X - William W. Walker, Jr., Ph.D.

VARIABILITY OF TROPHIC STATE INDICATORS IN VERMONT LAKESAND IMPLICATIONS FOR LEAP ERROR ANALYSESprepared forVermont Agency of Environmental ConservationWater Quality DivisionLakes ProgramMontpelier, VermontbyWilliam W. Walker, Jr.Environmental Engineer1127 Lowell RoadConcord, MA 01742617-369-8061August 1983

This report develops refined estimates of LEAP error terms, basedupon the general approach outlined previously (Walker, W.W., "Structureand Calibration of an Error Analysis Framework for the Vermont LakeEutrophication Analysis Procedure", 1982). The major refinementincorporated below is consideration of the effects of within-year andamong-year variability in observed lake water quality on the accuraciesof observed mean values. Results provide a useful framework fordesigning monitoring programs as well as a basis for estimation of LEAPmodel error terms.Lay monitoring data from the 98 STORET stations listed in theAppendix have been used as a basis for the analysis. Since variationsamong stations within a given lake would depend upon site-specificcharacteristics (morphometry, nutrient loading distributions, stationslocations, etc.), each station is treated separately below. Thispermits a focus on temporal variance components. LEAP is designed topredict long-term average conditions and temporal variability imposeslimitations on the accuracy of observed mean values derived from limitedsampling data.Table 1 presents results of nested analyses of variance for springtotal P, summer chlorophyll-a, and summer transparency. The conceptualmodel described previously divides measurement variance into threecomponents: (1) among lakes (stations); (2) among years, within lakes;and (3) among samples, within years. Because spring phosphorusmeasurements were generally conducted only once per year, it isimpossible to distinguish components (2) and (3) and they are lumped inTable 1. Within-year variations in phosphorus are treated separatelybelow. Two sets of ANOVA's are given in Table 1. One uses uses allnon-missing observations for each variable and examines all threevariance components. The second is based upon station-year means andincludes only station-years with non-missing values for phosphorus,chlorophyll-a, and transprency; this permits assessment of covariancecomponents (Table 2).The analysis indicates that among-lake (or among-station) variancedominates all three measurements and accounts for 73%, 56%, and 76% of

the total observed variance m phosphorus, chlorophyll, and transparency*when all data are considered. The among-sample variance component isstrongest in the case of chlorophyll-a (.19 or 33%); this reflectsseasonal variations in algal populations, as well as sampling andanalytical errors, and indicates, for example, that an average of 19samplesper summer would be required to estimate a yearly-mean value towithin a coefficient of variation of 10%. Diagnostic plots indicatethat the among-sample standard deviation in log(chlorophyll) increaseswith mean chlorophyll-a level from about .38 to .55; while this mayreflect a tendancy for algal populations to be more variable in moreeutrophic systems, the pooled estimate (.44) is adequate for the presentpurposes. Covariance among the three variables is generally strongeramong stations than among years (Table 2). The year-to-year componentis more strongly influenced by measurement and sampling errors, whichwould not be expected to correlate as strongly across variables.Another possible factor is that wet years may tend to have higherphosphorus levels (because of increased runoff) but lower chlorophyllconcentrations (because of lower light intensities and temperatures).varianceLimited data from four lakes are available to estimatewithin-yearin spring phosphorus concentrations (Table 3). Based upon theaverage coefficient of variation, a pooled estimate for thewithin-yearcomponent is .063 (natural log scales) and accounts for 66% of theyear/sample variance component estimated in Table 1.Table 4 summarizes the estimated variance components for eachvariable and computes estimated variances, coefficients of variation,and 90% confidence factors for long-term means of each variable as afunction of monitoring years. The within-year sampling regime (m)corresponds approximately to the lay monitoring design (1 phosphorus, 12chlorophyll, and 12 secchi measurements per year). The variancecomponents and formula can be used to assess alternative samplingdesigns. Generally, the variance in the long-term mean is greatest forphosphorus because only one sample is taken per year and the within-yearvariance component is appreciable. Increasing within-year samplingfrequency (even to 2 samples/year) may be cost—effective for determining-2-

long-term means in spring phosphorus.As shown in Table 5, a portion of the year-to-year variance in eachmeasurement is not random, but associated with specific years, i.e., theyearly effects are correlated across stations. Fixed yearly effects arestatistically significant for each variable and account for a total of23.1%, 3.3%, and 2.0% of the total within-station variance ofphosphorus, chlorophyll, and transparency. They are of practicalsignificance only in the case of phosphorus. Duncan's Multiple RangeTest applied to yearly means suggests an increasing trend in phosphorusranging from a mean of 8.2 in 1979 to 12.3 in 1982. Except for theslightly lower chlorophyll-a mean in 1979 (3.82 V6. mean of 4.10 forother years), the apparent trend in phosphorus is not correlated withtrends in the other measurements.Applied to the total estimated within-station phosphorus variancecomponent of .095 (V(Y) + V(W) in Table 4), the fixed yearly effectsaccount for 23.1% or .022; this, in turn, accounts for 68.8% of theestimated among—year variance component for phosphorus (.032). Thus,about two thirds of the year-to-year variance in spring phosphorusduring these four years was not random but associated with specificyears. This suggests a deterministic component which may be related tohydrologic or climatologic variations which influence all lakes, changesin laboratory or sampling procedures, or actual trends in concentrationattributed to increases in nutrient loadings. Additional analysis(particularly vs. streamflow data) may shed additional light on the thesources of these variations. The non-random nature of the year-to-yearphosphorus variations raises a question, however, as to theapplicability of the estimated variance components in future years andwarrants continued study. In applying these results to estimate errorterms for future LEAF applications, we are assuming the the lakeconditions and sampling procedures in 1979-82 period were reasonablyrepresentative.The above results can be used to estimate the model and data errorcomponents of LEAF submodels following the general approach outlinedpreviously. Table 6 lists the estimated data error variances for the 18-3-

calibration lakes, based upon the variance components listed in Table 4and the average monitoring program designs for these lakes. Table 7estimates the model error variances by difference from the total anddata error variance, considering the effects of error propagationthrough the model network. The total error variances (V(R)) in Table 7correspond to use of the linear form of the LEAP internal loadingfunction. While they are slightly greater than the errors derived fromthe exponential function, the linear function is recommended because ofsensitivity considerations. Resulting input standard deviations for themodel error terms in LEAF are listed in Table 8.Table 9 lists observed and predicted phosphorus concentrations forLEAF applications to 18 calibration lakes and the 24 additional lakescompiled by VDWR. Monitoring years and estimated variance componentsare also listed. A major distinction between the first and second setof lakes is apparent in terms of sampling frequency; the latter weregenerally sampled less frequently and LEAP error variance would beexpected to be greater for these lakes. Based upon the error termsestimated above, observations and predictions differ by more than twostandard errors for Morey in the first group. Under-prediction in thiscase is attributed primarily to unusual iron/sulfur chemistry leading toenhanced internal recycling of phosphorus. The effects of unusualoutlet hydraulic conditions leading to increased phosphorus loadings maybe responsible for the under-prediction in the case of Harvey's Lake (t• 1.99). In the second group of lakes, t values exceed 2.0 in 3 cases.Northeast Developer's "pond" is a relatively new, artificial impoundmentand is atypical of the natural lakes used in model development.Internal recycling may be more efficient than predicted in Valleybecause the lake is permanently stratified. The deviation of Hosmer(t=2.7) is unexplained.Excluding Northeast Developer's and Valley on the basis of theiratypical characteristics, the total error variance in the second set oflakes is .111. This compares well with the average data and modelvariance components (Vtotal in Table 9) of .108 for the same set oflakes. Calculated error statistics for the second set of lakes support

the error terms estimated from the calibration lakes and indicate thatthe general magnitudes of deviations in the second set are expectedbased upon sampling frequency and model error considerations. Thenumbers of lakes in the second set with chlorophyll and transparencydata are inadequate as a basis to test the error terms for thesevariables.One limitation of the error analysis is that the estimated variancecomponents and their effects on the accuracies of lake-mean valuesassume that the samples are statistically independent. Some autocorrelationwould be expected in chlorophyll and transparency samplestaken at a weekly frequency. Additional analysis would be required toquantify this and its effects on the error analysis. Generally, serialcorrelation would tend to become increasingly important at relativelyhigh sampling freqencies and would tend to decrease the value of highfrequencysampling for determination of summer-mean values. Serialcorrelation is not likely to influence the spring phosphorus estimates,however.Minor code modifications and procedures to permit consideration oferrors in the observed means in LEAP applications have been describedpreviously. By considering both model and data error components, LEAPcan be used to identify lakes which have unusual watershed, loading, orinternal recycling characteristics leading to high phosphorus levels.These would be logical candidates for further direct investigation.Deviations of Harvey's, Morey, Valley, and Northeast Developer's are allpositive and explained independently of the model. More detailedinvestigation of Hosmer may lead to an explanation for its deviation(also positive). In this sense, LEAP can be used as a screening tool toidentify problem lakes.Another type of application would involve projecting impacts ofchanges in nutrient loadings and/or land use on water quality. If dataare available for the lake in question, the model error terms (means)can be calibrated so that water quality observations and predictionsmatch, similar to the approach demonstrated for Lake Morey. If themodel is calibrated in this way, however, the error term standard-5-

deviations estimated above are nocalibration, projections would be moreupon the error terms. While thereerror in this type of application, thebe used to assess the effects ofvariables on predictions.longer valid* Because of theaccurate than predicted basedis no good way of assessing modelerror analysis routine can stilluncertainty in the model input

Table 1Results of Nested Analyses of VarianceDOFAllMSDataVCPercentDOFYearly -Means*MS VCPercentSpring Total PTotal 285Station 84Year/Sample 201.4511.241.122.456.334.122100.073.126.91345777.355.705.095.355.265.095100.074.026.1Chlorophyll-aTotal 2620Station 64Year 87Sample 2469.57314.2881.236.191.578.324.058.191100.056.310.433.31345777.414.875.069.419.345- .069100.083.117.0Secchi DepthTotal 3521Station 93Year 203Sample 3225.2768.300.302.045.281.212.022.045100.076.18.015.91345777.255.557.027.255.228.027100.090.010.3DOF « Degrees of FreedomMS m Mean SquareVC = Variance ComponentPercent - VC as Percent of Totalall components on natural log scalesincluding only station-years with non-missing values forall variables

Table 2Nested Analysis of CovarianceDOF VC-X VC-Y CC-XY R Bg BrX - Chl-aY«= Secchi -including within-yearvariationsTotalStationYearSample1919 .60764 .35387 .0561768 .198.342.267.027.049-.251-.210-.013-.032-.551-.685-.346-.324-.75-.87-.69-.50-.41-.60-.24-.16X - Chl-aY • Secchi(yearlymeans)TotalStationYear134 .41757 .34677 .071.256.229.027-.235-.223-.014-.720-.794-.335-.78-.82-.61-.56-.65-.20X •- Total PY - Chl-a(yearlymeans)TotalStationYear134 .35757 .26577 .093.417.346.071.237.221.016.613.730.1991.081.14.87.66.83.17X • Total P Y •= Secchi (yearly means)TotalStationYear1345777.357.265.093.256.229.027-.115-.120.003-.379-.486.069-.85-.93.53-.32-.45.03DOF = Degrees of FreedomVC-X, VC-Y « Variance Component of X and YCC-XY • Covariance Component of X and YR • Covariance Component CorrelationBg •= geometric regression slopeBr = regression slopeall components on natural log scales

Table 3Within-Year Variance of Spring Phosphorusin Four Vermont LakesLake Year NStandard Coef. ofMean Deviation VariationHarveysKeiserMoreyPeacham797979799810811.75.536.26.32.31.97.81.6.20.34.22.25mean .25Estimated Within-Year Variance Component(natural log scales).25 •= .063

Table 4Estimation of Variances in Observed Means as a Functionof Sample YearsSample Variances of Means* Coefs. of Variation 90% Confidence Fac.Years P Chl-a Secchi P Chl-a Secchi P Chl-a Secchi12345678910.0950.0475.0317.0238.0190.0158.0136.0119.0106.0095.0739.0370.0246.0185.0148.0123.0106.0092.0082.0074.0258.0129.0086.0064.0052.0043.0037.0032.0029.0026.3082.2179.1780.1541.1378.1258.1165.1090.1027.0975.2719.1922.1570.1359.1216.1110.1028.0961.0906.0860.1605.1135.0926.0802.0718.0655.0607.0567.0535.05071.8521.5461.4271.3611.3171.2861.2621.2441.2281.2151.7221.4691.3691.3121.2751.2491.2281.2121.1991.1881.3781.2551.2041.1741.1541.1401.1291.1201.1131.107* natural log scalesCoef. of Variation ••.5Variance90% Confidence Factor » F90 » exp( 2 CV)Ye/F90 < Y < Ye x F90 at 90% confidence levelY = actual mean, Ye • estimated meanVariance of Means Computed from:where,V(0) - V(Y)/n + V(W)/ (n m)V(0) •• variance of observed long-term meanV(Y) «= between-year variance componentV(W) = within-year variance componentn = number of yearsm "= number of samples per yearVariable Chl-a SecchiV(Y)V(W)m.032 .058 .022.063 .191 .0451 12 12

Table 5Test for Fixed Yearly Effects on Trophic State IndicatorsTotal P Chlorophyll-a TransparencySource DOF SS DOF SS DOF SSTotal 285 24.74 2620 282.27 3521 184.14Station 84 19.66** 64 172.39** 93 145.57**Year 3 1.07** 3 3.58** 3 .76**Error 198 3.56 2553 106.30 3425 37.81Year/(Error+Year) 23.1% 3.3% 2.0%Fixed Yearly Means:1982 1.090 A*1981 1.035 B1980 .955 C1979 .915 CRange .180.607 A.619 A.613 A.582 B.025.661 A.648 B.646 B.652 AB.015DOF " degrees of freedomSS = sum of squares** Effect significant at p < .01* Means followed by same letter are not significantly differentat p < .05, based upon Duncan's multiple range testall statistics on base-10 log scales

Table 6Estimated Data Error Components for LEAP Calibration LakesVariable V(Y) n V(W) m V(0)Spring P .032 5 .063 1 .019Mean Chlorophyll-a .058 4 .191 12 .019Maximum Chlorophyll-a * .083 4 .021Mean Transparency .022 4 .045 12 .006Oxygen Depletion Rate * .020 2 .010V(Y) = year-to-year variance componentn = average number sample-years per lakeV(W) • within-year variance componentm = average number os samples/yearV(0) • error variance in mean estimate » V(Y) / n + V(n)/ (n m)* developed previously (Walker ,1982)all variances expressed on natural log scales

Table 7Model Error ComponentsModel Error TermV(R) V(0) V(M)V(M) Components39/40 Spring Phosphorus .075 .019 .056 .028 (Watershed, 39).028 (Retention, 40)41 Phosphorus/Chl-a .197 .019 .178 .94 x .056 (Spring P).129 (P/Chl-a, 41)42 Mean Chl-a/Max Chl-a .287 .021 .266 1.16 x .178 (Mean Chl-a).026 (Mean/Max Chl-a, 42)43 Mean Chl-a/Secchi .107 .006 .101 .65 x .178 (Mean Chl-a).026 (Mean Chl-a/Secchi, 43)44 Phosphorus/HOD .105 .010 .095 .91 x .056 (Spring P).049 (P/HOD, 44)V(R) « total variance of residuals, excluding Star Lake (outlier) formean and maximum chlorophyll-aV(0) • average observed data error component (Table ?)V(M) • estimated total model error » V(R) - V(0)V(M) Components = breakdown of model error, sum BV(M)All variances on natural log scale

Table 8Summary of Calibrated Error Terms for LEAF InputStandardVariable Mean Deviation39 Watershed Model Error40 Retention Model Error41 F/Chl-a Model Error42 Mean Chl-a/Max Chl^a Model Error43 Mean Chl-a/Secchi Model Error44 Phosphorus/HOD Model ErrorNotes:1.01.01.01.01.01.0.17.17.36.16.16.21Standard Deviations estimated from square roots ofvariance terms estimated in Table ?;Standard Deviations of input variables 1-38should be set to 0.0.

