investigation of Järvinen and Haapala not carried out in Lake Pyhäjärvi and there (1980). Water quality data of rainwater was fore the initial values of the two phytoplank assumed to be: ammonia nitrogen 15.0 mg ton groups were estimated on the basis of ob m—2 month—’, nitrate nitrogen 15.0 mg m—2 servations carried out earlier (Lepistö et al. month1, pH 4.8, alkalinity 2.7 mval m 1979). month1. The concentration of phosphate Meteorological data was obtained from the phosphorus was assumed to be half of the mea Meteorological Yearbook of Finland (1981) sured amount of total phosphorus. The con according to observations carried out at centrations of dissolved oxygen and total dis Tampere airport. The method of obtain solved solids were assumed to be 10 mg 11 ing the weather data for every eight hour and 50 mg l, respectively. The temperature period was identical to that used previ of the rain water was assumed tobe the same ously in applying this model (Niemi 1978, as that of the surface water of the lake. Kinnunen et al. 1982). The data on the Investigations carried out in small hydrolo quantity and quality of the water of the gical basins were used in estimating the diffuse Tammerkoski rapids was obtained from the loadings entering Lake Pyhäjärvi. The drain hydrological data bank of the National Board age area of the case study area was estimated of Waters. Daily discharge data of the Tam to be 159 km2. The runoff entering the lake merkoski rapids was used. The water quality from the drainage area was calculated on the data was interpolated for each time step on basis of observations made at the small re the basis of rnonthly observations. search basin of Paunulanpuro. The runoff vai Near Lake Pyhäjärvi there are several ues were obtained as monthly averages from wastewater treatment plants, which discharge the hydrological office of the National Board wastewater to the lake. The most important of Waters. Kauppi (1979) presented the esti plants are the wastewater treatment plants of mates of total phosphorus and total nitrogen Viinikankoski, Rahola, Tako Oy, Kyösti, loadings from this drainage area, expressed as Loukonlahti and Suomen Trikoo, which were average values for the years 1964—1974. Ac included in the model. cording to the same investigation the annual The loadings of these six wastewater treat diffuse load can be divided into three seasons: ment plants were added together and were spring (April and May); summer and autumn treated in the model as one major source of (June-November); and winter (December loading. The water quality data for each plant March). In this work the spring period was was obtained from the regular reports of the assumed to continue until the end of May al plants. Because phosphate phosphorus and though in Kauppi’s study (1979) the spring ammonia nitrogen are not analyzed from period ended earlier, i.e. on the l9th of May. wastewater it was assumed that the concen The nutrient concentrations were calculated trations of these substances were equal to the for each of these three periods on the basis concentrations of total phosphorus and total of the average loading values for the years nitrogen, respectively. The temperature of the 1964—1974 and of observed runoff values for wastewater was estimated on the basis of 1980 and 1981. The nutrient concentration monthly observations carried out in the three remained constant within each period. In cal largest plants. The following estimates of the culating the nutrient concentrations it was as quality of the water leavingthe wastewater sumed that phosphate phosphorus• concentra treatment plant were assumed: pH-7.0, total tion is half of the total phosphorus concentra dissolved solids 200 mg 1’, coliform bacteria tion and ammonia and nitrate nitrogen con 10 per 100 ml, dissolved oxygen 5.0 mg 1—’ centrations are one third and one fourth of and alkalinity 8.0 mg l. The data on the wa the total nitrogen concentration, respectively. ter quantity and quality of the wastewater The nutrient concentrations obtained for the 81