* rt *< W O wrt ,-s fD oi cr Hn O O P rt 01 Mfl> O rt ^ HM D 4M P CO• •OHH OJ H p- n It IIO fD H n nH H OfD Otn fD rtP Prt po> p.PMP.< CD(D rt» H>rt 05* wP n>rt ex.® ..P- < >iP H-H PP- PP f>P fDOfD OM>OM> OCTH 0)fD fD?.gP. fDe aPp» nsn n03N>^K3l-'*-^rotOh-'h-K5U>K5O O O O O O O O O O O O O O O O O O O O O O O OO O O O O O O O O O O O O O O O O O O O O O O Ot0P > O>l>>4SvovO4sis>rOW.p-vo^v£>tois>.p>4>-vOvO4Nl>>Js4> , X>NJN500UiUiOO^*-K5COVjiOOCn*-4^0000Ui ^ H N W O C « W 0 0 l - , O v 0 « I I O M N O v l W 0 ) OrC p* P p- rt cr rt P -- - O - • O - P - - - - - - fD -P P M p p fD H n p. 01 H-> o HcrP* n K" fl) S3fD V!HO fDB rt H P •*!P' 01H crpP fDfD rtI Io o oI I I IoooooooooooooO N * " O H U O N U I O H U I H O H HUOO^O x O ^ v O P U i P ^ U I N l W O i OP H N W M * - N N W W ( O I O M M I O kC6-*>l'j>>rOCri*-J4s^'N>l>>£»cr>voui4!»OOPPPUiPNIO\WvOvOUiS>tOKIV0 4MPvOWCMfifrfrOiONvlOO^frOOHH | O M W I O * > M M N W M M W N M N• M O N J M ^ U ^ - « O ^ U W O P « > C T > O V4S't- , K»'>JrOOiUi^«l4>L«>OCO|sj6500VOO O ^ t O I O O ^ U i P U H U W H P ^ O^UlUlUI^-UiUUIf^UiUIUI'D'UlUiUlooooooooooooooooooooooooooooooooM H H M I s J P H H P P H P l O P P H•P>vO\OSO*VOVOVDooooooooooooooooooooooooooooooooOUlUIUlOUiUlUiMUiUiUiOUlUlUiI I I I I IO l - ' l - ' O O P - ' O O r O O O I - ' O O O OK>COrO

APPENDIXSTORET Station Codes Used in AnalysisPRIM. SEC. BASIN LAKE503776 9ARR1 LAMOILLE RIVER503777 9ARR2 LAMOILLE RIVER503778 9ARR3 LAMOILLE RIVER503779 9ARR4 LAMOILLE RIVER504813 9AVB1 ST. FRANCIS RIVER504814 9AVB2 ST. FRANCIS RIVER503105 9BEE1 CASTLETON RIVER503117 9BEE2 CASTLETON RIVER503006 9BIG1 WALLOOMSAC RIVER503007 9BIG2 WALLOOMSAC RIVER503112 9B0M1 CASTLETON RIVER503109 9BOM2 CASTLETON RIVER503089 9BUR1 HUBBARDTON RIVER503118 9BUR2 HUBBARDTON RIVER503512 9CAR1 PIKE RIVER503526 9CAR2 PIKE RIVER503784 9CAS1 GREENSBORO BROOK503787 9CAS2 GREENSBORO BROOK504254 9COL1 WEST RIVER504255 9COL2 WEST RIVER504939 9DER1 CLYDE RIVER504850 9DER2 CLYDE RIVER503179 9DUN1 LEICESTER RIVER503182 9DUN2 LEICESTER RIVER503180 9DUN3 LEICESTER RIVER504926 9ECH1 CLYDE RIVER504858 9ECH2 CLYDE RIVER503796 9ELM1 LAMOILLE RIVER503797 9ELM2 LAMOILLE RIVER503193 9FER1 OTTER CREEK503194 9FER2 OTTER CREEK503680 9FRD1 BLACK CREEK503681 9FRD2 BLACK CREEK504533 9FRL1 OMPOMPANOOSUC RIVER504532 9FRL2 OMPOMPANOOSUC RIVER503080 9GLE1 OTTER CREEK503079 9GLE2 OTTER CREEK503898 9GRE1 KINGSBURY BRANCH503894 9GRE2 KINGSBURY BROOK504536 9GR01 WELLS RIVER504537 9GR02 WELLS RIVER504702 9HAL1 CONNECTICUT RIVER504703 9HAL2 CONNECTICUT RIVER504526 9HAR1 PEACHAM BROOK504540 9HAR2 PEACHAM BROOK503082 9HOR1 HUBBARDTON RIVER503091 9HOR2 HUBBARDTON RIVERARROWHEAD MTAVERILLBEEBEBIG PONDBOMOSEENBURRCARMICASPIANCOLEDERBYDUNMOREECHOELMOREFERNFAIRFIELDFAIRLEEGLENGREENWOODGROTONHALLSHARVEYSHORTONIA

PRIM. SEC. BASIN LAKE503507 9IR01503537 9IR02503508 9IR03504863 9ISL1504857 9ISL2504625 9J0E1504626 9J0E2504701 9MAI1504709 9MAI2504541 9MAR1504529 9MAR2503687 9MET1503697 9MET2504705 9MOR1504707 9M0R2503895 9NEL1503897 9NEL2504153 9NIN1504158 9NXH2504905 9PAR1504908 9PAR2503141 9PIN1503142 9PIN2503003 9PRN1LAPLATTE RIVERLAPLATTE RIVERLAPLATTE RIVERCLYDE RIVERCLYDE RIVERJOES BROOKJOES BROOKPAULS STREAMPAULS STREAMSTEVENS RIVERSTEVENS RIVERMISSISSQUOI RIVERMISSISSQUOI RIVERCONNECTICUT RIVERCONNECTICUT RIVERKINGSBURY BRANCHKINGSBURY BRANCHBLACK RIVERBLACK RIVERROARING BROOKROARING BROOKOTTAUQUECHEE RIVEROTTAUQUECHEE RIVERWALLOOMSAC RIVER503004 9PRN2 WALLOOMSAC RIVER504356 9RAP1 N. BRANCH DEERFIELE » RAPONDA504359 9RAP2 N. BRANCH DEERFIELI >504155 9RES1 BLACK RIVER RESCUE504157 9RES2 BLACK RIVER504869 9SAL1 CLYDE RIVER SALEM504865 9SAL2 CLYDE RIVER504889 9SEY1 ECHO LAKESEYMOUR504872 9SEY2 ECHO LAKE504915 9SHA1 BARTON RIVER SHADOW504843 9SHA2 BARTON RIVER503084 9SNS1 HUBBARDTON RIVER SUNSET503104 9SNS2 HUBBARDTON RIVER503183 9STA1 MILL RIVERSTAR503184 9STA2 MILL RIVER503094 9STC1 MILL BROOK503116 9STC2 MILL BROOK503899 9VAL1 KINGSBURY BRANCH VALLEY503900 9VAL2 KINGSBURY BRANCH504922 9WIL1504920 9WIL2503887 9W001503888 9W002504151 9W0W1504160 9W0W2BARTON RIVERBARTON RIVERKINGSBURY BRANCHKINGSBURY BRANCHOTTAUQUECHEE RIVEROTTAUQUECHEE RIVERIROQUOISISLAND PONDJOESMAIDSTONEMARTINSMETCALFMOREYNELSONNINEVAHPARKERPINNEOPARANWILLOUGHBYWOODBURYWOODWARD>

STRUCTURE AND CALIBRATION OF AN ERROR ANALYSIS FRAMEWORK FORTHE VERMONT LAKE EUTROPHICATION ANALYSIS PROCEDUREprepared forVermont Agency of Environmental ConservationWater Quality DivisionLakes ProgramMontpelier, VermontbyWilliam W. Walker, Jr.Environmental Engineer1127 Lowell RoadConcord, MA 01742617-369-8061November 1982

The structure, calibration, and use of the error calculationroutine in LEAP are based upon first-order error analysis concepts. An"error" is defined as a "residual", or difference between an observedand predicted lake condition (phosphorus, chlorophyll-a, transparency,or oxygen depletion rate). Figure 1 summarizes the pathways involved incalculating the mean and variance of each lake response variable andresidual. Because both the measurement and the model errors tend to bemultiplicative and to increase with the corresponding mean estimate,they are best expressed on a logarithmic scale. According tofundamental error analysis concepts, the total residual variance can bepartitioned into the following sources:where,V(R) = V(I) + V(M) + V(0)V(R) = total residual varianceV(I) = variance attributed to uncertainty in model inputvariables and parameter estimatesV(M) = variance attributed to model errorV(0) = variance attributed to- measurement errors in theobserved lake response variableNote that the variance of a predicted value is given by the sum of thefirst two terms and that the last term is included only when comparingobserved and predicted lake conditions. In the Vermont LEAP subroutine,V(l) terms are calculated from the standard deviations specified foreach input variable (subscripts 1 - 38). These values should reflectthe uncertainty in each input variable, which would tend to be lake- anddata-specific. V(M-) terms are calculated from the standard deviationsspecified for each model error variable (subscripts 39 - 44) and wouldtend to be constant across lakes.For these types of models, previous work (Walker, 1977, 1982b)indicates that the error "balance" tends to be dominated by the model(V(M)) and data (V(0)) terms. Because the V(I) terms are lake- andinformation-specific, it is necessary initially to ignore them incalibrating the other terms based upon observed and predicted lake1

Figure 1LEAP PathwaysPredicted Lake ConditionsInputVariables.^Spring PMax Chl-ar p Mean Chl-a1_^ Residuals•TransparencyOxygen DepletionObserved LakeConditions

conditions. The V(I) terms can still be included in LEAP applications,although, in general, they will be relatively small.LEAP predictions correspond to long-term-average lake conditions.V(0) errors reflect errors in observed lake conditions based uponlimited sampling. In order to estimate the. V(0) terms, it is convenientto employ the following nested analysis of variance (Snedecor andCochran, 1972):V(0) = V(Y) / n + V(W) / ( n m )where,V(Y) = between-year variance componentV(W) = within-year variance componentn = number of sampling yearsm = average number of observations within each sample-yearEach of the above terms is specific for each observed variable. Thesecond term in the above equation incorporates analytical error, fieldsample collection error, and statistical sampling error; the last isattributed to temporal variablity in lake conditions within a givenyear. For an "adequate" sampling design, the second term should besmall in relation to the first; this is because "m" is relatively large.For example, a pooled estimate of V(Y) for chlorophyll in Vermont Lakesis .074 on a natural log scale (Walker, 1981). Previous analysis ofwithin-year variance in chlorophyll at 258 stations in Corps of Engineerreservoirs suggests a typical V(W) value of .32 (Walker, 1982c); thisvalue is conservative (high) because it is based upon a sampling regimewhich includes some spring and fall samples. For a one-year, weekly,June-August, sampling regime (n=l,m=12), terms of the above equation are.074 and .027, respectively. Definition of a within-year variancecomponent for spring phosphorus is indeterminant. Because of theseconsiderations, variance in the observations is approximatelyrepresented by considering V(Y) only, although this could be refined byanalysis of time series data from Vermont lakes. Table 1 summarizespooled estimates of V(Y), n, and V(0) for each response variable, based2

Table 1Pooled Estimates of Year-to-Year Variancein Observed Lake ConditionsVariableV(Y)nV(0)Spring P.0805.016Mean Chlorophyll-a.0744.019Maximum Chlorophyll-a.0834.021Transparency.0354.009Oxygen Depletion Rate.0202.010V(Y) = pooled estimate of year-to-year variancen = average number sample-years per lakeV(0) = error variance in mean estimate = V(Y) / nall variances expressed in natural logarithmsbased upon ANOVA's conducted on LEAP calibration lakes(Walker,1981b)

upon analyses of variance for the LEA? calibration lakes (Walker, 1981).Figure 2 summarizes LEAP residual distributions for the calibrationlakes. Given V(0) estimates, the model error terms can be estimated bydifference from the residual error variance. In this exercise, modelerror terms must be "tracked" through the network. For example, themodel error for chlorophyll is the cumulative effect of model errors inpredicting watershed phosphorus export, lake phosphorus retention, andthe lake phosphorus/chlorophyll relationship. For multiplicativemodels, the model error variance terms are additive on a logarithmicscale, although average sensitivity or slope must also be consideredusing the standard first-order error analysis procedure. For example,the chlorophyll-a model error is tracked using the following scheme:where,log(B) = a + b log(P)2V(MB) = b V(MP) + V(MB/P)a,b = regression coefficients for phosphorus/chlorophyllrelationshipV(MB) = total model error for chlorophyllV(MP) = total model error for spring phosphorusV(MB/P) = model error term for phosphorus/chlorophyllrelationshipFor models of more complex structure, an average sensitivity (b) isemployed in variance tracking, based upon analytic differentiation ofthe model at average values of the lake response variables.Results of the error tracking exercise are summarized in Table 2.Because observed phosphorus loadings are not available for Vermontlakes, it is not possible to separate the spring phosphorus model errorinto its two components (watershed export and lake phosphorusretention). In this case, the variance is assumed to be partitionedequally between the two sources; this assumption is of no consequenceto model applications because the variance in predicted lake conditionsreflects only the sum of these error terms and not their relative3

-psprminimum of interval0.6000.400 Hv Mo0.200 Sh Bo0.000 St Ho Ce-0.200 Wi Ir Fa-0.400 Pa Sh Ha-0.600Su CaCuFigure 2LEAP Error DistributionsSC Elfrequencyint cum 0 18182 167 144 73 30 0percentageint cum0.0 100.011.1 100.011.1 88.938.9 77.822.2 38.916.7 16.70.0 0.0univariate statistics for: r-psprnumber of cases =mean =variance =std deviation =mean square =coef of variationstd error of meant-test for m=0 =skewness coef. =kurtosis coef. =sum of squares •-av absolute dev -18.070957.0659931.256891.0673617» 3.62038= .06054981.17188.190726-.6537191.21251.197199five lowest values16 Halls -0.36817 Shadow -0.27511 Parker -0.2254 Curtis -0.1866 Fairfield -0.139r-chlminimum of interval1.4001.200 St1.000 Ca0.800 Sh0.6000.4000.200 Mo-0.000 Sh Bo Fa Ha Hv Cu-0.200 Ho Ir-0.400 El Pa-0.600 SC Suunivariate statistics for: r-chlnumber of cases = 16mean = .147021variance = .285535std deviation = .534354mean square = .289304coef or variation = 3.63456std error of mean = .133589t-test for m=0 = 1.10055skewness coef. = 1.09054kurtosis coef. = .321237sum of squares = 4.62886av absolute dev - .382537five lowest values16 Sunset -0.55011 SC -0.41510 Parker -0.2614 Elmore -0.2018 Iroquois -0.171missing values = 0maximum = .563368upper hinge = .228787median = .0634172lower hinge = -.150874minimum = -.367914h-spread = .379661prob(>t) = .256484large-sample z = .330347large-sample z = -.566137cor. sum of sqs = 1.12188av absolute value = .208169five highest values14 Star 0.1991 Bomoseen 0.31913 Shelburne 0.36710 Mofey 0.4807 Hveys 0.563missing values =maximum =upper hinge =median =lower hinge =minimum =h-spread =prob(>t) =large-sample z =large-sample z =cor. sum of sqs =av absolute valuefive highest values15 Shadow 0.1589 Morey 0.29312 Shelburne 0.9222 Carmi 1.02313 Star 1.40021.39954.259324.0318102-.193642-.550281.452966 ..2885171.78084.2622894.28302.364798frequency percentageint cum int cum0 16 0.0 100.01 16 6.3 100.01 15 6.3 93.81 14 6.3 87.513 0.0 81.30 13 0.0 81.31 13 6.3 81.36 12 37.5 75.06 12.54 12.5 37.525.02 2 12.5 12.5

-chlxminimum of interval1.4001.200 Ca St1.0000.8000.6000.400 Sh Bo0.200 Mo Ir-0.000 Sh Hv Fa-0.200 Ha Pa Cu-0.400 El Ho-0.600-0.800 Su SCunivariate statistics for: r-chlxnumber of cases = 16mean - .151378variance = .350952std deviation = .592412mean square •= .351933coef of variation = 3.91346std error of mean = .148103t-test for m=0 = 1.02211skewness coef. = .669165kurtosis coef. = .136169sum of squares = 5.63092av absolute dev = .431285five lowest values11 SC -0.781r-secchi 16 Sunset -0.659minimum 7 Hortonia of interval -0.3264 Elmore 0.400-0.2073 Curtis 0.200 SC Su -0.1170.000 Mo Sh Hv Pa El Ir-0.200 Ho Bo Fa Cu Ha-0.400 St-0.600-0.800 Ca-1.000 Shunivariate statistics for: r-secchinumber of cases =mean =variance =std deviation =mean square _ =coef of variationstd error of meant-test for m-0 =skewness coef. =kurtosis coef. =sum of squares =av absolute dev =16-.0842503.101904.319225.102634=-3.78901•- .0798062-1.05569-1.391451.157261.64214.219372five lowest values12 Shelburne -0.9002 Carmi -0.67913 Star -0.36014 Halls -0.1103 Curtis -0.108missing values = 2maximum = 1.35772upper hinge = .427412median = .0531025lower hinge = -.184901minimum = -.781423h-spread = .612313proM>t) = .324216large-sample z = 1.09274large-sample z = .111182cor. sum of sqs = 5.26428av absolute value = .425639five highest values9 Morey 0.3141 Bomoseen 0.46512 Shelburne 0.59413 Star 1.3402 Carmi 1.358missing values =maximum =upper hinge =median =lower hinge =minimum -h-spread =prob(>t) =large-sample z =large-sample z =cor. sum of sqs =av absolute valuefive highest values6 Hveys 0.10715 Shadow 0.1269 Morey 0.19516 Sunset 0.23911 SC 0.247frequencyint cum 0 16 2 160 140 140 14 n 142 123 103 72 40 22 2.246913.121528-.024271S-.109284-.90007.230812.308559-2.27223.9448991.52857= .219372frequencyint02651011percentageint cum0.0 100.012.5 100.00.0 87.50.0 87.50.0 87.512.5 87.512.5 75.018.8 62.518.8 43.812.5 25.00.0 12.512.5 12.5cum16161483221percentageint cum0.0 100.012.5 100.037.5 87.531.3 50.06.3 18.80.0 12.56.3 12.56.3 6.3

yhodminimum of interval0.8000.600 Sh0.400 Hv0.200 Ho0.000 SC Bo Mo Ir-0.200 Ha Fa Pa-0.400 Su Cafrequencymt cum 0 121 121 111 104 93 52 2percentagemt cum0.0 100.08.3 100.08.3 91.78.3 83.333.3 75.025.0 41.716.7 16.7univariate statistics for: r-hodnumber of cases = 12mean = .120334variance = .0917856std deviation = .302961mean square = .098617coef or variation = 2.51767std error of mean = .0874574t-test for m=0 = 1.37591skewness coef. = .569876kurtosis coef. = -.897996sum of squares = 1.1834av absolute dev = .236635five lowest values2 Carmi12 Sunset-0.209-0.2078 Parker -0.1403 Fairfield -0.13910 Halls -0.123missing values = 6maximum = .66229 5upper hinge = .338888median = .142411lower hinge = -.139607minimum = -.209161h-spread = .478495probOt) = .194138large-sample z = .805927large-sample z = -.634979cor. sum of sqs = 1.00964av absolute value = .256691five highest values1 Bomoseen 0.1609 SC 0.1635 Hortonia 0.3984 Hveys 0.59411 Shadow 0.662

Table 2Model Error ComponentsModel Error Term V.(R) V(0) V(M) V(M) Components39/40 Spring Phosphorus .069 .016 .053 .026 (Watershed, 39).027 (Retention, 40)241 Phosphorus/Chl-a .176 .019 .158 .94 x .053 (Spring P).111 (P/Chl-a, 41)242 Mean Chl-a/Max Chl-a .252 .021 .231 1.16 x .158 (Mean Chl-a).018 (Mean/Max Chl-a, 42)243 Mean Chl-a/Secchi .100 .009 .091 .65 x .158 (Mean Chl-a).024 (Mean Chl-a/Secchi, 43)244 Phosphorus/HOD .097 .010 .087 .91 x .053 (Spring P).043 (P/HOD, 44)V(R) = total variance of residuals, excluding Star Lake (outlier) formean and maximum chlorophyll-aV(0) = average observed data error component (Table 1)V(M) = estimated total model error = V(R) - V(0)V(M) Components = breakdown of model error, sum = V(M)All variances on natural log scale

Table 3Summary of Calibrated Error Terms for LEAP InputVariableStandardMean Deviation39 Watershed Model ErrorL .1640 Retention Model ErrorL .1641 P/Chl-a Model Error ] L .3342 Mean Chl-a/Max Chl-a Model Error J L .1343 Mean Chl-a/Secchi Model Error ] L .1544 Phosphorus/HOD Model ErrorL .21Notes :Standard Deviations estimated from square roots ofvariance terms estimated in Table 2;Standard Deviations of input variables 1-38should be set to 0.0.

magnitudes. Corresponding input values for the model error terms arespecified in Table 3.With a slight modification of the LEAP model subroutine, it can bereadily employed to test whether a given lake conforms to the modelstructure and calibration. Output variables 26-30 were originally usedto compare observed and predicted lake conditions on log scales.Because of the structure of the LEAP error analysis framework, it isnecessary to express these comparisons as ratios, rather than logratios; i.e., the "LOG" transformation in line 4052 of the subroutineshould be removed in order to provide a means for hypothesis testing.The effects of variance in the observed conditions (V(0)) can beincorporated by specifying non-zero standard deviations for inputvariables 20-24. These standard deviations can be estimated from thenumber of sampling years and the pooled estimates of year-to-yearvariance given in Table 1. For example, for a lake with a meanchlorophyll-a of 4.4 mg/m3 based upon 3 years of sampling, the inputstandard deviation for variable 21 would be:.5S(21) = 4.4 (.074 / 3 ) = .69With these modifications, output from the LEAP error analysis procedurewill include estimates of mean, standard deviation, minimum(2.5%) andmaximum(97.5%) for each observed/predicted comparison (output variables26-30). The mean equals the ratio of the observed to the predictedresponse and would therefore equal 1.0 in the case of perfect fit. Ifthe estimated range (min to max) does not include 1.0, then theprobability that the lake conforms to LEAP (or vice-versa) is less than5%. This range reflects the combined effects of model error (V(M)),input data errors (V(l), if any are specified), and observed response(V(0)) error. A lake which does not "conform" would be consideredatypical of the lakes used in LEAP calibration. It would also bepossible to modify the model code to permit direct probabilitycalculations.4

REFERENCESSnedecor, G.W. and W.G. Cochran, Statistical Methods , Iowa StateUniversity Press, Ames, Iowa, 1972.Walker, W.W., "Some Analytical Methods Applied to Lake Water QualityProblems", Doctoral Thesis, Division of Engineering and Applied Physics,Harvard University, 1977.Walker, W.W., "Calibration of LEAP to Vermont Lakes: Data Base Summary",prepared for Vermont Agency of Environmental Conservation, Water QualityDivision, December 1981.Walker, W.W., "Calibration and Testing of a Eutrophication AnalysisProcedure for Vermont Lakes", prepared for Vermont Agency ofEnvironmental Conservation, Water Quality Division, January 1982a.Walker, W.W., _"A Sensitivity and Error Analysis Framework for LakeEutrophication Modelling", Water Resources Bulletin , Vol. 18, No.l,February 1982b.Walker, W.W., "Empirical Methods for Predicting Eutrophication inImpoundments: Phase II Model Testing", prepared for the Office of theChief, Corps of Engineers, U.S. Army, Waterways Experiment Station,Vicksburg, Mississippi, draft, March 1982c.

Calibration of LEAP to Vermont LakesInterim Report:Data Base SummaryPreliminary Model Testingprepared forVermont Agency of Environmental ConservationWater Quality DivisionMontpelier, VermontbyWilliam W. Walker, Jr., Ph.D.Environmental Engineer1127 Lowell RoadConcord, Massachusetts 01742December 1981

INTRODUCTIONThe Lake Eutrophication Analysis Procedure (LEAP) is a system whichcan be used for predicting lake trophic status and related water qualityconditions, given certain land use, morphometric, and hydrologicinformation. (Walker, 1979,81). Regional calibration and testing of theframework are required prior to application in a planning context. Thisreport summarizes the data base which has been established for 15Vermont lakes using information provided by the Agency of EnvironmentalConservation (AEC). With slight modifications, the original version ofLEAP has been applied to each of the 15 lakes and the results comparedwith observed water quality data. The purposes of this report are toprovide:(1) a concise data summary of the model input variablesand observed water quality conditions to be usedin model calibration and testing(2) opportunity for the AEC to review the data baseand suggest any corrections or refinements(3) indications of model sensitivity to key variableswhich may need better definition(4) indications of the types of model changes whichmay be required in order to improve the frameworkFollowing review of this report and refinements in the data base,modifications in the model structure will be made to account forimportant factors which are not currently considered in LEAP. Resultsindicate two such factors: (1) differences in phosphorus export betweenwatersheds with sedimentary and glacial till soils; and (2) effects ofinternal phosphorus loading (recycling) on the lake phosphorus balanceand productivity. A second report will describe the final data base and1

the structure, calibration, and testing of the refined model framework.Appendices to this report contain the following:ABCDEdata base tablesLEAP subroutine and variable definitionsmodel predictions and sensitivity analysescharts of observed and predicted lake conditionsanalyses of variance in observed lake water qualityThe following sections describe the modified model and discuss keyresults.LEAP MODIFICATIONSBefore preliminary testing, the original LEAP framework has beenmodified to include the following:(1) additional land use categories, to conform with the landuse data base;(2) a means of accounting for phosphorus retention by upstreamlakes or reservoirs;(3) a phosphorus input term for shoreline septic systems;(4) modifications of the oxygen depletion rate calculationto account for the effects of multi-basin lakes;(5) modification of the function used to calculate summerchlorophyll based upon spring phosphorus concentration;(6) modification of the function used to calculate transparencyto account for variations in non-algal turbidity or color;(7) expansion of the list of output variables to include loadingcomponents;(8) provision of alternative means for estimating mean hypolimneticdepths required for oxygen depletion calculations.Each of these changes is discussed below.Sensitivity analyses indicate2

that none of the changes are very substantial in terms of impact onpredicted productivities of the study lakes. The original phosphorusexport and retention coefficients, which are derived from much largerwatershed and lake data bases, have been retained. Thus, we areessentially testing the original model framework.The watershed of each lake has been classified according to nineland use categories, in accordance with the types of data available fromplanning maps and landsat:(1) lake surface area(2) upstream lake surface area(3) wetlands (exclusive of upstream lakes)(4) conifer forest(5) hardwood forest(6) mixed forest(7) untilled agriculture(8) tilled agriculture(9) urbanThe model framework has been modified to permit specification of aseparate export concentration for each land use category. Inpreliminary testing, categories (2) through (7) have been treated asnon-contributing areas (the "forested" category in the original LEAPframework, with an export concentration of 15 mg/m3). Category (8)corresponds to the old "agricultural" classification, with an exportconcentration of 57 mg/m3, since the data base used to develop theoriginal export function treated "agricultural" and"cleared-unproductive" separately (Omernik, 1977). Category (9) isunchanged, with an export concentration of 139 mg/m3.The land use data file (Appendix A) contains three sets of land usedistributions for each lake, keyed by the following symbols:(1) LS = based upon Landsat(2) PM = based upon planning map3

(3) XX - "best"The last category is based upon the first two and subjective assessmentsprovided by the AEC for each lake. Land use distributions have beenadjusted so that the total watershed area equals the value specified inthe classification survey report (Vermont AEC, 1980). The adjustmentprocedure has used direct estimates for landuse categories (1), (2),(3), (7), (8), and (9). Total forested area ((4)+(5)+(6)) has beencalculated by difference from the total watershed area and the abovecategories. Finally, forested areas have been partitioned among theconifer, hardwood, and forested categories using percentages calculatedfrom landsat data, since forest types are generally not available fromplanning maps. Appendix A lists the estimated land use breakdowns bylake and source, with areas expressed in acres and as fractions of therespective total watershed areas.The framework has also been modified to account for phosphorusretention by upstream lakes or reservoirs. Surface and drainage areasof upstream lakes and reservoirs were provided by the AEC for each studylake. Using the settling velocity model (Chapra, 1975), the amount ofphosphorus trapped by each upstream lake can be calculated from thefollowing system of equations:WT - QR CI AW U / (U + QS)QS - QR AW / ALor,WT = QR CI FUFU•- AW AI U / (AL U + QR AW)where,WT » P load trapped by upstream lake (kg/yr)QR " average runoff rate = .6 m/yrCI - average inflow concentration of upstream lake (mg/m3)4

AW = watershed area of upstream lake (km2)U — phosphorus settling velocity = 12 m/yrAL = upstream lake surface area (km2)QS = surface overflow rate of upstream lake (m/yr)FU = adjustment factor for P retention in upstreamlake (km2)The parameters CI and FU are input for each lake. Since upstream lakesare generally remote from intensive uses, inflow concentrations areinitially assumed to equal the forested value (CI » 15 mg/m3), althoughthis parameter could be adjusted if there are significant agriculturalor urban uses above a given upstream lake. FU values have beencalculated from the surface and drainage areas of each upstream lake andsummed to provide a model input value for each study lake. In the modelsubroutine, the estimated WT value is subtracted from the totalundeveloped loading in constructing the lake phosphorus budget.Calculations are generally insensitive to the upstream lake factors.Bomoseen had the highest value (15.95 km2 vs. a total watershed area of95.7 km2)The phosphorus input from shoreline septic systems is estimatedfrom the following equation:WS • US FSwhere,WS = septic input (kg/yr)US•• septic system use factor (annual population equivalents)FS • septic phosphorus export factor (kg/capita-yr)US and FS are model input variables for each lake. A nominal value of.05 kg/capita-yr has been used in preliminary calculations. This isbased upon .5 kg/cap-yr input to systems (under phosphorus detergentban) and an assumed 90% treatment efficiency. On the average, theassumed FS value may be a bit low. FS values can be adjusted from lake5

to lake, based upon soil suitability, setbacks, age, etc.. Future workwill consider alternative values.Septic system use, US, is estimated from the following:WS - 3 (.5 R-SEAS + R-PERM) +(.5 USE-DAY + USE-NT)/365where,R-SEAS - number of seasonal shoreline residencesR-PERM = number of year-round shoreline residencesUSE-DAY = day use of resorts/parks (capita-days)USE-NT » night use of resorts/parks (capita-days)The above scheme assumes an average of 3 people per dwelling. AppendixA lists the above residence and use factors for each lake, along withthe calculated annual population equivalent, residence counts were notavailable for a few lakes. Approximate estimates have been derived bycounting structures on USGS or lake contour maps, increasing the totalby 20% (based upon average relationship between map count and tabulatedvalues for other lakes), and assuming that 75% of the residences areseasonal. Septic contributions calculated in the above way account fora maximum of 9.4% of the total phosphorus loadings for the study lakes(St. Catherine) and results are generally insensitive to this loadingcomponent. The assumed FS value may be biased on the low side, however,and needs further investigation.The scheme for calculating oxygen depletion rate is based upon itsrelationship with mean depth and phosphorus concentration (Walker,1979).The mean depth used in this calculation should reflect the morphometryof the stratified portion of the lake. Use of the nominal mean depthwill give biased predictions in lakes which have more than onehypolimnetic basin or relatively large, shallow, isolated embayments.For example, Lake Quinsigamond, Massachusetts, has three separatehypolimnetic basins. Depletion rates vary by a factor of two among thebasins but there is little variation in surface water quality. Themodel has been shown to apply using mean depths calculated separately6

for each basin (Walker, 1981).Accordingly, an additional input term ("basin mean depth") has beenincluded in the LEAF framework for use in oxygen depletion ratecalculations. Based upon a review of contour maps of the study lakes,only Eortonia is considered to have a basin mean depth which issubstantially different from the nominal mean depth. Roughly half ofthe surface area is in the southwest portion of the lake (maximum depth15 ft) which is remote from the stratified northeastern portion (maximumdepth 60 ft). In this case, the basin mean depth is estimated at 8.9 m,as compared with the nominal mean depth of 5.6 m. In all other studylakes, the basin mean depths equal the nominal mean depths.The function relating spring phosphorus to summer chlorophyll hasbeen replaced by the regression equation published in the classificationsurvey report (Vermont AEG, 1980). Figure 1 plots observed chlorophyllvs. observed spring phosphorus values for the study lakes, in relationto the regression equation. The chlorophyll productivities of Star andCarmi are considerably greater than the predicted values. In thesecases, the observed summer chlorophyll/spring phosphorus ratio exceeds1, which is roughly the algal physiologic ratio. This indicates thatseasonal or internal loadings are significant in these lakes and thatsummer phosphorus values would be more representative of trophic statusthan spring overturn values. The model tends to over-predict thechlorophyll values in lakes with meta- or hypolimnetic algal populations(Harveys, St Catherine, Hortonia, Morey). This may be related to thesampling difficulties which are present when most of the chlorophyll islocated within a relatively narrow depth range. While sampling andyear-to-year variablities influence the scatter in Figure 1,variabilities in the observed chlorophyll/phosphorus relationshipsimpose limitations on the accuracy of chlorophyll predictions derivedfrom the LEAP framework.The transparency function has been modified to provide a means ofincorporating the effects of non-algal turbidity and color:1/S - A0 + Al B7

Figure 1Observed Summer Chlorophyll-a vs. Observed Spring Phosphorus0

where,S = mean summer secchi depth (m)AO • non-algal turbidity/color term (1/m)Al = chlorophyll sensitivity = .025 m2/mgB = mean summer chlorophyll-a (mg/m3)The function is similar to those developed for Harveys Lake (VermontAEC, 1981) and for Corps of Engineer reservoirs (Walker, 1981). Theintercept term, AO, can be increased in lakes with high non-algalturbidity or color. One Platinum-Cobalt Unit of true color is theequivalent of about .003 1/m in AO.Observed chlorophyll and transparency data (Figure 2) indicate thatStar Lake (A0=.68) and Elmore (A0=.2) have somewhat higher non-algallight extinction coefficients then the others. A nominal value of .081/m has been used for all lakes in preliminary testing. Observed Secchiand chlorophyll data can be used to adjust AO in lake applications.Predicted transparencies are sensitive to AO only relativelyunproductive lakes, in which transparency is relatively insensitive tochlorophyll. The model will give biased transparency predictions inlakes with substantial metalimnetic or hypolimnetic algal populations,depending upon how chlorophyll samples are taken. A positive bias isapparent for Morey, in which case most of the chlorophyll is locatedbelow the Secchi depth.In addition to the above modifications, the output variable listhas been expanded to include an itemized accounting of the loadingcomponents (undeveloped, agricultural, urban, septic, and atmospheric).This provides insight into factors controlling the nutrient budget ofeach lake. Mean hypolimnetic depths required for oxygen depletioncalculation can be input directly (as derived from hypsiographs).Anticipating that detailed hypsiographs would not be available for manyfuture applications, an option has been included to permit estimation ofmean hypolimnetic depth based upon input values for mean depth, maximum8

*Figure 2Observed Summer Secchi Depth vs. Observed Summer Chlorophyll-a^m^¥^^^^^iml^^lm^l

depth, and thermoclme depth. The calculation procedure assumes thatlake surface area and volume vary as single term power functions intotal depth. If the input value for mean hypolimnetic depth is zero,the estimation procedure is used. If both the mean hypolimnetic depthand the thermocline depth are zero, the lake is assumed to beunstratified and oxygen depletion calculations are by-passed.DATA BASEThe data base is assembled in a series of tables in Appendix A. Thefollowing should be considered in reviewing the data:(1) Observed water quality data have been tabulated by lake and year,based upon information in the classification survey report, STORETprintout, lake reports, and the tabulated 1981 data. Where possible,missing chlorophyll and transparency values have been filled with laymonitoring data. Hypolimnetic oxygen depletion rates have beenre-calculated for all lakes, based upon profiles derived from STORET andthe hypsiographs, and using a consistent averaging procedure.Thermocline depths are based upon the Cornett/Rigler definition. Thedepletion rates tabulated in the Harveys Lake report (1981) are probablybiased on the low side, because profiles measured after the bottom ofthe hypolimnion reached anoxic conditions were apparently included inthe slope estimation procedure.(2) The morphometric file includes two estimates for mean depth: onederived from the classification survey report, and the other, from thehypsiographs. In most cases, aggreement is reasonable, with theexception of Harveys, and, perhaps, Winona and Carmi. These values needfurther checking. Aside from Harveys and Bomoseen, hypsiographs valueshave been used in the model testing described below.(3) Total drainage areas have been taken directly from theclassification survey report. Is this the most reliable source?9