Table 2. Nutrient concentrations of the runoff waters in lowed closely the procedure used eariier with the drainage area of Lake Pyhäjärvi in 1980 and 1981. this model (Kinnunen et al. 1982). Phosphate Ammonia Nitrate The calibration was carried out by trial and Year Season phosphorus nitrogen nitrogen error, by changing the value of one or two pa g 11. 1—1 tg 1—1 rameters at a time and making a computer 1980 Spring 37.5 398 298 run. The computed and observed results were Summer and compared and if necessary new vaiues were Autumn 23.0 276 207 Winter 16.7 given to the parametres to improve the agree 279 209 ment 1981 Spring 14.9 158 118 between the simuiated and observed vai Summer and ues. Autumn 3.4 58 43 Temperature was calibrated first and after Winter 7.9 131 98 about fifteen computer runs the simulated vai ues were in reasonably good agreement with the observed ones. After temperature simula tion had been accomplished the other state runoff are dispiayed in Tabie 2. In addition it was assumed that the dissoived oxygen con variabies were caiibrated. Special attention centration, pH, aikalinity and total dissoived was paid to the caiibration of oxygen, nu soiids concentration of the runoff waters were rients and to some extent of aigae. However, the exact calibration two 11.9 mg 11, 6.0, 6.0 mg i— and 50 mg 11, of the aigal groups respectiveiy. The temperature of the runoff was not possibie because of the iack of phyto water was assumed to be the same as that of piankton biomass observations. After about the surface water of the lake. fifteen additional computer runs the model was considered to be caiibrated to the case study lake. Verification of the model was attempted 3.3 Caiibration and verification with the corresponding data of 1981 using the same values for parameters as were used in the The model was calibrated with the data set of calibration. The agreement between the calcu the open water season of 1980. The observed lated and observed values was found to be data of 22 May, 1980 was used as the initial poor and the model could not be considered data and the simulation was continued untii verified. Therefore, vaiues of five parameters Autumn. The initial values of some state van were changed, after which the agreement was abies, e.g. zoopiankton, aigae and detritus, of the same order as in the caiibration. The had to be estimated because of the iack of ob modei was therefore in effect caiibrated sepa served data. rately with the data sets of 1980 and 1981. At the beginning of the caiibration the vai The values of the parametres that were dif ues of the parameters were the same as those ferent in these two caiibrations are presented used in the simulation of Lake Päijänne. Dur in Tabie 3. This shows that the model is rather ing the course of the calibration the values of sensitive to the parameter values, because the parametres were changed so that the mod small changes in the parameters introduced ei would respond to the observations carried considerabie changes in the model output. out in the lake. The calibration procedure foi Caiibration resuits are presented in Figs. 2—8.

Table 3. Values of parameters having different values in the calibration runs with the data of 1980 and 1981.

Parameter value in the calibration with the data of Parameter 1980 1981 Maximum diffusion coefficient 3Q1Q— 30.106 Phosphorus leaching coefficient 5.0 2.0 Phosphorus precipitation coefficient 0.015 0.15 HaIf saturation constant 0.021 0.017 of phosphate phosphorus of aigal group 1. HaIf saturation constant 0.021 0.017 of phosphate phosphorus for aigal group 2.

6 408401737k 82

m m 29.5.1980 40 40

30 30

E 20 20

10 - 10

0 0 0 6 8 12 16 20°C 4 8 12 16 20°C

m 40 40 12.6.198o_,,••;,r•••••••O•••••Or

30 c E

ii 20 20

10 10

0 0 6 8 12 16 20°C 12 16 20 0C

m m 40 40

30 30 w E 20 20

10 10

0 0 0 4 8 12 16 20°C 0 4 8 12 16 20°C Temperoture Temperoture

Fig. 2. Calibration results of temperature with the data of 1980. 83

m 40

30 c 4) 6 4) ui 20

10

0 0 6 8 12 16 20°C

m m 40 40

30 30 c 4) 6 4) ui 20 20

10 10

.0 0 0 4 8 12 16 20°C 0 4 8 12 16 20°C

m m 40 60

30 30 c 4) 6 4) 1j20 20

10 10

0 0 0 4 8 12 16 20°C 0 4 8 12 16 20°C Temperoture Temperature Fig. 1 Calibration results of temperature with the data of 1981. 84

m 60

30

E 20

10

0 0 4 8 12 16 20°C

m m 40 40

30 30 c E w 20 20

10 10

0 0 0 4 8 12 16 20°C 0 4 8 12 16 20°C Temperoture

m

— 21.9.1981 40

30 c E 20

10

0 0 4 8 12 16 20°C

Temperature -

Fig. 3. Continued 85

III w 16.6.1981 . 7.6.18oZF 60

30

20 - 20 .