(4) Runoff rates (back-calculated from residence times, mean depths,drainage areas, and surface areas in the classification survey report)seem highly variable, with range from .31 to .89 m/yr. Previousexperience indicates that long-term-average runoff rates in New Englandare rarely outside of the .5 to .75 m/yr range. These values needchecking.(5) Any additional insights that could be provided with regard to the"best" land use estimates would be helpful, particularly for lakes withlarge variances between landsat and the planning maps. Based upon thecharts in Appendix C, problem lakes in this regard include Cedar,Curtis, Iroquois, Park, and Shelburne. Tilled and urban land uses arethe most important for export calculations.(6) The ANOVA's in Appendix E should be reviewed. Relatively highwithin-lake (year-to-year) variability may indicate problems withmeasurements in certain lakes and years. For example, the year-to-yearvariance in spring phosphorus is .35 for Star Lake, in relation to anaverage year-to-year variance of .08 for all lakes.PRELIMINARY TESTINGTo provide preliminary indications of error and sensitivity, therevised model has been applied to each lake. The distributions ofpredicted variables and sensitivity analyses for spring phosphorusconcentrations are tabulated in Appendix C. These calculations arebased upon the "best" set of land use estimates for each lake. AppendixD contains a series of charts comparing observed values of springphosphorus, mean chlorophyll, maximum chlorophyll, transparency, andoxygen depletion rate with predicted values for each set of land useestimates. Symbols in the charts are used to identify differentsampling years and different land use estimates. Review of the chartsprovides insights into year-to-year variability in observed conditions,10

sensitivity to assumed land use distributions, and differences betweenobserved and predicted distributions. Comments on the results and/ordata base are indicated below the chart for each lake.Figure 3 contains stem-and-leaf diagrams of residuals for each ofthe six response variables. Uniform scales have been used to permitdirect comparisons of the error distributions on a natural logarithmicscale. While the LEAP framework provides direct error estimates, factorof two accuracy is generally a "rule-of-thumb" for empirical lake models(Vollenweider and Kerekes, 1981). These are indicated for each variablein Figure 3 (+/- .7 log units).Table 1 summarizes error statistics calculated from the observedand predicted lake characteristics. For reasons discussed below,Shelburne Pond is not typical of the other Vermont lakes or of the lakesused in development of LEAP and has been excluded from the errorstatistic calculations. Analyses of variance have been conducted on theobserved lake data to provide indications of year-to-year variability inrelation to lake-to-lake variability. Details on the ANOVA calculationsare given in Appendix E. ANOVA's indicate that differences among lakesare strongest in the case of TDO (F - 57.8) and weakest in the case ofHOD (F =5.0). The strength of the TDO variations is due primarily todifferences in mean hypolimnetic depths, rather than productivity.A.02 far 40 6 To .OSS torWithin-lake (i.e. year-to-year) variance ranges from.of variability "Tl»imposes limitations on the accuracies of the observed mean values usedin model testing. The pooled error variance of the observed means isappreciable in relation to the total error variance only for the oxygenstatistics, which have much lower total error variance than the othervariables. Thus, the observed data base appears to be adequate formodel testing - we cannot attribute lack of fit exclusively tomeasurement errors or variability in observed lake conditions, althoughthey may be important for certain lakes and variables.For spring phosphorus, the key variable in the framework, the errorvariance estimated by LEAP (.12) is comparable to the observed errorvariance (.11). Thus, the residual spread in Figure 3 (excludingShelburne) is consistent with data from other lakes and watersheds used11

Jkstem-and-leaf diagram for r-pspr>1.60 She1.401.201.000.80_ 3 * Accuv^'70.60 Mor0.40 Iro0.20 Fai Har Win-0.00 Bom Car-0.20 Hor Par St Sta-0.40 Ced Elm-0.60 Cur Figure 3-0.80-1.00LEAP Err °r Distributions-1.20 _ ,-1.40-1.60stem-and-leaf diagram for r-chla>1.60 She1.401.201.00 Car Sta0.80 n0.6~0~ ~ ' d0.40 Fai Mor0.20 Iro-0.00-0.20 Bom Har Par-0.40 Cur-JL.6 0 Elm Hor St-0.80 **" ~~-1.00-1.20-1.40-1.60tem-and-leafdiagramXf> 1.60 She1.40 Car1.201.000.80 Sta nO.FOIfoflor *•0.40 Fai0.20 Bom-0.00 Par-0.20 Har-0.40-0.60 Cur Elm-0.80 ffor-1.00 St-1.20-1.40-1.60X^•Ccu^^•4 >£h (of*se'*—t/fr*c/&JJ ci.

stem-and-leaf diagram for r-secchi1.601.401.201.00jp_.80_ _2X ftca+ruty0.60 "" '0.400.20 Har St-0.00 Bom Cur Hor Mor-0.20 Elm Par Figure 3-0.40 Fai Iro^.•^ - — • LEAP Error Distributions-0.80 Car (continued)-1.00-1.20-1.40/ -1.60 She Stastem-and-leaf diagram for r-hod1.601.401.201.00 A0.400.20 Mor-0.00 Fai Har Iro-0.20 Bom Car Hor Par St-0.40-0..6J).-0.80-1.00-1.20-1.40-1.60stem-and-leaf diagram for r-tdo1.601.401.201.00oUo0.20-0.00 Bom Car Hor Par St-0.20 Fai Har Iro-0.40 Mor-0.60 _ ,•^.tflT """"-1.00-1.20-1.40-1.60

Table 1Summary of Error StatisticsVariableStatistic Spring P- Chl-a Chl-max Secchi HOD TDOMODEL ERROR STATISTICS - excludingShelburneNumber of Lakes 14 13 11 13Estimated MSE (a)Observed MSE (b)Var (Obs. Means) (c).12.11.016.25.32.020.26.51.036.26.28.008.18.02.009.18.02.009R-Squared -.16 .16 .16 .06 .24 .96.... ANOVA STATISTICS - excluding Shelburne —uamg o aeiourneAmong-Lake Mean Sq. .534 1.523Within-Lake Mean Sq.(d) .080 .072F 6.7 21.01.352.08815.3.986.02835.1.100.0205.01.160.02057.8ANOVA STATISTICS - all lakes (e)-Among-Lake Mean Sq. 1.710 2.321Within-Lake Mean Sq.(d) .080 .074F 21.2 31.42.025.08324.31.547.03544.6.100.0205.01.160.02057.8Note: all statistics on natural log scalesabcdeerror variance of estimated values calculated by LEAPobserved error variancepooled estimate of error variance of observed mean values, basedamong-year variance and average number of sample yearsrepresents year-to-year variance within lakesdetails given in Appendix E

in development of LEAP. Observed variance is greater than estimatedvariance in the case of chlorophyll (.32 vs .25) and maximum chlorophyll(.51 vs .26).R-squared values are apxeciable only in the case of TDO (.96 vs-.16 to .24 for the other variables). The strength of the TDO predictionis attributed to the hypolimnetic depth effect discussed above. Thus,while it does explain variance in hypolimnetic oxygen status, theexisting framework does not explain much variance in the observedproductivities of this collection of lakes. This is attributed to therelatively low variance in lake productivities (most are mesotrophic),inherent "factor-of-two" accuracy limits of the empirical modellingapproach, and to the effects of factors which are not considered in themodels. The error analyses for predicted spring phosphorus in AppendixC indicate that the predominant sources of prediction error are in thephosphorus retention and export models. Further analyses discussedbelow indicate that there are good possibilities for improving thesemodels and reducing prediction error.A systematic residuals analysis has been undertaken to determinewhether errors in the predictions are associated with any of the lake orwatershed characteristics included in the data base. Any suchassociations may indicate need for re-calibration or re-structuring ofthe model framework. Shelburne is clearly an "outlier" in relation tothe other lakes. It is the only lake outside of the factor-of-two errorlimits for spring phosphorus. Based upon review of the errorhistograms, bivariate residuals plots, review of soils data for eachlake, and independent evidence from the literature, much of the errorvariance for Shelburne and other lakes can be attributed to the effectsof two factors which are not currently considered in the LEAP framework:(1) the differences in phosphorus export between watersheds withsedimentary soils and those with non-sedimentary soils; and (2) theeffects of internal loading (recycling) on the phosphorus balance andproductivity. These effects are discussed in detail below.The export functions used in the LEAP framework have beencalibrated to data from over 100 watersheds in the Northeast, most of12

which have soils of glacial origin (Meta Systems, 1978). The analysisindicated that export coefficients for forested and agricultural landsuses are two to five times greater in the Midwest (western Ohio,Indiana, Illinois), which have soils of sedimentary origin. Greaterphosphorus exports from sedimentary watersheds have been identifiedelsewhere. For example, Dillon and Kirchner (1975) found that meanexport values from sedimentary watersheds were about 2.5 times greaterthan those from igneous watersheds of plutonic origin. This isprimarily attributed to high phosphorus concentrations and often poordrainage characteristics of lake-plain soils, as compared with glacialtill.Of the 15 study lakes, three have watersheds with sedimentarysoils: Shelburne (approx. 75%), Winona (approx. 40%), and Fairview(approx. 15%). The existing LEAP framework under-predicts springphosphorus concentrations in each of these lakes, by factors of 6, 1.36,and 1.36, respectively. The bias for Shelburne is reduced to a factorof 4 if land use estimates from the planning map are used. Thus, itseems reasonable that these biases could be at least partially offset byusing higher phosphorus export concentrations to represent thesedimentary portions of the watersheds. This will be investigated infuture model refinements.Another factor contributing to bias in the case of Shelburne, isthat the watershed contains a high percentage of soils (approx. 75%) inHydrologic Soil Group D. These high-clay, poorly-drained soils arehighly subject to surface runoff, the predominant transport mechanismfor total phosphorus. The watershed also contains a high percentage oftilled agricultural land use which, combined with the poorly-drained,lake-plain soil types, would tend to result in higher exportcoefficients than predicted by the LEAP framework. The location,watershed characteristics and water quality or Lake Shelburne suggestnaturally eutrophic conditions which are not typical of other Vermontlakes.Another curious aspect of Lake Shelburne is the apparent decreasingtrend in spring total phosphorus, ranging from 147 mg/m3 in 1977 to 7213

mg/m3 in 1981. Model bias reduces from a factor of 6 to a factor of 4if 1981 measurements are used. This trend contrasts with many of theother lakes, which showed increasing trends over the same period. Forexample, Harveys increased from 10 to 22 mg/m3, Morey, from 17 to 48mg/m3, Elmore, from 10 to 15 mg/m3, Parker from 14 to 21 mg/m3, StCatherines, from 10 to 17 mg/m3, and Star, from 10 to 23 mg/m3. These"trends" do not necessarily reflect significant variations in culturalimpacts and are probably related to climatologic variations or to thetiming of spring phosphorus sampling in relation to peak runoff andturnover.Year-to-year variations in observed spring phosphorus levels areconsiderable and it would be useful to develop a better basis forinterpreting them. One approach would be use watershed model relatingdirect runoff to precipitation sequences, similar to that developed inrecent work for the New Haven Water Company (Walker, 1981), based uponthe SCS Soil Cover Complex Method. The model simulates annual sequencesof storm events and calculates direct runoff by event, as well as annualtotals. By simulating multiple years, a feeling for year-to-yearvariations in loadings can be obtained. This provides some help ininterpreting lake water quality measurements made during a given year inrelation to the long—term-average and, thus, distinguishingnaturally-induced from culturally-induced variations. Input data forthis model are generally available from soils maps, land use maps, andweather stations which record daily total precipitation. Essentially,this framework is a second-order LEAP which attempts to simulateyear-to-year variations, as opposed to the current version, which isfocused on the long-term average.Internal phosphorus recycling is another factor possiblycontributing to errors in the spring phosphorus predictions from theexisting framework. This effect is demonstrated most effectively usinga modelling exercise. The current framework explicitly considersexternal phosphorus loading only. The effects of internal loading areimplicitly included in the phosphorus retention function. If thenutrient balance equation is modified to account for internal loading14

directly, the equation for predicting lake phosphorus concentrationbecomes:P = (WX + WI) (1 - RP) / (QSAS)where,P = lake phosphorus concentration (mg/m3)WX - external phosphorus loading (kg/yr)WI = internal phosphorus loading (kg/yr)RP a phosphorus retention coefficientQS » surface overflow rate (m/yr)AS = lake surface area (km2)The potential for internal loading is directly related to the oxygenstatus of the hypolimnion, since phosphorus release rates from anoxicsediments are generally orders of magnitude greater than release ratesfrom oxic sediments. If we assume that the internal loading per unit ofhypolimnetic area is proportional to the length of the anoxic period,the following expression results:WI == FI TANOX AHTDO => 12 ZH / HOD» FI (200 - TDO) AHwhere,FI = effective P release rate from anoxic sediment (mg/m2-day)TANOX «• length of anoxic period (days/yr)TDO =» days of oxygen supply at spring turnoverAH = hypolimnetic surface area (km2)ZH •• mean hypolimnetic depth (m)HOD = areal oxygen depletion rate (g/m2-day)The above expression assumes an average stratified period of 200 daysand spring oxygen concentration of 12 g/m3. Combining the aboveequations, the effect of internal loading on the predicted phosphorus15

concentration is given by:PI - (1-RP) WX / (AS QS)P2 - (1-RP) [WX + FI (200 - TDO) AH] / (AS QS)P2/P1 - 1 - FI (200 - TDO) AH / WXwhere,Pi • lake P calc. from external loading (mg/m3)P2 = lake P calc. from external and internal loading (mg/m3)If the internal loading model is realistic, the expression on theright-hand side of the last equation reflects the potential significanceof internal loading in relation to the external loading and should berelated to the errors in the calculated phosphorus concentrations.Variables AH and WX are measured or calculated in the LEAP framework.TDO can be observed or calculated, based upon external phosphorusloadings, but is generally more sensitive to morphometry (meanhypoiimnetic depth). If the model is to be useful, the remainingparameter, FI, should be reasonably constant across lakes and can beestimated from the slope of the line in Figure 4, which plots the errorin the spring phosphorus prediction (expressed as P/PE — 1) as afunction of the internal loading parameter derived from the right-handside of the above equation. According to the model, internal loadingshould be of greater significance in lakes further to the right,including Morey, Iroquois, Fairview, and Carmi. Most of the lakes areclose the line with the following equation:P/PE -1 -. -.2 + 1.2(200 - TDO) AH / WXThe slope of the line indicates an average net release rate from of 1.2mg/m2-day from anoxic hypoiimnetic areas. The apparent negativeintercept (-.2) may reflect a 20% bias in the phosphorus retentionfunction for lakes which do not stratify or do not reach completelyanoxic conditions. Since the retention function was developed on a16

Figure 4Error in Spring Phosphorus Estimate vs. Internal Loading Term1.501 5U«11.39 j.291.181.081.9710.87!0.7610.66 10.5510.45 1I0.3'H HarWin0.24 I0.131•r 0.03 1*• -0.081-0.18lCedSta"-Par"Hor StCI-0.50+_0.289,ANOXIC+-0.42A H / -W,+-0.56+0.70

collection of oxic and anoxic lakes, it is possible that such a biascould exist.The significance of the negative intercept needs furtherinvestigation with the refined data base, since some of the negativeresiduals may be attributed to problems with the input or observedvalues. As expected, the three lakes with partially sedimentarywatersheds (Shelburne, Winona, and Fairview) still have positive biaseswhen the effects of internal loading are considered. Modifications ofthe above model may be required to account for vertical variations inthe volumetric oxygen depletion rate, as observed in Harvey's Lake(Vermont AEC,1981).The above model provides a direct means of calculating internalloading which could be incorporated into the lake nutrient balances.The only additional input requirement would be the hypolimnetic surfacearea, which could be estimated from contour maps or from the mean depth,maximum depth, and thermocline depth using a scheme which is similar tothat employed for estimating mean hypolimnetic depths. Modification ofthe existing LEAP to account for sedimentary soils and internal loadingshould reduce prediction error considerably and improve its usefulnessas a planning tool for Vermont lakes. This work will proceed once thedata base has been reviewed and, where possible, refined.17

REFERENCESChapra, S.C. "Comment on 'An Empirical Method of Estimating theRetention of Phosphorus in Lakes' by W.B. Kirchner and P.J. Dillon",Water Resources Research , Vol. 11, No. 6, pp. 1033-1034, 1975.Cornett, R.J. and F.H. Rigler, "Hypolimnetic Oxygen Deficits: TheirPrediction and Interpretation", Science , Vol. 205, pp. 580-581, 10August 1979.Dillon, P.J. and B. Kirchner, "The Effects of Geology and Land Use onthe Export of Phosphorus from Watersheds", Water Research , Vol. 9, pp.135-148, 1975.Meta Systems, Inc., "Land Use-Nutrient Export Relationships in theNortheast", in "A Preliminary Analysis of the Potential Impacts ofWatershed Development on the Eutrophication of Lake Chamberlain",prepared for New Haven Water Company, Connecticut, June 1978.Omernik, J.M., "Nonpoint Source - Stream Nutrient Level Relationships:A Nationwhide Study", EPA Ecological Research Series, CorvallisEnvironmental Research Laboratory, U.S. Environmental Protection Agency,Corvallis, Oregon, EPA-600/3-77-105, September 1977.Stauffer, R.E., "Sampling Strategies for Estimating the Magnitude andImportance of Internal Phosphorus Supplies in Lakes", Water ChemistryLaboratory, University of Wisconsin, draft report prepared for U.S.Environmental Protection Agency, Corvallis Research Laboratory, May1980.Vermont Agency of Environmental Conservation, "Vermont LakeClassification Survey", Department of Water Resources and EnvironmentalEngineering, Water Quality Division, Montpelier, December 1980.Vermont Agency of Environmental Conservation, "Harvey's Lake Diagnostic/Feasibility Study", Interim Report, Department of Water Resources andEnvironmental Engineering, Water Quality Division, Montpelier, January1981.Vollenweider, R.A. and J. Kerekes, "The Loading Concept as Basis forControlling Eutrophication - Philosophy and Preliminary Results of the0ECD Programme on Eutrophication", Prog. Wat. Tech. , Vol. 12, pp.5-38, 1980.Walker, W.W., "Use of Hypolimnetic Oxygen Depletion Rate as a TrophicState Index for Lakes", Water Resources Research , Vol. 15, No. 6, pp.1463-1470, December 1979.Walker, W.W., "A Lake Eutrophication Analysis Procedure Applied to LakeMorey", prepared for the Vermont Department of Water Resources, WaterQuality Division, Lakes Program, August 1980.

Walker, W.W., "A Sensitivity and Error Analysis Framework for LakeEutrophication Modelling", accepted for publication in Water ResourcesBulletin , 1981.Walker, W.W., "Empirical Methods for Predicting Eutrophication inImpoundments: Phase II Model Testing", prepared for the Office of theChief, Corps of Engineers, U.S. Army, Waterways Experiment Station,Vicksburg, Mississippi, draft, 1981.Walker, W.W., "Water Quality Data Analysis, Model Development, and ModelApplications for Lake Quinsigamond", prepared for Meta Systems Inc.,Environmental Design and Planning Inc., Massachusetts Division of WaterPollution Control, and U.S. Environmental Protection Agency, May 1981.Walker, W.W., "Reservoir Eutrophication as a Factor in ManagingWater-Supply Reservoirs", prepared for Meta Systems, Inc., and New HavenWater Company, Connecticut, draft, November 1981.

APPENDIX AData BaseObserved VJater Quality Data by Lake and YearObserved Water Quality Data - Geometric MeansLake Morphometry - ILake Morphometry - IILand Use FractionsLand Uses - AcresShoreline Septic Use DataOther Model Input VariablesSoils Data - by LakeSoils Data - by Name

tn tn cort rr rro o oto as to mrt rt rtET Cf tTCD (D CDH H MnatoM5CI•t)(0H*?*ftnm T) TI gca co n> ol-t l-t l-l H.JK 4 ?r ?v CO COPip iBoiCDoneHrtH'COne>irtHCO~J ,^-^C»C»^J~J--JCX3CXP^~J-^-^O0rX>^-J-JO0C»~-J^-^C»C»-vl--J~JHHHH tOt— P-l—'I— M M I-' p"I—•I—• I—• I—• I—-I— H H H M H I-'I— MMI-'HOOHC^«O^C»OM Ul v| C> ^ • M NUDOi •P-OOVDNJVOUlCONJ U l J l U N J W t O l O f - O i u l U l 4 > t O WO W I vo ys «-J I - J O O I O l — Io co o o en woocwI O H v D P ^ H C » s l I I OlL0^fjNJ^S3Ov0~-JJ>N3l-'v000 I0 0 0 0 > 0 0 0 0 O O O o o r - J O O O O O O . t > < o ^ JIII COol i l tp< rO00 CO. O . Oo oI I IN3O.00p 1 |o o oNi UlUl -Pii b b io ocr> ex>I b b io ovO 00 Ul --J* • . •1 O O NO O 1o o o oco •-•CO .£>I b Joo o. . .O O Cio o oIIUloI I I I1 1co Co .p- Co 4> Co ui Ui ( M O W M N N ^ U I Wvovo I COO0-P- I --I Co Co O i_nO O O O ^ J O O O O OO H » J C M j i C K O > O N 3o o o o o o o o oII0>-~JC7>CT>CT>tsJCON5^ljONiOOCOCOU i O I - ' O o i O W O N W u i i O t o U WO O O O O O O C O O O O O O OI I II 03oI I I I!-• t—'a .00 ~~lo oJ> INOOo o oI I I I Ul 4> Ul I I ICO CoI -t> I I IuiI I I I I I I I IoI I I N 5 I^J• W W

St CatherSt CatherShelburneShelburneShelburneShelburneShelburneStarStarStarStarWinonaWinonaWinonaWinonaWinona80.0081.0077.007 8.0079.0080.0081.0078.00[79.00^80.0081.0077.0078.0079.0080.0081.0012.0017.00147.00128.00135.0099.00 .72.0010.007.0022.0023.0040.0022.0020.0029.0023.00----------------3.004.90-59.3096.80---17.0023.00---------132.00150.00---33.0042.00------6.605.40-0.700.34---0.900.80------0.39Lake Water Quality Data - by Yearyr = year sampledpspr = spring phosphorus concentration (mg/m3)psum = sunaner epilimnetic phosphorus concentration (mg/m3)chla = summer chlorophyll-a (mg/m3)chlmax = maximum summer chlorophyll-s (mg/m3)secchi * secchi depth (m)hod = hypolimnetic oxygen depletion rate (g/m2-day)

•LakeBomoseenCarmiCedarCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonapspr14.8319.9617.1312.1512.4819.9713.8411.6729.7227.0715.3311.75112.6113.7225.94P sum12 .14---13 .04----11 .34---—-chla5.3721.79-6.304.2210.453.573.6710.519.526.213.1875.7619.77-chlnax14.8364.99-11.508.4521.657.156.0036.7420.4615.684.40140.7137.23-secchi4.641.84-3.803.232.856.6 54.932.654.723.796.380.490.85-hod0.380.27---0.45,W3cu^.0.410.450.510.400.40---tdo113.6841.78---75.70106.3160.8447.4791.50162.60---zhyp3.600.940.000.000.002.84U>-rkQ-3.632.302.003.055.420.000.000.00Lake Water Quality Data - Geometric Meanspspr = spring phosphorus concentration (mg/m3)psum = summer epilimnetic phosphorus concentration (mg/m3)chla = summer chlorophyll-a (mg/m3)chlmax •• maximum summer chlorophyll-a (mg/m3)secchi =• secchi depth (m)hod " hypolimnetic oxygen depletion rate (g/m2-day)tdo =» days of oxygen supply at spring turnover = 12 zhyp / hodzhyp - mean hypolimnetic depth (m)

lakeBomoseenCarmiCedarCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonadarea95.7030.873.333.7320.4815.6420.6817.868.9220.6821.4930.1619.272.8711.30sarea9.575.570.460.31 ,0.91/L48// zm zm-hg8.20 -6.20 5.442.00 1.9213.30 3.323.40 3.497.10 7.23Q]_JJ& 22.915.50 i I 5.595.90 5.787.60 8.307.70 7.6110.90 10.723.60 3.611.50 1.481.20 1.02zx19.8010.104.009.805.2012.8044.2018.3011.3013.1014.7019.507.902.402.70t1.901.800.890.460.171.003.301.301.601.800.522.500.780.180.32zthera10.008.000.000.000.008.0010.0011.007.509.008.0010.000.000.000.00zhyp3.600.940.000.000.002.84•4r5 J 1 ?©*3.632.302.003.055.420.000.000.00fe.«-/V.20.Lake Morphometry - Idarea» total drainage area (km2)sarea • lake surface area (km2)zm » mean depth (from classification report) (m)zm-hg • mean depth (calculated from hypsiographs) (m)zx » maximum depth (m)t = hydraulic residence time (yrs)ztherm = thermocline depth (m)zhyp = mean hypolimnion depth (m)

lakeBomoseenCanniCedarCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonaelev411.00435.00492.001219.001139.00550.00893.00484.00684.00416.001299.00484.00330.001851.00467 .00maxelev shore2034.00 34610.00861.00 12816.00900.00 2665.001707.00. 5122.002000.00 4218.001130.00 9048.002379.00 6338.001125.00 12651.001250.00 5201.001776.00 7662.002157.00 4830.002100.00 15820.00830.00 7423.002300.00 2150.001806.00 4152.00runoff0.430.620.310.600.890.850.430.430.340.450.670.500.440.670.32voldev1.241.841.501.011.961.661.200.901.57.1.741.571.681.371.881.33shordev3.151.531.112.591.251.861.392.641.611.461.382.401.551.261.20ulakfac15.950.610.000.000.770.000.005.010.000.000.261.280.000.000.00Lake Morphometry - IIelev * surface elevation (ft, msl)maxelev = maximum watershed elevation (ft, msl)shore * shoreline length (m)runoff • runoff rate from watershed (m/yr)voldev = volume development ratioshordev = shoreline development ratioulakfac = upstream lake P retention factor (km2)

lakeBomoseen LSBomoseen PMBomoseen XXCarmi LSCarmi PMCarmi XXCedar LSCedar PMCedar XXCurtis LSCurtis PMCurtis XXElmore LSElmore PMElmore XXFairfield LSFairfield PMFairfield XXHarveys LSHarveys PMHarveys XXHortonia LSHortonia PMHortonia XXIroquois LSIroquois PMIroquois XXMorey LSMorey PMMorey XXParker LSParker PMParker XXSt Cath LSSt Cath PMSt Cath XXShelburne LSShelburne PMShelburne XXStar LSStar PMStar XXWinona LSWinona PMWinona XXcon if0.0920.0810.0820.0180.0170.0160.1420.0940.1230.2480.1370.1620.2880.2220.2490.1020.0910. .0940, .4270.3160.3450.1040.0840.0900.0950.0850.0770.1880.1660.1660.4380.2010.2340.1370.1060.1120.0470.0350.0400.3180.2750.2770.1180.0910.099hdwd0.5140.4530.4620.2780.2530.2480.2860.1890.2480.4680.2590.3050.4930.3790.4240.6120.5480.5650.2910.2150.2350.5010.4040.4330.3640.3260.2940.6130.5430.5430.3310.1520.1770.5590.4290.4550.1250.0910.1040.3660.3160.3180.4280.3300.357Land Use Fractionsmxdf0.1330.1220.1240.1020.0930.0910.1420.0940.1240.0910.0500.0590.0650.0500.0560.0740.0670.0690.0730.0540.0590.1610.1300.1390.0690.0620.0560.0690.0610.0610.0970.0450.0520.1180.0900.0950.0590.0430.0490.1550.1340.1350.0820.0630.069untld0.0600.0690.0670.2220.1200.2070.1370.0300.1730.0900.0800.2290.0330.0900.0990.0490.0300.0470.0630.1100.1780.0400.1100.0760.2310.0000.2310.0100.1000.0850.0610.1300.3170.0300.0600.1200.3620.0300.3750.0200.1230.1180.1190.1870.160tild0.0100.0300.0250.1210.2000.1200.1190.3600.1300.0100.2200.0720.0220.1200.0330.0290.0700.0310.0110.1400.0290.0100.0300.0190.0880.2720.0880.0100.0100.0100.0100.3800.1360.0100.1290.041'0.1660.5400.1720.0100.0100.0100.0280.0920.080urban0.0100.0690.0630.0000.0600.0600.0000.0600.0300.0000.1610.0800.0000.0400.0400.0000.0600.0600.0000.0300.0190.0000.0600.0600.0000.1010.1010.0000.0100.0250.0000.0300.0230.0100.0500.0400.0100.0300.0300.0200.0310.0310.0090.0200.020lake0.1000.1000.1000.1800.1800.1300.1380.1380.1380.0830.0830.0830.0440.0440.0440.1200.1200.1200.0800.0800.0800.1020.1020.1020.0930.0930.0930.1050.1050.1050.0450.0450.0450.1140.1140.1140.0940.0940.0940.0800.0800.0800.0840.0840.084conif = conifer foresthdwd = hardwood forest.untld = untilled agriculturetild = tilled agricultureurban = urbanlake = lake surfae areaulake = upstream lakeswet In = wetlands (exclusive of upstream lakes)ulake0.0200.0200.0200.0020.0020.0020.0000.0000.0000.0000.0000.0000.0050.0050.0050.0000.0000.0000.0000.0000.0000.0330.0330.0330.0000.0000.0000.0000.0000.0000.0020.0020 .0020.0030.0030.0030.0000.0000.0000.0000.0000.0000.0000.0000.000wet In0.0570.0570.0570.0760.0760.0760.0340.0340.0340.0100.0100.0100.0490.0490.0490.0140.0140.0140.0550.0550.0550.0480.0480.0480.0610.0610.0610.0040.0040.0040.0150.0150.0150.0190.0190.0190.1370.1370.1370.0310.0310.0310.1310.1310.131

lakeBomoseen LSBomoseen PMBomoseen XXCarmi LSCarmi PMCarmi XXCedar LSCedar PMCedar XXCurtis LSCurtis PMCurtis XXElmore LSElmore PMElmore XXFairfield LSFairfield PMFairfield XXEarveys LSHarveys PMHarveys XXHortonia LSHortonia PMHortonia XXIroquois LSIroquois PMIroquois XXMorey LSMorey PMMorey XXParker LSParker PMParker XXSt Cath LSSt Cath PMSt Cath XXShelburne LSShelburne PMShelburne XXStar LSStar PMStar XXWinona LSWinona PMWinona XXtfor17 573154901578830362761270747031140774441148542823297368930442727281240422987326633812722291911631043940444539343935459921132458606046564933110180591759551451817 5313521464untld14181636158416969151578113251428374211168456502188116181321562"907178485335508050851511434325690168022444489417 2214317 84148783333523447tild23670159192515249159829610792036611160616811327012054715148451328519359919451505054201672074962305788256981677777258223urban2361636150004574570492501487402022020232232015396026 526 5022222205212801591207437030049143143142222265656lake236423642364137613761376114114114777777225225225464464464410410410450450450205205205538538538240240240852852852450450450575757235235235ulake468468468121212000000242424000000146146146000000121212212121000000000wet In134313431343580580580282828999249249249545454281281281212212212134134134222222787878144144144650650650222222367367367Land Uses - Acrestfor = total forestuntld = untilled agriculturetild = tilled agricultureurban = urbanlake = lake surface areaulake = upstream lakeswet In • wetlands (exclusive of upstream lakes)

lakeBomoseenCarmiCedarCurt isElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonaseas-res269.0225.031.030.0120.0100.085.050.030.075.096.0280.00.06.07.0perm-res90.075.05.05.025.00.017.017.010.025.04.090.00.01.00.0day-use6929.018709.00.00.021371.00.00.00.00.00.00.040212.00.00.00.0nt-use1276 5.033322.00.00.010730.00.00.00.00.068000.00.025304.00.00.00.0sload718.0679.461.560.0313.7150.017 8.5126.075.0373.8156.0814.40.012.010.5Shoreline Septic Use Dataseas-res = number of seasonal, shoreline residencesperm-res = number of permanent, shoreline residencesday-use = daytime use of state parks/resortsnight-use = overnight use of state parks/resortssload =» computed shoreline septic use (capita/yr)= 3 (perm-res + .5 seas-res) + (day-use/2 + night-use)/365•

lakeBomoseenCarmiCedarCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonaulakfac15.950.610.000.000.770.000.005.010.000.000.261.280.000.000.00sload717.97679.4361.5060.00313.68150.00178.50126.007 5.00373.82156.00814.420.0012.0010.50otherld0.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00zbasin8.205.441.923.323.497.2317.708.905.788.307.6110.723.611.481.02Other Model Input Variablesulakfac - upstream lake P retention factor (km2)sload = shoreline septic use (capita/yr)otherld - other direct P loading (kg/yr)zbasin = mean depth of stratified part of lake (m)

Soils Data - by LakeLAKE % SOIL ASSOCIATION ORIGIN HSGEomoseen 80 Nassau-Dutchess T C/D BBomoseen 10 Dutchess-Nassau T B C/DBomoseen 10 Bernardston-Pittston T CCarmi 45 Peru-Stowe T CCarmi 45 Cabot-Westbury T CCarmi 10 Carlisle-Terric T DB BCedar 70 Nellis-Amenia TB CCedar 30 Berkshire-Mario _ T C/DCurtis 90 Glover-Calais T C CCurtis 10 Calais-Buckland T C CElmore 50 Peru-Mario T C/D C CElmore 30 Lyman-Marlo-Peru T D CElmore 20 Cabot-Peru T c CFairfield 60 Woodstock-Tunbridge T c CD DFairfield 15 Peru-Stowe TA AFairfield 10 Carlise-Terric Med. TC C BFairfield 10 Windsor-Missisquoi S c CFairfield 5 Scantic-Raynham-Bing. S c BHarveys 70 Paxton-Woodbridge T DHarveys 15 Woodstock-Colrain T B C/DHarveys 15 Muck & Peat-Peacham T C CHortonia 70 Dutchess-Nassau T C DHortonia 30 Bernardston-Pittston T C/D CIroquois 33 Peru-Cabot T C CIroquois 33 Lyman-Marlou T C C BC C CIroquois 33 Peru-Marlou TD CMorey 80 Tunbridge-Woodstock-Colr TC/D CMorey 20 Tunbridge-Woodstock-Buckl T C CParker 50 Cabot-Buckland T DParker 40 Glover-Calais T C/D BParker 5 Buckland-Calais T C CParker 5 Muck & Peat Peacham T B BDSt Cather 70 Nassau-Dutchess T DDSt Cather 25 Bernardston-Pittstown TDB BSt Cather 5 Stockbridge-Amenia T C C C/DShelburne 50 Vergennes-Covington S C C C/DShelburne 25 Muck & Peat S C/D C CShelburne 25 Farmington-Nellis-Stockbr T C/D BStar 60 Peru-Marlo-Lyman T DStar 30 Marlo-Peru-Lyman T C BB CStar 10 Lyman-Marlo-Peru TABAWinona 40 Lyman-Berkshire-Marlow TWinona % = approx. 20 percent Muck & of Peat drainage Assoc basin in given S assoc.Winona ORIGIN = soil 20 Raynham-Amenia origin (S=sedimentary , T=Till) SWinona HSG = hydrologic 10 Berkshire-Marlow soil groupTWinona 10 Colton-Stetson-Adams S