10 — 10

0 0 0 2 4 6 8 mgL1 0 2 4 6 8 mg[1

III m 21.71980 60 40 .j 7.198ZJ c 30 30 E 0 ui 20 20

10 10 .-

0 i 1 ‘—1 0 0 2 4 6 8mg11 0 2 4 6 8 mg[1

m w • 10.9.1981 . 40

30 c 0 E - . 0 ui 20

- .

10 — .

- . 0 6 6 0 2 4 6 8 mg11 DissoLved oxygen Dissolved oxygen

Fig. 4. Calibration results of dissolved oxygen with the dätaof198O and 1981. 86

m 40

— 30 30 c w E 0) w 20 20

10 10

0 0 0 400 800 1200 ,LL91

m m 40 40

— 30 30 c E w W 20 20

10 10

0 0 0 400 800 1200 ,gL1 0 400 800 1200 Ag11

m m 60 40

.30 30 c E W 20 20

10 10

0 — 0 0 400 800 1200 ,tg[1 600 800 1200 ,gE1 Ammonio nitrogen Ammonio nitrogen

Fig. 5. Calibration results of ammonia nitrogen with the data of 1980 and 1981. 87

m 17. 6.1980 40

c E UJ 20

10

..

1 UQ LI 100 200 300 ,gr1

m m 21.7. 1980 60 60

30 30 c E w LLJ 20 20

10 10

0 0 0 100 200 300 ,LgL1 0 100 200 300 ,ig

m m 10.9.1981 40 40 —

30 30 c E w w 20 20

10 10

0 0 0 100 200 300 ,g 0 100 200 300 ,ag11 Nitrote nitrogen Nitrote nitrogen

Fig. 6. Calibration results of nitrate nitrogen with the data of 1980 and 1981. 88

E w

m 40

* 30 c E 20

10

0

m 40 r

30 Lj.9.i981

E w ui 20

10

0 40 60 80 100 ,gL 20 40 60 80 100 .ag11 140 Phosphate phosphonis Phosphate phosphorus

Fig. 7. Calibration results of phosphate phosphorus with the data of 1980 and 1981.

4. SENSITIVITY OF THE MODEL TO diffuse Ioading and Ioading entering the lake THE INPUT DATA through precipitation are carried out at much longer time intervais and interpolation of the The values of the forcing functions must be observations is necessary in order to obtain given to the model once during every simula data for each eight hour time step. In pre tion time step, the length of which can be se paring the input data from observations, van lected. In this application the length of the ous computer programs other that that of the simulation time step was eight hours. The ob model must be used. This increases consider servations of the water quality and quantity ably the time and computer resources neces of the tributary, wastewater treatment plants, sary for obtaining the data. If the time inter 89

750 Table 4. Combinations of tributary data and meteorolo Catibratiori 1980 gical data used to test the sensitivity of the model to the input data. 500 Frequency of giving of input data Once every Weekly Monthly

250 - Run No Data 8 hours average average 3 1. Tributary data + Meteorological data +

0 2. Tributary data + Meteorological data + .0 o fbi, 3. Tributary data + Ca[ibration 1981 Meteorological data + 4. Tributary data + •2, 500 Meteorological data + a 5. Tributary data + Meteorological data + 250 — 6. Tributary data + Meteorological data +

Ii Calibration Tributary data + May June JuLy August September run Meteorological data +

Fig. 8. Calibration results of phytoplankton biomass with the data of 1980 and 1981. vai of giving the input data could be increased sentations of such systems, are abstract and e.g. to weekiy or monthly average values the imperfect in comparison with the real ecosys application of the model would become easier. tems which they are intended to represent. The sensitivity of the model to variations Even the most complex of modeis can include in the input data was tested by making six only a fraction of ali the important reactions computer runs with different sets of input da of a water body. Therefore at the present ta. In these runs the tributary data (including stage of model development a model can never the water quality and quantity of the tribu simuiate a water body accurateiy in ali condi tary, wastewater treatment plants, diffuse load ti°ns. and precipitation load) and meteorologicai da Modeis are often evaluated according to ta were given to the model in various combi three basic properties, namely realism, preci nations of weekly or monthly averages (Table sion and generality (Walters 1971). By realism 4). is understood how weil the mathematical The results of these six computer runs equatlons of the model correspond to the re were compared with the results of the calibra actions occurring in nature. Precision implies tion run, in which the input data was given to how accurately the model produces the data the model every eight hours. It was found on which it is based, whiie generality is the that the results of the runs 1,2,3 and 4 were range of appiicability of the model. Conse approximately equai to those of the calibra quently, a model cannot include these three tion run.The results of the runs 5 and 6 pres properties to an equai extent. Typicaliy, mod ented in Figs. 9—14, were also close to the re deis can be built by paying attention to some sults of the caiibration run. of these properties. For example if the objec tive of modeling is to attain understanding of the aquatic system it is more important that the model is capabie of showing the direction of research than that its numericai precision should be high. On the other hand if the ob jective of the model is to develop a tool for 5. DISCUSSIQN water management, precision is generally the goal and the same attention is not given to Water bodies are complex ecological systems reaiism and generality. affected by a great number of interacting proc In the present application the objective was esses. Modeis, which are mathematical repre to apply a model in order to assess its suit 90

m 29 5198,J 40

— 30 c 0) E 0) W 20

10 —

u 1 0 4 8 12 16 20°C

m m 17.1980 40 •12O6O1980,;OLJ 40

— 30 30 c 03 E 0) W 20 20

10 10

4 8 12 16 20°C 0

m 8.9.1980 40

30 c 0) E 0) 20

10

8 12 0 4 8 12 16 20°C Temperature Temperature

Fig. 9. The effect of input data on simulated temperature resuks. O = observations — = calibration run of 1980, input data given once every eight hours = input data given as weekly averages O = input data given as monthly averages 91

m m 17.6.1980 40 40

30 c 5) c E . 5) 5) 2 5) ILJ 20 ui 20-

10 —

0 0 2 4 6 8 mg[1 12 0 400 800 1200,.g11

c 5, c E 5) 53 2 53 ui ui

c 53 2 5) uJ

4 6 400 800 1200,.eg(’ Dissolved oxygen Ammonio nitrogen

Fig. 10. The effect of input data on simulated dissolved Fig. 11. The effect on input data on simu oxygen tesults lated ammonia nitrogen results. O = observations O = observatjons

— — = calibration run of 1980, input data given once = calibration run of 1980, input data every eight hours •givenonce eve eight hours

- - — - - -= input data given as weekly averages = input data given as weekly averages = input data given as monthly averages = input data given as monthy averages 92

Oli O m O :1 17.6.1980 17.6.1980 60 — 40 Oy’

30 .30— c E -. Iii 20 W20— E 10

O 0: 10- 0 0 20 40 50 80 100 140 0

1 1 0 100 200 300

- 1•1 O 21.7.1980 c 40 @0 E aO —..12J w — 30 c 0 E W 20

10

0 100 200 300 ,teg[1 c aO E III aO 8.9.1980 LLI 60 — :‘I .JI.