Soils DataNAME HSG EROD ORIGINAdamsAmeniaBerkshireBernardstonBinghamvilleBucklandCabotCalaisCarlisleColrainColtonCovingtonDutchessFaraingtonGloverLymanMarioMissisquoiNassauNell isPaxtonPeruPittstownRaynhamScanticStetsonStockbridgeSt oweTerricTunbridgeVergennesWindsorWoodbridgeWoodstockBBBCBCDCA/DBADBC/DC/D .C/DCAC/DBCCCCC "'BBCBCDACc.49.28.20.32.20.17.24.24.20.17.20SSTTSTTTTTSSTTTTTSTTTTTSSSTTTTSSTTHSG = hydrologic soil groupEROD = erodibility factorORIGIN = soil origin(S=sedimentary,T=til1)

APPENDIX BLEAP Subroutine and VariablesRevised Lake Model SubroutineInput Variable List (Lake Morey)Output Variable List

3480 REM revised lake model subroutine3490 A1=X(1)+X(2)+X(3)+X(4)+X(5)+X(6)+X(8)+X(9 ):'watershed3500 Y(2)=X(1)*X(24)+X(2)*X(25)+X(3)*X(26)+X(33510 Y(2)=X(35)*X(15)*(Y(2)-X(20)*X(10))3520 Y(3)=X(35)*X(15)*(X(4)*X(27)+X(5)*X(28))3530 Y(4)=X(35)*X(15)*X(6)*X(29)3540 Y(5)=X(7)*X(32)3550 Y(6)=X(16)*X(21)+X(17)'septic'total3560 Y(1)»Y(2)+Y(3)+Y(4)+Y(5)+Y(6)3570 Y(7)=Y(1)/((A1+X(7))*X(15))3580 Y(8)=(A1+X(7))*X(15)/X(7)3590 Y(9)=X(11)/Y(8)3600 Y(10)=1!/(1!+.82*(Y(9)".45)*X(36))3610 Y(11)=Y(10)*Y(7)3620 Y(12)=X(37)*EXP(-.69+.94*LOG(Y(ll)))3630 Y(13)=X(38)*EXP(-.354+1.088*LOG(Y(ll)))3640 Y(14)=X(39)/(.025*Y(12)+X(23))3650 Z3=-15.6 + 46.1*LOG(Y(11))/2.30336 55 IF X(13)=0 AND X(14)=0 THEN Y(15)=0:Y(16)3660 Z5=LOG(X(18))3670 Y(15)= -3.58 + .0204*Z3 + 1.976*Z5-.3S46*3680 Y(15)= X(40)*(10~Y(15))3690 IF X(14)0 THEN Y(16)=X(14):GOTO 37103700 Y(16) - X(11)*(X(12)-X(13))/(X(12))3710 Y(17) = X(22)*Y(16)/Y(15)3720 Y(18) = Y(10)*Y(9)3730 Z3=Y(l)/((l! + .82*(Y(9r.45))*(X(7)*Y(3)))3740 Y(19)= 1E-03*(Z3".82)*(Y(1)/(X(7)))".183750 Z3 = -(Y(19)"(-.25))3760 Y(20)= EXP(-18.51-20.49*Z3)3770 Y(21)= EXP(-36.77-29 .33*Z3)37 80 Y(22)= EXP(-53.8-35.6 5*Z3)3790 Z3 = Y(20)+Y(21)+Y(22)3800 Y(20) = Y(20)/Z33810 Y(21)= Y(21)/Z33820 Y(22) - Y(22)/Z33830 RETURNOk)*X(30)+X(9)*X(31)'natural p'agric p'urban p'atmos loadload'inflow cone'overflow rate'residence time'1-rp'p spring'chl-a'chl-a max'secchi depth'tsi=0:Y(17)=0:GOTO 3720Z5*Z5: 'hod:'hypolimnetic depth:'days of oxygen supply:'p residence time'discriminantscore: 'prob(eutrophic):'prob(mesotrophic):'prob(oligotrophic)

•oowHHOi-l1Sfioaco covO 00PJ PIrt rtrt rtO Ort rt1 1co oro a*O t-'0 1V 0p. JOXCO-JwrtHOHIoa*h-"1{acoo^Pirtrtort1hifaIDrtCDaCOUlPIHrtOrt1SJPrtCO0"o.CO*-oc9Bv:oco coCO Nio •>C rtv£ rt^1I JoCOCO CO M N M M NH O ^ O C O —1 CT> l_ns: cjcHcsffiO H OT(D:t) rt H- 3 H- P O P D rort to aI B s4 t- 1 rt X i-t 3 H rt •aO t-' rt P t-< H- ro &.*< CO p. H a n> i- 1 H- O* I-" rto. *! Hi H- 3 H-a ro & >-• o CD A- 00 r>V ** a. cu 13 (0 ^ O H H- -3 £> A. O O- rt O hj• ^ f •tf 00 Hy o „ f B rt > ^'•mO ~~»o« (a OQ H-QQ >X) O rt O•*!(0~* O.0QBfC P ->. o rtp -•> ro•*. 5*- B sH O OQr>TJ r- •«•. rtco nj o-„B >tf O 0 o^OQ co •«OQ H CO rt0 ~-»« ^ - - 9 B3 0 0 OQ 0QOQ (W Co ~-~ir^-N*•»» •—N3**-^ >B B S >_ 1 S 0 OQ CO CO 0CO COv^*-• (-•CH Olo^BCOI— 1WUl H H HId vj Ul (Jl UlM M I O4>- CO ts5»-• I—W O NSJ(-•COA!*•0Q**^»oto-ali OHaHit- 1o«oaooHia•aCOrtt- 1P5^(DCOBOQ>X>acs3c«Jcn O00wJUCOH-aj>(D(Ua>~J O^ enPI COa a*rt oH Hcu n>t-«h3 p.ar 1 n>opaeaoHiHiB» Ol^!o &- n> 1ft) *0•ortrt ?^ H-3-OQ O000•-^•< aHCOfDOQiT)•^^yr 30o o o*-w H^t3Ot- 1H-BH"aH-Oao(Dt3rtCo |o 1— |vO 00o-~lS S >arta- crBK>ocrsn>rt B•° -3t—iVO CO 00FOrort**1WOr ^oroto >nroX- to0N5r K3ON3 rCT-CnaHO*(0a>HroCor KJO to OHH-t-»I- 1roCL>0QHH-Ot>HroCoroo*-c3artH-i- 1t- 1roa-t>0QHH-n>HroCo^ ri—• I—Co S> l-«SH-X^ioHWCoHP-roa. s;Ooa.O O O O O O O O O O O O O O O O O O O O O O ^ - O O C n O O I - ' l — O O O H U M v J N i o f r [>O O O O O O O O O O O O O O O O O 0 0 O C n O O O O O 0 > O O O O O v 0 O 0 0 t s 3 O O C ^ C 0 4 i > ghr|H>Horopr N>>nroPr N>»-••—OoH-HiroH*riOrt>rtCDPr K)COMHtaMt>wt- 1PIwStoHI-' CO t)O O O O O O O O O C O C O H - 0 > C O C O C O C O O O O C O O O O O O O O O O O O O O O O O O O O• • • • oN i C o C O C o u i C O O O O O O O C O O O O O O U i O O O O O O H - O O O O O O O O O O O O O O PIC O M 3 v 0 - ^ C n O O O O O O O O O O O O H - ' O K - O O O O O C O O O O O O O O O O O O O O O c^a^^ooQ

LIST OUTPUT VARIABLESinput file - runlDEPENDENT VARIABLES1 Total P Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m3 «edF-8 Overflow Rate m/yr «£—9 Residence Time yr10 1 - P Retent CoefIIP Spring mg/m312 Chlorophyll-a mg/m313 Max Chl-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Hypol Depth m17 Days of 02 Supply18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)23 Dummy24 Dummy25 DummyCODE42222o22444442422111000

APPENDIX CLEAP Output By Lake

unl—sensitivityPREDICTED VARIABLEMEANSTD DEVCVPOKiAOseet\5% 95%m. Total P Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone rag/m38 Overflow Rate m/yr9 Residence Time yr10 1 - P Retent Coef11 P Spring mg/ra312 Chlorophyll-a mg/m313 Max Chi-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Hypol Depth in17 Days of 02 Supply18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)1145.72358.04100.36364.31287.1035.9027.744.321.900.4813.245.6911.674.500.413.60105.620.910.020.000.540.46377.29165.1143.95174.9095.707.187.271.300.570.145.022.926.622.300.170.0044.890.330.000.010.260.260.330.460.440.480.330.200.260.300.300.290.380.510.570.510.420.000.430.370.251.990.480.57592.99142.3641.80139.47147.4024.0616.422.361.050.266.212.043.751.620.183.6045.140.440.01-0.010.02-0.062213.64900.51240.97951.61559.1953.5546.867.883.450.8628.2615.8936.3012.510.963.60247.121.890.030.011.050.98runl-sensitivityREDICTED VARIABLE: 11 P Spring mg/m3k AN = 13.2436STD. DE^i = 5.01862•INPUT VARIABLEMEANCVCV= .378948DY/DXDLY/DLX% VAR.1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km28 Upstr Lake Area km29 Wetland Area km210 Upstr Lake Ret Fac km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr20 Inflow Cone of Upst Lakes (rag21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m330 Upstream Lake P mg/m3I Wetland P mg/m3• -I Atmos P Load kg/km2-yr35 Error - Watshd36 Error - P Reten7.8644.1711.896.412.396.079.571.895.4415.958.200.43717.9715.000.0515.0015.0015.0015.0057.00139.0015.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.3010.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.550-0.031-0.031-0.031-0.0310.1780.587-0.084-0.031-0.031-0.075-0.378-1.4050.001-0.0808.2990.0390.2200.0590.0320.0120.0300.0090.0270.1119.510-6.885-0.018-0.103-0.028-0.0150.0320.269-0.060-0.004-0.013-0.090-0.234-0.0460.031-0.0900.0310.0440.2500.0670.0360.0510.3180.0110.0310.2510.718-0.5200.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1320.0000.2260.0270.0551.7350.1260.0370.0223.5020.0030.0264.85832.31756.932

unl-sensitivityREDICTED VARIABLEMEANSTD DEV•1 Total P Load kg/yr670.51 185.742 Undeveloped P Load kg/yr 118.82 46.263 Agricultural P Load kg/yr 190.79 72.264159.83 68.47Urban Runoff P Load kg/yr167.10555.70Atmospheric P Load kg/yr33.97 6.796 Septic P Load kg/yr34.95 8.507 Inflow P Cone mg/m33.44 0.728 Overflow Rate ra/yr1.58 0.3390.50 0.14Residence Time yr1017.41 6.301 - P Retent Coef7.36 3.7011 P Spring mg/m315.71 8.7112 Chlorophyll-a mg/m33.79 1.9913 Max Chl-a mg/m30.33 0.13140.94 0.00Secchi Depth m34.70 14.2815 Oxygen Depl Rate g/m2-day0.79 0.2516 Hypol Depth m0.02 0.0117 Days of 02 Supply0.01 0.0218 P Residence Time yrs0.74 0.16190.25 0.18TS Discr Score20runl-sensitivityProb(Eutrophic)21 Prob(Mesotrophic)PREDICTED 22 Prob(Oligotrophic)VARIABLE: 11 P Spring mg/m3^lEAN = 17.4114STD. DEV = 6.30023 CVCV0.280.390.3 80.430.330.200.240.210.210.280.360.500.550.520.410.000.410.320.241.550.220.71= .361845Carm)5%385.3054.5489.4567.8685.7922.7721.492.271.040.298.442.695.191.330.140.9415.230.420.02-0.030.42-0.1195%1166.85258.88406.92376.48325.4750.6856.845.232.390.8735.9020.1047.6010.820.740.9479.041.490.040.051.050.61INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km28 Upstr Lake Area km29 Wetland Area km210 Upstr Lake Ret Fac km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr20 Inflow Cone of Upst Lakes (mg21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m3• 27 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m3Upstream Lake P mg/m3Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error Watshd36 Error - P Reten0.507.652.816.393.701.855.570.052.350.615.440.62679.4315.000.0515.0015.0015.0015.0057.00139.0015.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.2090.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.550-0.195-0.194-0.194-0.1940.4831.806-0.361-0.198-0.194-0.242-0.719-2.0420.001-0.01017,6430.0080.1230.0450.1030.0600.0300.0010.0380.14512.190-8.693-0.006-0.085-0.031-0.0710.1030.192-0.115-0.001-0.026-0.009-0.225-0.0730.051-0.0090.0510.0070.1060.0390.0890.1960.2380.0010.0330.2490.700-0.4990.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1780.0000.0020.0780.0020.3460.0460.2410.3572.1590.0000.0335.27133.69357.595

unl-sensitivity.PREDICTED VARIABLEMEANSTD DEVCV5%Cccto^95%w l Total P Load kg/yr39.732 Undeveloped P Load kg/yr8.203 Agricultural P Load kg/yr 10.344.314 Urban Runoff P Load kg/yr13.805 Atmospheric P Load kg/yr3.086 Septic P Load kg/yr38.437 Inflow P Cone mg/m32.258 Overflow Rate m/yr0.860.579 Residence Time yr21.7810 1 - P Retent Coef9.0811 P Spring mg/m320.0412 Chlorophyll-a mg/m33.2613 Max Chl-a mg/m30.000.0014 Secchi Depth m0.0015 Oxygen Depl Rate g/m2-day0.4916 Hypol Depth m0.0317 Days of 02 Supply0.0318 P Residence Time yrs0.8019 TS Discr Score0.1720 runl-sensitivityProb(Eutrophic)21 Prob(Mesotrophic)^fclEDICTED 22 Prob(Oligotrophic)VARIABLE: 11 P Spring mg/m3^EAN = 21.7771STD. DEV = 7.2424612.774.335.422.424.600.6110.610.940.360.147.244.4010.661.720.000.000.000.200.010.040.090.12CV0.320.530.520.560.330.200.280.420.410.250.330.480.530.530.000.000.000.410.221.290.110.72= .33257220.892.853.631.407.092.0622.120.970.370.3411.203.456.921.130.000.000.000.210.02-0.040.63-0.0775.5423.5729.5113.2526.384.5966.765.191.960.9442.3523.9258.099.390.000.000.001.100.040.100.970.41INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m3'31 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - WatshdError - P Reten0.410.830.410.580.430.100.460.111.920.3161.500.0515.0015.0015.0015.0057.00139.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.419 •0.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550-2.708-2.706-2.709-2.7074.43018.3841.960-2.711-2.196-15.9510.02733.7100.0700.1400.0700.0980.0740.0170.0190.25212.527-9.397-0.051-0.103-0.051-0.0720.0880.0840.041-0.014-0.194-0.2270.0770.0770.0480.0970.0480.0680.1930.1080.0130.3470.575-0.4310.0000.0000.0000.0000.0000.0000.0000.0000.0008.1980.0000.2170.0840.3390.0840.1650.4110.5290.00612.12026.92550.922

unl-sensitivityPREDICTED VARIABLE MEAN STD DEV CVCarirX?^2- 95%1 Total P Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m38 Overflow Rate m/yr9 Residence Time yr10 1 - P Retent Coef11 P Spring mg/m312 Chlorophyll-a mg/m313 Max Chl-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Kypol Depth m17 Days of 02 Supply18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)runl-sensitivityEDICTED VARIABLE: 11 P Spring mg/m3INPUT• AN =VARIABLEMEAN20.4006STD. DEV1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area11 Mean Depth mkm215 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshdk6 Error - P Reten71.6917.8916.7724.7 39.303.0032.237.170.460.6320.408.5418.673.410.000.000.000.290.030.060.840.11= 6.821630.601.140.220.850.270.300.310.043.320.6060.000.0515.0015.0015.0015.0057.00139.0015.0030.001.001.0023.117.036.4810.703.100.608.691.560.100.136.824.159.961.790.000.000.000.080.010.080.020.10CV0.0000.0000.0000.0000.0000.0000.0000.0000.0000.2180.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550CV0.320.390.3 9G.430.33C.200.270.220.220.200.330.490.530.530.000.000.000.270.271.460.030.96= .334384DY/DX-2.018-2.015-2.021-2.0175.10119.003-6.861-2.031-1.009-0.2020.01417.0710.1020.1930.03 80.1450.0460.0510.0060.08816.900-7.46237.628.157.7410.414.772.0118.804.640.300.4210.453.236.421.190.000.000.000.170.02-0.110.79-0.10DLY/DLX-0.060-0.112-0.022-0.0840.0670.278-0.104-0.004-0.164-0.0060.0420.0420.07 50.1420.0280.106,- **—.-i-Qlcx...1[1136.6239.2836.3358.7618.114.4855.2811.100.710.9539.8222.5554.259.760.000.000.000.500.050.220.880.32% VAR0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0010.0000.0630.2020.7220.0270.404~CL_178