— 30

E 20 40 60 80 100 ,ag 1 Phosphate phosphorus LiJ 20 Fig. 130 The effect of input data on simulated phosphate phosphorus results 10 O = observations = calibration run of 1980, input data given once every eight hours

- — —= input data given as weekly averages O = input data given as monthly averages 0 100 200 300 ,ug11 Nitrote nitrogen

Fig. 12. The effect of input data on simu lated nitrogen results O = observations = calibration run of 1980, input data given once every eight hours — input data given as weekly averages = input data given as monthly averages 93

the case study area were available, and the dis charge had to be calculated. The main fea tures of the case study area are rapid flow through the lake, i.e. short detention time and consequently weak temperature stratifica tion, high input of nutrients through tribu taries and wastewater treatment plants, high 0 nutrient concentrations and anaerobic hypo 3 limnion. The effect of bottom sediments on -v water quality is considerable,particuiary dur UI ing the 0 anoxic period which lasts typically E 0 from June to August. Occurrence of unequal -o distributed flows can cause considerable c ly 0 horizontal vanations in water quality. Ali c 0 these features make this lake difficult for mod 0• 0 eling work. -c The data for running the model, especialiy 0. the values concerning some state variables should ideally have been observed more fre quently, i.e. the biomass values of phyto piankton were totally missing from the cali bration and verification years. In preparing the data concerning meteorology, hydrology, tributaries and wastewater treatment plants, Nay June July August interpolation had to he used. Frequent use of interpolation introduces an error, but is in Fig. 14. The effect of input data on simulated phyto avoidable in this type of work. Estimation of plankton results the diffuse loading data was based on the in 0—0= aigal group 1, calibration run of 1980 input vestigation of Kauppi (1979). This method is data given once every eight hours

— an efficient of means of caiculating = aigal group 2, calibration run of 1980, input diffuse data given once every eight hours load but requires measurements of field per ci —o = aigal group 1, input data given as weekly centages of the drainage basin, which had to averages be estimated. A — A= aigal group 2, input data given as weekly Despite the shortcomings in observation averages frequency discussed above, the data can in V V = aigal group 1, input data given as monthly averages general be considered adequate for this type V —V= aigal group 2, input data given as monthly of appiication and is seldom more accurate in averages ecological modeling. This implies therefore that the applicability of the model for this case study lake is not primarily a question of ability for predictive purposes. Therefore, the the adaquacy of available data. main emphasis was placed on the numerical Calibration of the model was carried out precision of the model, and calibration was using a trial and error method, i.e. by making continued until the model gave the best over a computer run and comparing the simulated ali agreement with the observations. and observed results, changing the values of Finnish watercourses typically consist of one or two parameters at a time and making a chains of lakes which are in contact with each new run. Altogether 46 runs were needed to other through straits or rivers. In lake mod produce the presented results. The observa eling it is therefore sometimes difficult to tion of the sensitivity of the model made du specify the case study area and the natural ring its earlier application (Kinnunen et al. boundaries of the lake. This problem was also 1982) was verified in this application. The encountered in the present study. Specifica sensitivity of the model to various parameters tion of the case study area was carried out on e.g. haif saturation constants, decay rates of the basis of morphological characteristics of nutrients, growth rates and temperature cor the lake as weli as of flow measurements. No rection rates is obvious. It is also evident that measurements of the output discharge from it is difficult to ohtain laboratory or field mea 94 surements of the correct values of parameters. calibration. This shows that the model is not The number of parameters that can he cali capable of accurately simulating such a dynam brated in the model is about one hundred for ic system as the present case study lake. On only some of which are recommended values the other hand it also shows that the struc avaiiable in the literature. Consequentiy the ture of the model is quite good in representing number of parameters that has to be main the real lake because only five parameters out ly estimated by calibration remains large and of about one hunclred had to be changed. the correct combination of parameters for Methods for estimating parameters in eco producing the best fit with the observations, logicai modeis have been presented e.g. by if such a combiantion exists, is hard to find. Lewis and Nir (1978), Benson (1979 and Seeking this combination requires several Jørgensen et al. (1981). Evaluation of the computer runs and considerable computer re problems encountered in modeling complex sources as well as some intuition on the part aquatic systems has been presented e.g. by the of the user. Canada Center for Inland Waters (1979), by In every model there are certain parameters Simons and Lam (1980) and by Fedra et al. which have a paramount effect on the behav (1981). iour of the model. In this model these are the The test in which the effect of input data parameters affecting advection and diffusion. on the simuiated results was investigated If these parameters are not at the correct level yieided important information. It was found the model does not produce the observed re that giving the meteorological data and the sults with any combination of the remainder quantity and quality of ali inflowing waters of the parameters. Therefore in the calibra such as tributaries, wastewater treatment tion of the FINNECO model it is important plants, diffuse loading and precipitation as first to calibrate the temperature stratification weekly or monthly averages gave nearly as and water stage of the lake. After this has good results for most of the state variables, been accompiished attention can be directed except for phytopiankton biomass as were ob to the calibration of the other state variabies. tained by giving these data once during every Strict division of the order of calibrating the eight hours. This implies that gathering the other state variabies cannot be given. The input data for the model becomes easier and concentrations of nutrients and phytoplank does not require as much computer time as ton biomass are closely interrelated and they was needed erlier. The slight difference be must be calibrated simultaneously, while the tween the simulated results with these two calibration of some other state variabies such sets of input data could be decreased by cali as coliforms, sodiumlignosuiphonate and total bration. However in the present work the dissolved soiids can be calibrated separately model was not calibrated using weekly or because other state variabies are not affected monthly input data. by them. It must he pointed out that part of the Calibration by the trial and error method is tributary input data for every eight hour time not rewarding because there are no definite step was obtained by interpolation on the ba limits or criteria according to which the caii sis of measured weekly or monthly averages. bration should he finished. The user tends to Therefore the precision of this set of input think that the next parameter combination data does not correspond to the data input may give even better results, as is sometimes frequency. Meteorological data, on the other the case, and calibration continues, leading hand, was originaliy obtained as hourly or dai sometimes to better agreement but sometimes iy averages. However, this experiment shows giving poorer results. It is therefore important that the lenght of the time step couid be in to recognize the stage at which the agreement creased from eight hours. This should de between the model output and observations is crease the time needed for preparing the input reasonably good and to stop modeling. data and reduce computer costs. In this work the model was calibrated with Ts the FINNECO model suitable for mak the data of 1980 and gave reasonably good re ing prognoses of the water quaiity of the sults. However, verification could not be ob modelediake? This application shows that at tained with the calibrated model using the da prsent it is not suitäble for this pui-pöse in ta of 1981. After changing the values of five this lake. However, application of the model parameters the agreement between the simu to this lake with complex .hydrology and high lated and observed results was as good as in nutrient concentrations gives valuable infor 95