unl-sensitivityPREDICTED VARIABLEMEANSTD DEVCVCtwior^5%95%Total P Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m38 Overflow Rate ra/yr9 Residence Time yr10 1 - P Retent Coef11 P Spring mg/m312 Chlorophyll-a mg/ra313 Max Chl-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Hypol Depth m17 Days of 02 Supply18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 runl-sensitivityProb (OligotrophiaJEDICTED VARIABLE:CAN = 16.3591INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area kra28 Upstr Lake Area km29 Wetland Area km210 Upstr Lake Ret Fac km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr20 Inflow Cone of Upst Lakes (mg21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m3£^ Upstream Lake P mg/m3^PL Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshd36 Error - P Reten408.98203.5061.47101.0227.3015.6822.4720.000,170.7316.366.9414.683.950.000.000.000.130.030.040.820.14127.7073.1821.5640.559.103.146.382.930.030.115.243.317.682.010.000.000.000.020.010.060.070.130.310.360.350.400.330.200.280.150.140.150.320.480.520.510.000.000.000.200.281.630.080.9811 P Spring mg/m3STD. DEV. = 5.24078 CV = .320359MEAN5.098.691.152.030.680.820.910.101.010.773.490.89313.6815.000.0515.0015.0015.0015.0057.00139.0015.0015.0030.001.001.00CV0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1460.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.550DY/DX-0. 167-0. 167-0. 168-0. 1681. 3252386931691685335723220020275471810. 3090. 0410, 0720, 0240, 0290, 0030, 0360, 0364,-1.-0,-0.-0.-0.0,0,-0.12.014-A,640.43 8219.0299.1330.4845.2614.0210.5112.7414.930.130.548.622.675.161.430.000.000.000.090.02-0.090.69-0.13DLY/DLX-0.052-0.089-0.012-0.0210.0550.212-0.094-0.001-0.010-0.025-0.1220.0170.038-0.0250.0330.1660.2830.0370.0660.0840.2470.0030.0330.0670.895-0.271763.69417.76123.96225.4653.1723.4039.6526.800.230.9831.0518.0141.8010.920.000.000.000.190.050.170.960.40% VAR.0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0060.0000.0250.0571.0743.1280.0550.1710.0842.9570.0000.0420.48370.22921.689

unl-sensitivity^EDICTED VARIABLEMEANSTD DEVCVf&irfieU(• 141'5%95%1 Total P Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m38 Overflow Rate m/yr9 Residence Time yr10 1 - P Retent Coef11 P Spring mg/m312 Chlorophyll-a mg/m313 Max Chi-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Hypol Depth m17 Days of 02 Supply ,18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)356.93148.5333.07111.4356.407.5026.747.101.020.5514.646.2513.014.230.402.8485.270.560.020.010.690.30106.0155.0511.5844.9818.801.507.241.080.150.145.363.167.252.180.170.0035.330.150.010.020.210.230.300.370.350.400.330.200.270.150.150.250.370.510.560.510.410.000.410.280.271.880.310.77197.0770.7716.4149.7028.965.0315.565.240.750.337.042.284.271.510.172.8437.240.320.01-0.020.26-0.16646.47311.7266.64249.80109.8511.1945.959.631.38.0.9030.4417.1739.6711.840.912.84195.280.970.040.041.120.76runl-sensitivityEDICTED VARIABLE: 11 P Spring mg/m3^TO3. AN = 14.637STD. DEV = 5.35797CV= .366057INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m33 2 Atmos P Load kg/km2-yr35 Error - Watshd^fc Error - P Reten1.478.831.080.730.490.941.880.227.230.85150.000.0515.0015.0015.0015.0057.00139.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1520.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550-0.220-0.219-0.220-0.2201.2504.118-1.094-0.220-0.4100.4290.0026.1510.0510.3090.0380.0260.0170.0330.0080.07712.016-6.595-0.022-0.132-0.016-0.0110.0420.264-0.141-0.003-0.2030.0250.0210.0210.0530.3170.0390.0260.0660.3120.0080.1580.821-0.4510.0000.0000.000• o.ooo0.0000.0000.0000.0000.0000.0110.0000.0130.0832.9970.0450.0210.0403.6170.0022.07045.26945.833

unl-sensitivityPREDICTED VARIABLEMEANSTD DEVCVRa*v*Y^*j \J • 4.J.Z.5% 95%1 Total P Load kg/yr213.122 undeveloped P Load kg/yr92.743 Agricultural P Load kg/yr 38.404 Urban Runoff P Load kg/yr 23.265 Atmospheric P Load kg/yr49.808.936 Septic P Load kg/yr23.947 Inflow P Cone mg/m35.368 Overflow Rate ra/yr3.300.429 Residence Time yr9.9610 1 - P Retent Coef4.3511 P Spring mg/m38.5612 Chlorophyll-a mg/m35.3013 Max Chl-a mg/ra30.3515.1014 Secchi Depth m519.2715 Oxygen Depl Rate g/m2-dayyrs1.3716 Hypol Depth m0.0217 Days of 02 Supply0.0018 P Residence Time0.3119 TS Discr Score0.6920 runl-sensitivityProb(Eutrophic)21 Prob(Hesotrophic)22 EDICTED Prob(Oligotrophic)VARIABLE: 11 P Spring mg/m3^WE, TEAM = 9.95992STD. DEV= 3.9754769.1341.0817.1011.1816.60.1.796.061.620.990.143.982.294.992.630.150.00229.220.530.000.000.230.23CV0.320.440.450.480.330.200.250.300.300.330.400.530.580.500.440.000.440.390.242.3.10.750.34= .3 99146111.4038.2415.768.9025.575.9814.422.931.810.224.481.522.661.960.1415.10214.770.630.01-0.00-0.150.23407.72224.9093.5460.8197.0013.3139.729.816.000.8022.1312.4927.5014.300.8415.101255.462.980.030.000.771.15INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR•12345679111516212425262728293132Conifer For Area kra2Hardwood For Area km2Mixed For Area km2Untilled Agric Area kra2Tilled Agric Area km2Urban Area km2Lake Area km2Wetland Area km2Mean Depth mRunoff m/yrShoreline Septic Use cap/yrSeptic P Factor (kg/cap-yr)Conifer For P mg/m3Hardwood For P mg/m3Mixed For P mg/m3Untilled Agric P mg/m3Tilled Agric P mg/m3Urban P mg/ra3Wetland P mg/m3Atmos P Load kg/km2-yrError - WatshdError - P Reten7.144.871.223.670.600.391.661.1417.700.43178.500.0515.0015.0015.0015.0057.00139.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.3020.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550-0.053-0.053-0.053-0.0530.7922.441-0.524-0.053-0.147-0.2780.0028.3430.1440.0980.0240.0740.0120.0080.0230.0787.215-5.782-0.038-0.026-0.006-0.0200.0480.095-0.087-0.006-0.261-0.0120.0420.0420.2160.1470.0370.1110.0690.1090.0340.2340.724-0.5810.0000.0000.0000.0000.0000.0000.0000.0000.0000.0080.0000.0441.1750.5460.0340.3110.0360.3720.0303.80829.64863.988

unl-sensitivityPREDICTED VARIABLE MEAN STD DEV CVHortWiet95%Total P Load kg/yr195.80"2 Undeveloped P Load kg/yr53.393 Agricultural P Load kg/yr 17.224 Urban Runoff P Load kg/yr 64.295 Atmospheric P Load kg/yr54.606.306 Septic P Load kg/yr25.437 Inflow P Cone mg/m34.238 Overflow Rate m/yr1.329 Residence Time yr0.5210 1 - P Retent Coef13.185.6611 P Spring mg/m311.6012 Chlorophyll-a mg/m34.5113 Max Chl-a mg/m30.4314 Secchi Depth m3.6315 Oxygen Depl Rate g/m2-day 100.620.6816 Hypol Depth m0.0217 Days of 02 Supply0.0018 P Residence Time yrs0.5119 TS Discr Score0.4820 Prob(Eutrophic)21 Prob(Mesotrophic)22 PREDICTED Prob(Oligotrophic)VARIABLE: 11 P Spring rag/m3^pul runl-sensitivity- 13.1773 STD. DE\/ - 4.7753863.1925.967.5930.8818.201.266.711.280.390.144.782.856.442.290.180.0041.450.240.000.000.260.26CV0.320.490.440.480.330.200.260.300.300.270.360.500.550.510.410.000.410.350.252.010.500.54= .362393102.6820.197.1324.6028.034.2215.002.310.730.306.382.073.831.640.193.6344.140.340.01-0.01-0.00-0.04373.36141.2041.57168.01106.359.4043.107.732.400.8927.2015.4835.1812.430.993.63229.351.390.030.011.031.01INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR•123456789101115162021242526272829303536Conifer For Area km2Hardwood For Area km2Mixed For Area km2Untilled Agric Area km2Tilled Agric Area km2Urban Area km2Lake Area km2Upstr Lake Area km2Wetland Area km2Upstr Lake Ret Fac km2Mean Depth mRunoff m/yrShoreline Septic Use cap/yrInflow Cone of Upst Lakes (mgSeptic P Factor (kg/cap-yr)Conifer For P mg/m3Hardwood For P mg/m3Mixed For P mg/m3Untilled Agric P mg/m3Tilled Agric P mg/m3Urban P mg/m3Upstream Lake P mg/m3Wetland P mg/m3Atmos P Load kg/km2-yrError - WatshdError - P Reten1.617.732.481.360.341.071.820.590.865.015.590.43126.0015.000.0515.0015.0015.0015.0057.00139.0015.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.3020.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.550-0.142-0.142-0.142-0.1431.0763.454-0.125-0.143-0.143-0.435-0.508-2.8400.003-0.1458.4800.0470.2240.07 20.0390.0100.0310.0170.0250.1229.07 9-6.318-0.017-0.083-0.027-0.0150.0280.281-0.017-0.006-0.009-0.166-0.216-0.0930.032-0.1660.0320.0530.2550.0820.0450.0430.3280.0200.0280.2790.689-0.4790.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.5980.0000.8340.0320.0861.9850.2050.0610.0174.0840.0120.0246.57932.52952.955

unl-sensitivityPREDICTED VARIABLEMEANTotal P Load kg/yr19.842 Undeveloped P Load kg/yr22.383 Agricultural P Load kg/yr 25.9442.874 Urban Runoff P Load kg/yr24.905 Atmospheric P Load kg/yr3.756 Septic P Load kg/yr39.167 Inflow P Cone mg/m33.693 Overflow Rate m/yr1.579 Residence Time yr0.5010 1 - P Retent Coef19.5411 P Spring mg/m38.2017.8212 Chlorophyll-a mg/m33.5113 Max Chl-a mg/m30.4014 Secchi Depth m2.3015 Oxygen Depl Rate g/m2-day 69.6316 Hypol Depth m0.780.0317 Days of 02 Supply0.0318 P Residence Time yrs0.8119 TS Discr Score0.1720 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)EDICTED VARIABLE: 11 P Spring mg/m3runl-sensitivity• AN = 19.5437STD. DEV. = 7.30121INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshd^.6 Error - P RetenMEAN0.692.620.502.060.780.90-0.830.545.780.3475.000.0515.0015.0015.0015.0057.00139.0015.0030.001.001.00STD DEV46.0311.2012.8322.828.300.7510.541.400.590.147.304.1810.041.880.170.0029.280.310.010.040.100.140.0000.0000.0000.0000.0000.0000.0000.0000.0000.3790.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550CV0.380.500.490.530.330.200.270.380.380.290.370.510.560.540.420.000.420.400.261.540.120.86CV = .373583Tro^ucts5%'55.598.239.6514.7812.782.5122.861.730.740.289.262.965.781.200.172.3030.020.350.02-0.060.61-0.12CV DY/DX DLY/DLX-0.857-0.856-0.857-0.8561.4916.076-2.094-0.858-0.759-0.7380.00812.23 00.0380.1470.0280.1150.0440.0500.0300.13514.871-9.740-0.030-0.115-0.022-0.0900.0600.27 9-0.089-0.024-0.224-0.0130.0310.0310.0290.1130.0210.0830.1280.3580.0230.2080.761-0.49895%258.3560.8769.74124.2848.505.5967.087.873.320.8941.2622.7554.9710.240.922.30161.471.740.050.121.000.45% VAR.0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0170.0000.0280.0250.3640.0130.2240.1444.5600.0163.43737.33953.835

unl-sensitivityPREDICTED VARIABLEMEANSTD DEVCV5% '/ 95%1 Total P Load kg/yr240.122 Undeveloped P Load kg/yr 106.953 Agricultural P Load kg/yr 16.9132.174 Urban Runoff P Load kg/yr65.405 Atmospheric P Load kg/yr18.696 Septic P Load kg/yr26.097 Inflow P Cone mg/ra34.228 Overflow Rate m/yr1.970.479 Residence Time yr12.3510 1 - P Retent Coef5.3311 P Spring mg/m310.8212 Chlorophyll-a mg/m34.6913 Max Chl-a mg/m30.392.0014 Secchi Depth m62.0215 Oxygen Depl Rate g/m2-day0.9316 Hypol Depth m0.0217 Days of 02 Supply0.0013 P Residence Time yrs0.460.5419 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 EDICTED Prob(Oligotrophic)VARIABLE: 11 P Spring mg/m3runl-sensitivityAN = 12.355STD. DEV = 4.5531471.1647.487.4815.2621.803.746.501.230.570.144.552.706.052.360.160.0025.870.340.000.000.240.24CV0.300.'440.440.470.330.200.250.290.290.300.370.510.560.500.420.000.420.360.231.920.520.45= .368526132.7544.016.9812.4633.5812.5315.852.351.100.265.911.933.531.710.172.0026.930.450.01-0.00-0.020.05434.34259.8740.9783.08127.3827.8842.937.573.510.8625.8214.6933.1112.850.892.00142.821.920.030.010.941.02INPUT VARIABLEMEANCVDY/DXDLY/DLX% VAR1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area11 Mean Depth mkm215 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardv/ood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m3• 32 Atmos P Load kg/km2-yr35 Error - WatshdError - P Reten3.4411.231.261.760.200.522.180.098.300.45373.820.0515.0015.0015.0015.0057.00139.0015.0030.001.001.000.0000.0000.0000.0000.0000.0000.0000.0000.0000.2920.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.550-0.112-0.112-0.112-0.1120.8492.727-0.251-0.113-0.351-3.1040.00319.2360.0790.2570.0290.0400.0050.0120.0020.1128.028-6.470-0.031-0.102-0.011-0.0160.0140.115-0.044-0.001-0.236-0.1120.07 80.07 80.0960.3120.0350.049. 0.0220.1340.0020.2720.650-0.5240.0000.0000.0000.0000.0000.0000.0000.0000.0000.7850.0000.17 80.2702.8720.0360.0700.0040.6570.0006.06927.98161.077

#runl-sensitivityPREDICTED VARIABLE234D678910111213141516171819202122Total P Load kg/yrUndeveloped P Load kg/yrAgricultural P Load kg/yrUrban Runoff P Load kg/yrAtmospheric P Load kg/yrSeptic P Load kg/yrInflow P Cone mg/m3Overflow Rate m/yrResidence Time yr1 - P Retent CoefP Spring mg/m3Chlorophyll-a mg/m3Max Chl-a mg/m3Secchi Depth mOxygen Depl Rate g/m2-dayHypol Depth mDays of 02 SupplyP Residence Time yrsTS Discr ScoreProb(Eutrophic)Prob(Mesotrophic)Prob(Oligotrophic)runl-sensitivityMEAN362.16100.88179.2545.1429.107.8025.2114.810.510.6215.686.6714.024.050.453.0581.570.320.030.030.800.17PREDICTED VARIABLE: 11 P Spring ng/m3^!AN = 15.6 829STD. DEV. = 5.45552INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km28 Upstr Lake Area km29 Wetland Area km210 Upstr Lake Ret Fac km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr20 Inflow Cone of Upst Lakes (mg21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m3^0 Upstream Lake P mg/m3^B. Wetland P mg/m3^W2 Atmos P Load kg/km2-yr35 Error - Watshd36 Error - P RetenMEAN5.043.801.116.802.910.490.970.050.320.267.610.67156.0015.000.0515.0015.0015.0015.0057,00139.0015.0015.0030.001.001.00STD DEV119.2738.3066.6619.029.701.567.062.880.100.135.463.297.622.080.180.0032.600.080.010.050.110.16CV0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1950.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.550CVCV0.330.380.370.420.330.200.280.190.190.210.350.490.540.510.400.000.400.260.281.700.140.93= .347865DY/DX-0.171-0.171-0.172-0.1711.0433.417-2.046-0.169-0.172-0.434-0.3491.5980.002-0.0076.7540.1460.1100.0320.1970.0840.0140.0010.0090.04214.085-5.906DLY/DLXParker5%187.4447.2185.2019.4314.945.2314.4110.040.350.417.822.484.731.450.203.0536.670.190.02-0.060.57-0.15-0.055-0.041-0.012-0.0740.1940.106-0.127-0.001-0.003-0.007-0.1690.0680.022-0.0070.0220.1390.1050.0310.1880.3070.1250.0010.0090.0800.898-0.37795%699.75215.56377.11104.8456.6811.6444.1321.850.760.9531.4517.9041.5911.331.003.05181.400.540.050.121.030.49% VAR.0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.1450.0000.0020.0150.6420.3660.0311.1720.9490.6380.0000.0030.59359.99035.453