mation concerning the suitability of the com tekoaltaiden veden laadun simulointiin. Järvi plex model to the simulation of water quality tai tekoallas oletetaan mallissa horisontaali and introduces some of the problems that are suunnassa homogeeniseksi, joten ero veden encountered. In an earlier application of the laadussa ilmenee ainoastaan vertikaalisuunnas FINNECO model better results were obtain sa. Mallin simuloimat tilamuuttujat ovat: ed (Kinnunen et al. 1982). In that study, lämpötila, liuennut happi, biologinen hapen however, the area investigated was more sta kulutus, alkaliniteetti, pH, epäorgaaninen hii ble, with only minor variations in water quali li, epäorgaaninen sedimentti, hiilidioksidi, ty and therefore the model gave better numer ammonium-, nitriitti- ja nitraattityppi, fos ical accurracy. Additional case studies are faattifosfori, epäorgaanisen sedimentin fosfo needed in order to again more information ri, hygieniaindikaattori, natriumlignosulfo concerning the suitability of the FINNECO naatti, kasviplankton ja eläinplankton. Mallin model to various types of lakes. ulkoisia muuttujia ovat säätiedot sekä simu loitavaan järveen laskevien vesien kuten jo kien, jätevedenpuhdistamoiden, sateen ja ha jakuormituksen määrä- ja laatutiedot. Mallissa otetaan huomioon tuulen vettä sekoittava vaikutus ja se soveltuu myös talvikauden si mulointiin. Mallia sovellettiin Tampereen Py ACKNOWLEDGEMENTS häjärveen avovesikausina 1980 ja 1981. Pyhä- järvi on läpivirtausjärvi, johon kohdistuu eri Jorma Niemi gathered the necessary data for tyyppistä jätevesikuormitusta. Kuormitukses running the model, calibrated it and wrote ta johtuen veden laatu on heikko, ravinnepi this paper. Juhani Eloranta operated the toisuudet ovat korkeita ja alusvedessä vallit computer program of the model and wrote va sevat ajoittain hapettomat olosuhteet. Kohde- rious programs for preparing the required da alue asettaa suuret vaatimukset sovellettaval ta sets. le mallille. We would like to thank the personnel of Malli kalibroitiin vuoden 1980 tiedoilla. the Tampere Water District Office, especially Noin 30 tietokoneajon jälkeen lasketut ja ha limnologists Jaakko Keränen and Kirsti Kro vaitut tulokset vastasivat toisiaan tyydyttä gerus, for carring out special field investigati västi. Malli yritettiin verifioida kalibrointiajon ons for obtaining the necessary data for the parametrikombinaatiolla ja vuoden 1981 vas application of the model. Further, we would taavilla tiedoilla, mutta malli ei verifioitu like to thank the head of the Tampere Water nut tyydyttävästi. Muutettaessa viiden para District Office, Altti Luoma, for encourage metrin arvoa saatiin myös tällä aineistolla ment during the course of this work. We are kohtuullisen hyvä yhteensopivuus laskettujen grateful to Mr Michael Bailey for revising the ja havaittujen arvojen välille. Tämä viittaa sii English of the manuscript. hen, että mallin rakenne jäljittelee tyydyttä västi ekosysteemin toimintaa, sillä mallissa on Helsinki, February 1983 kaikkiaan kalibroitavia parametrejä noin 100. Mallia voidaan tällä hetkellä pitää kalibroituna Jorma Niemi, Juhani Eloranta Pyhäjärveen kahden eri vuoden aineistolla, mutta ei verifioituna. Kaikkiaan mallilla teh tiin tässä työssä 46 tietokoneajoa. Lisäksi tutkittiin mallin herkkyyttä ulkoi sille muuttujille, antamalla niiden arvot mallil le kahdeksan tunnin välein, viikkokeskiarvoi na ja kuukausikeskiarvoina. Havaittiin, että mallilla saadaan käytännössä lähes yhtä tarkat LOPPUTIIVISTELMÄ tulokset antamalla ulkoiset muuttujat kuu kausikeskiarvoina, lukuunottamatta kasvi FINNECO malli on yksidimensioinen veden planktontuloksia. jos ulkoisten muuttujien laatua simuloiva malli, joka kehitettiin vesihal arvot annetaan mallille harvemmin kuin ker lituksen ja Oy Suomen International Business ran kahdeksassa tunnissa se helpottaa mallin Machines Ab:n välisessä yhteisprojektissa vaatimien tietojen muokkausta ja säästää tie vuosina 1978—1981. Malli soveltuu järvien ja tokoneaikaa. 96

Pyhäjärvi osoittautui vaikeaksi for the lake model FINNECO and the river model soveltamis QUAL II. Mimeographed publications of the National kohteeksi mallille, eikä malli toiminut siinä Board of Waters, Finland. 425 p. niin hyvin kuin Päijännesovellutuksessa (Kin Kinnunen, K., Nyholm, B., Niemi, J.S., Frisk, T., Kylä nunen et al. 1982). Tulokseen vaikuttavat Harakka, T. & Kauranne, T. 1982. Water quality mm. järven virtausolosuhteet, hapellisten ja modeling of Finnish Water Bodies. Publications of the Water Research Institute, Finland, 46. 99 p. hapettomien vesimassojen vaihtelu, heikko Kuparinen, J. 1981. Sulfuttiselluloosatehtaan jätevesien ja lämpötilakerrostuneisuus sekä altaasta pois eräiden raskasmetalli- ja orgaanisten yhdisteiden vaiku virtaavan vesimäärän arvioiminen. Sovellutus tus vesistön bakteeritoimintaan. (The effects of the wastewaters of sulphite mill, some heavy metals and osoitti, että malli jäljittelee tyydyttävästi organic compounds on the bacterial activity in the wa luonnon ekosysteemin toimintaa, mutta että ter bodies). Publications of the National Board of Wa sen numeerinen tarkkuus ei ole riittävä tällä ters, Finland, 204. 55 p. kohdealueella. Lepistö, L., Kokkonen, P. & Puumala, R. 1979. Kasvi planktonin määristä ja koostumuksesta Vuoksen, Ky mijoen ja Kokemäenjoen vesistöalueilla kesällä 1971. (On the quantity and composition of phytoplankton biomass in the drainage bains of the rivers Vuoksi, Kymijoki and Kokemäenjoki in the summer 1971). Pu blications of the National Board of Waters, Finland, 172. 250 p. Lewis, S. & Nir, A. 1978. A study of parameter estimati on procedures of a model for lake phosphorus dyna REFERENCES mics. Ecol. Modelling 4: 99—117. Meteorological Yearbook of Finland. 1981. Climatological Benson, M. 1979. Parameter fitting in dynamic models. data 1980. Meteorological Yearbook of Finland, vol 80 EcoI. Modelling 6: 97—115. part 1a—1980. Published by the Finnish Meteorologi Canada Centre for Inland Waters. 1979. Assessment of cal Institute. 115 p. water quality simulation capability for Lake Ontario. Niemi, J.S. 1978. Application of an ecological simulation Scientific series No. 111. 220 p. model to Lake Päijänne. Publications of the Water Re Fedra, K., van Straten, G. & Beck, B.M. 1981 Uncertain search Institute, Finland, 28. 39 p. ty and arbitrariness in ecosystems modelling: a lake Sarkkula, J. & Forsius, J. 1977. Tampereen Pyhäjärven modelling example. Ecol. Modelling 13: 87—110. virtaustutkimus talvella 1977. (Flow measurements in Järvinen, 0. & Haapala, K. 1980. Sadeveden laatu Suo Lake Pyhäjärvi in the winter 1977). National Board of messa 1971—1977. Summary: The quality of wet and Waters, Finland, Hydrological Office. Mimeograph. 90 dry deposition in Finland according to observations made frorn 1971 to 1977. Publications of the National Simons, T.J. & Lam, D.C.L. 1980. Some limitations of Board of Waters, Finland, 198. 102 p. water quality models for large lakes: a case study of Jørgensen, S.E., Jørgensen. L.A., Kamp-Nielsen, L. & Lake Ontanio. Water Resour. Res. 16: 105—116. Mejer, H.F. 1981. Parameter estirnation in eutrophica Walters, C.J. 1971. Systems ecology: the systems appro tion modelling. Ecol. Modelling 13: 111—129. ach and mathematical models in ecology. In: Funda Kauppi, L. 1979. Effect of drainage basin characteristics mentals of Ecology pp 276—292. E.P. Odum (ed.). on the diffuse load of phosphorus and nitrogen. Pubii W.B. Sounders Company. 574 p. cations on the Water Research Institute, Finland, 30: Virtanen, M., Forsius, J. & Sarkkula, J. 1979. Measured 21—41. and modelled currents of a lake. Aqua Fennica 9: 3— Kauranne, T. 1983. Computer program documentation 15.