unl-sensitivityPREDICTED VARIABLEMEANSTD DEVCVStCnt^n***.5%95%Total P Load kg/yr435.46% ' Undeveloped P Load kg/yr 144.313 Agricultural P Load kg/yr 62.23434.20Urban Runoff P Load kg/yr5103.50Atmospheric P Load kg/yr40.726 Septic P Load kg/yr28.957 Inflow P Cone mg/m34.368 Overflow Rate m/yr2.4690.45Residence Time yr1012.991 - P Retent Coef5.5811 P Spring mg/m311.4212 Chlorophyll-a mg/m34.5513 Max Chl-a mg/m30.4714 Secchi Depth ra5.4215138.44Oxygen Depl Rate g/m2-day1.1016 Kypol Depth ra0.0217 Days of 02 Supply0.0013 P Residence Time yrs0.53190.47TS Discr Score20runl-sensitivityProb(Eutrophic)21 Prob(Mesotrophic)22 PREDICTED Prob(Oligotrophic)VARIABLE: 11 P Spring mg/m3JfeEAN * 12.9854 STD. DEV. = 4.9086INPUT VARIABLE1 Conifer For Area kra22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km28 Upstr Lake Area kra29 Wetland Area km210 Upstr Lake Ret Fac km211 Mean Depth ra15 Runoff m/yr16 Shoreline Septic Use cap/yr20 Inflow Cone of Upst Lakes (rag21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m3m28 Tilled Agric P mg/ra329 Urban P mg/m330 Upstream Lake P mg/m3Wetland P mg/ra3Atmos P Load kg/km2-yr35 Error - Watshd36 Error - P RetenMEAN3.3813.722.883.621.241.213.450.090.581.2810.720.50814.4215.000.0515.0015.0015.0015,0057.00139.0015.0015.0030.001.001.00124.5861.5025.6138.3734.503.147.021.140.630.144.912.866.472.320.200.0058.780.400.000.000.240.240.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.2610.0000.2000.2000.2000.2000.2000.2000.1110.2230.2000.2000.3330.3000.5500.290.420.410.460.330.200.240.260.260.310.380.510.570.510.420.000.420.360.231.820.440.51CV = .378009245.7361.9327.3233.3453.1427.3017.832.591.470.246.102.003.631.640.205.4259.210.540.01-0.010.06-0.01CV DY/DX DLY/DLX-0.101-0.100-0.101-0.1010.5241.743-0.360-0.100-0.100-0.223-0.299-2.1280.001-0.01924.2860.0500.2040.0430.0540.0180.0180.0010.0090.1038.6 85-7.121-0.026-0.106-0.022-0.0280.0500.153-0.096-0.001-0.005-0.022-0.247-0.0820.094-0.0220.0940.0580.2360.0490.0620.0810.1930.0010.0100.2380.669-0.548771.69338.58141.74209.48201.5960.7547.017.344.120.8327.6615.5835.4812.621.105.42323.642.260.030.011.000.95VAR.0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.3130.0000.0140.2450.0941.5560.0680.1080.0561.3020.0000.0034.39328.17463.671

unl-sensitivityPREDICTED VARIA3LE MEAN STD DEV CVSMU«r'^5- J -6 95%Total P Load kg/yr"2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m38 Overflow Rate m/yr9 Residence Time yr10 1 - P Retent Cpef11 P Spring mg/m312 Chlorophyll-a mg/m313 Max Chl-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Kypol Depth ra yrs17 Days of 02 Supply18 P Residence Time19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophic)• runl-sensitivity260.5141.49129.3435.0854.600.0031.014.620.780.5817.897.5516.183.720.000.000.000.450.030.020.770.21EDICTED VARIABLE: 11 P Spring mg/m3AN = 17.8859 STD. DEV. = 6.1602INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area km211 Mean Depth m15 Runoff m/yr24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshd36 Error - P RetenMEAN0.762.000.957.223.310.581.822.633.610.4415.0015.0015.0015.0057.0039.0015.0030.001.001.0090.4018.1456.2616.7718.200.008.131.380.230.146.163.718.751.940.000.000.000.150.010.030.140.170.0000.0000.0000.0000.0000.0000.0000.0000.0000.2980.2000.2000.2000.2000.1110.2230.2000.3330.3000.5500.350.440.430.480.330.000.260.300.300.240.340.490.540.520.000.000.000.330.261.630.190.81CV = .344416130.1417.3054.1913.4828.030.0018.362.540.430.368.982.825.491.310.000.000.000.230.02-0.040.48-0.13521.4799.49308.7091.29106.350.0052.388.381.410.9335.6220.1747.7310.560.000.000.000.880.040.081.060.56CV DY/DX DLY/DLX % VAR.-0.302-0.302-0.302-0.3010.9533.408-0.558-0.302-0.940-0.7570.0230.0600.0280.2160.0990.0170.0790.12514.137-7.539-0.013-0.03-4-0.016-0.1220.1760.110-0.057-0.044-0.190-0.0180.0190.0500.0240.1810.3150.1350.0660.2100.790-0.4210.0000.0000.0000.0000.0000.0000.0000.0000.0000.0260.0120.0850.0191.1081.0240.7600.1474.11547.39945.304

unl-sor.sitivityPREDICTED VARIABLEMEANSTD DEVCV5%siztv95%# Total ? Load kg/yr2 Undeveloped P Load kg/yr3 Agricultural P Load kg/yr4 Urban Runoff P Load kg/yr5 Atmospheric P Load kg/yr6 Septic P Load kg/yr7 Inflow P Cone mg/m38 Overflow Rate m/yr9 Residence Time yr10 1 - P Retent Coef11 P Spring mg/m312 Chlorophyll-a mg/m313 Max Chl-a mg/m314 Secchi Depth m15 Oxygen Depl Rate g/m2-day16 Hypol Depth m17 Days of 02 Supply18 P Residence Time yrs19 TS Discr Score20 Prob(Eutrophic)21 Prob(Mesotrophic)22 Prob(Oligotrophia)• runl-sensitivity42.1321.894.478.276.900.6021.988.330.180.7315.976.7814.304.010.000.000.000.130.020.010.740.25EDICTED VARIABLE: 11 P Spring mg/m3AN = 15.9679 STD. DEV. = 4.85027INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area11 Mean Depth mkm215 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/ia331 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshd^6 Error - P RetenMEAN0.790.910.390.340.030.090.230.091.480.6712.000.0515.0015.0015.0015.0057.00139.0015.0030.001.001.0013.008 .231.743.482.300.125.851.620.030.114.853.177.312.020.000.000.000.030.010.020.180.200.0000.0000.0000.0000.0000.0000.0000.0000.0000.1950.0000.2000.2000.2000.2000.2000.1110.2230.2000.3330.3000.5500.310.380.390.420.330.200.270.190.190.150.300.470.510.500.000.000.000.230.261.740.240.80CV = .30375122.7210.322.053 .563.540.4012.915.650.120.548.702.665.141.460.000.000.000.080.01-0.030.39-0.1578.1046.459.7419.2113.440.9037.4312.300.260.9829.3117.2739.7610.970.000.000.000.200.040.061.090.64CV DY/DX DLY/DLX % VAR.-1.079-1.078-1.081-1.0819.55030.298-2.029-1.082-1.325-1.2950.0194.5470.2010.2310.0980.0850.0070.0230.0230.08713.125-4.355-0.054-0.062-0.026-0.0230.0170.169-0.029-0.006-0.123-0.0540.0140.0.140.1890.2170.0920.0800.0260.1960.0210.1640.822-0.2730.0000.0000.0000.0000.0000.0000.0000.0000.0000.1200.0000.0091.5472.0470.3680.27 90.0092.0760.0193.23065.90324.391

unl-sensitivityPREDICTED VARIABLEMEANSTD DEVCVw5%Mca/io—95%• Total P Load kg/yrUndeveloped P Load kg/yrAgricultural P Load kg/yrkg/yrUrban Runoff P Loadkg/yrAtmospheric P LoadSeptic P Load kg/yrInflow P Cone mg/m39 Overflov; Rate m/yr1011 Residence Time yr12 1 - P Retent Coef13 P Spring iag/m314 Chlorophyll-a mg/m31516Max Chl-a mg/m317 Secchi Depth ra18 Oxygen Depl Rate yrs g/m2-day19 Hypol Depth m20 Days of 02 Supply2122P Residence TimeTS Discr Scorerunl-sensitivityProb(Eutrophic)Prob(Mesotrophic)Prob(Oligotrophic)•98.8035.0524.809.9428.500.5327.733.750.270.6919.058.0017.323.570.000.000.000.190.030.020.770.22REDICTED VARIABLE: 11 P Spring mg/m3EAN = 19.045 STD. DEV. = 5.72999INPUT VARIABLE1 Conifer For Area km22 Hardwood For Area km23 Mixed For Area km24 Untilled Agric Area km25 Tilled Agric Area km26 Urban Area km27 Lake Area km29 Wetland Area km211 Mean Depth m15 Runoff m/yr16 Shoreline Septic Use cap/yr21 Septic P Factor (kg/cap-yr)24 Conifer For P mg/m325 Hardwood For P mg/m326 Mixed For P mg/m327 Untilled Agric P mg/m328 Tilled Agric P mg/m329 Urban P mg/m331 Wetland P mg/m332 Atmos P Load kg/km2-yr35 Error - Watshd_36 Error - P RetenMEAN1.114.040.781.810.900.230.95,49,02,3210.500.0515.0015.0015.0015.0057.00139.0015.0030.001.001.0037.2218.3712.885.539.500.107.421.550.110.125.733.738.821.830.000.000.000.070.010.030.140.16CV0.0000.0000.0000.0000.0000.0000.0000.0000.0000.4120,000.20000.2000.2000.2000.2000.1110.2230.2000.3330.3000.5500.380.520.520.560.330.200.270.410.410.180.300.470.510.510.000.000.000.390.241.520.180.75CV = .300865DY/DX-0.536-0.534-0.536-0.5352.0156.9991.512-0.535-2.6 24-9.128 2 ,0260.010,0680 ,2450 ,0470.1100.0550.0140.0900.18313.450-5.94846.5112.288.773.2614.630.3516.251.640.120.481G.433.156.261.280.000.000.000.090.02-0.040.49-0.11DLY/DLX0.0310.1130.0220.0510.0960.0830.0750.042-0 1400.1510.0050.0050.0530.1930.0370.0870.1640.1010.0710.2880.7060.312209.90100.0070.1030.2455.510.7847.358.550.610.9934.7620.3247.979.950.000.000.000.410.040.071.040.54% VAR.0.0000.0000.0000.0000.0000.0000.0000.0000.0004.2880.0000.0010.1261.6480.0610.3310.3650.5560.22310.21449.59032.596

APPENDIX DCharts of Observed and Predicted Lake Responses by Lake and Yearobs: = observed values, symbol = last digit of yearest: = estimated values, symbols L = landsat land useP = planning map land useX = "best" land use* = overlay of two or more symbolsminimum scale values = 0 in each case

case:Bomo+ + + + + + + + + + +variable: 1 spring p scale max = 35obs:| 0 7* 9 Iest: I L XP I+ + + + + + + + + + +variable: 2 summer chl-a scale max = 25obs: I 09 1 8 Iest: I L XP I+ + + + +- + + + + + : +variable: 3 max chl-a scale max = 50obs: 1 9 8 Iest:I L * I+ + + + •—+ + •—+ + + + +variable: 4 secchi scale max = 8obs:I 18 9 0 Iest:I * L I+-,--.—+ + + ,.+ + + + + + +variable: 5 areal hod scale max = .75obs:| • . * Iest:I L XP I+ + + +Bomoseen+ + + H + + +good agreement for all variablesPM closer than LS

case:Carm+ + + + + + + + + + +variable: 1 spring p scale max = 35obs:I 0* 1 8 Iest: I L X P I+ + + + + + + + + + —+variable: 2 summer chl-a scale max = 25obs:I 1 8 0 9 1est: j L X P I+ + + + + + + + + + +variable: 3 max chl-a scale max = 50obs: [* iest:I L X P I+ +—. +_ + + + + + + + +variable: 4 secchi scale max = 8obs:I *1 8 Iest: | P X L I+ + + + +- + + + + + +variable: 5 areal hod scale max = .75obs:I 7 1 Iest:I L X P ICarmi+ + + + + + + + + +—: +model under-predicts lake productivity, as measured bychlorophyll-a and transparencyPM closer than LSchlorophyll increases dramatically in summerstrong positive bias in observed chlorophyll vs obs. spring P plotobserved summer chl-a/spring P » 1.1internal and/or seasonal point-source loading indicatedsummer phosphorus measurements would be useful and likelyto be considerably greater than spring values

case:Ceda+ + + + + +-- + + + + +variable: 1 spring p scale max = 35obs: I 8 0 9 7 1 1est: j L X P I+ + + + +. ,— + + + + + +variable: 2 summer chl-a scale max = 25obs: | ]est:| L X P I+ + .—+ + + + + + + + +variable: 3 max chl-a scale max = 50obs: |Iest: I L X P I+ + .— + + + + __ + ..+ + + +variable: 4 secchi scale max = 8obs: | . Iest:I . p X L Ivariable: 5 areal hod scale max = .75obs: IIest:1*I+ + + + + + + _+ + + +Cedarmodel over-predicts spring Pwide variance between LS and PM, LS closerneed better land use definitionatmospheric loading component greatestassumed runoff value (.31 m/yr) seems low compared withtypical (.6 m/yr) and may be partially responsiblefor error, based upon sensitivity analysis

case:Curt+ + + + + + + + + + +variable: 1 spring p scale max = 35obs: i 0 9 7 1 8 Iest: i L . X P I+ + + + + + +— + + + +variable: 2 summer chl-a scale max = 25obs: I 7 Iest: i L X P I+ ._+ + + + + —+ + + + +variable: 3 max chl-a scale max = 50obs: ! 7 |est: | L X P i+ +_ + + + + ~ _ + + + + +variable: 4 secchi scale max = 8obs: I 7 Iest: ! • P X L I+ + ._+ + _ + + + + + + +variable: 5 areal hod scale max = .75obs:iIest: I *I+ + +—. + + +— +— + + + +Curtislarge variance between LS and PM, LS closer to observed dataLS/FM variance due to differences in urban land useneed better definition of urban land use

case:Elmo+ + + + + + + + + +variable: 1 spring pscale maxobs: I 7 * 01est: I L X P+ + + + + + + + + +variable: 2 summer chl-ascale maxobs: | 8 * *est:| L X P+ + + + + + + + + +variable: 3 max chl-ascale maxobs:| *7 0est: I L X P+ + + + + + + + + +variable: 4 secchiscale maxobs:I 0*1est: I P X L+ + + + + + + + + +variable: 5 areal hodscale maxobs: Iest: |*+ + + + + + + +—, + +Elmoremodel over-predicts productivity, as measured by P and Chl-aLS closer than PMOver-prediction of transparency due to higher non-algalturbidity/color than other lakes, as indicated byobserved secchi vs. chl-a plot

case:Fair+ b + h h +- 1 (. + 1-variable: 1 spring p scale maxobs:| * 1 0est: I L XPvariable: 2 summer chl-a scale maxobs:I 8 1 0 9est: I L XP+ + + + + + + ;— + + +variable: 3 max chl-a scale maxobs:I 8 9est:I L XP+ + + + + + + + + +variable: 4 secchi scale maxobs:I 91 0 8est:I PX L+ + + +__ + .—+ + + + +variable: 5 areal hod scale maxobs:I 1est: I L X PFairfieldproductivity under-predicted by all land use estimatesPM closer than LSphosphorus export probably underestimated for approx. 15%of watershed which is sedimentaryinternal loading may also be important, as indicated byinternal loading model

case:Harv+ + + + 1- b + + + + hvariable: 1 spring p scale max = 35obs:| 78 90 1 |est: I L X P I+ + + + + + + + + + +variable: 2 summer chl-a scale max = 25obs: 1 8 * 0 Iest: | LX P I+ + + +—, + + +_ + + + +variable: 3 max chl-a scale max = 50obs: | 8 7 90 Iest:[ L X P. |+ + + + + + + + + + +variable: 4 secchi scale max = 8obs:| 9 1 *0 Iest: I P X L I+ + + + + + + + + + +variable: 5 areal hod scale max = .75obs:| * |est: | L X P I+ + + + +~ + + .+ + +Harveysslight under-prediction of spring P estimates, exaggerated+by high 1981 spring P value, which may be an outlier (check)bias in P may be due to effect of S. Peacham Brook, as describedin 1981 reportagreement reasonable for chlorophyll, hodobserved chl estimates variable because of hypolimnetic populationsunder-prediction of transparency due to low turbidity andhypolimnetic algal populations, which have no impact on secchicheck morphometry (mean depth)

caserHort+ + + + + +_. + + + + +variable: 1 spring p scale max = 35obs:I 7 8 0 9 1 Iest:I L * I+ •—+ + + + + + + + + +variable: 2 summer chl-a scale max = 25obs:19 0 1 Iest:I L * I+ + + + + + +—• + + + +variable: 3 max chl-a scale max = 50obs: I * Iest:I L XP I+ + + + + + +- + + + +variable: 4 secchi* scale max = 8obs:I 1 0 9 Iest: I * L I+ + + +—. + + + + + + +variable: 5 areal hod scale max = .75obs:I 7 1 9 Iest: | L XP I+ + + + + + + -+ + + +Hortoniareasonable agreement for all variablesPM closer than PSmeta/hypolimnetic algal populations

case:Iroq+ + + + + + + + + + [.variable: 1 spring p scale max = 35obs:I 3 9 10 71est: I L X P I+ + + + + + + + + + +variable: 2 summer chl-a scale max = 25obs:| 0 79 8 1 Iest: I L X P I+ + + + + +— + + + + +variable: 3 max chl-a scale max = 50obs:I 9 5 1est:I L X P ' I+ +— + + + + + + + + +variable: 4 secchi scale max = 8obs:I 1 8* 0 Iest: I P X L I+— + +; + + —+ + + + + +variable: 5 areal hod scale max = .75obs:I 8 7 * Iest: | L X P !+ + + + + + + + + + +Iroquoissubstantial under-prediction of P and productivityPM closer than LSerrors explained by internal loading model

case:Park+ +_, + + +_ + + + + +variable: 1 spring p scale maxobs:I 8 7 0 9 1est:I L X ?+ + + + + + + + + +variable: 2 summer chl-a scale masobs:I 7 19 0est:I L X P+ + —+ + + +_ + + + +variable: 3 max chl-a scale maxobs:I 7 9est:| L X P+—• + + + + + + + + -+variable: 4 secchi scale maxobs:I 1 70 9est:I P X Lvariable: 5 areal hod scale maxobs: I 1est:I L X P+ __+ + .—+ + + + + + +Parkergood agreement between best land use estimates and datalarge variance been LS and PMland uee needs better definition

case:St C+ + + + + + + + + + +variable: 1 spring p scale max = 35obs:I *9 0 1 Iest: I L X P !+ + + + + + + + + + +variable: 2 summer chl-a scale max = 25o b s : 1 8 * 1Iest:I L X P I+ + h h + h + 1- 1- + Hvariable: 3 max chl-a scale max = 50o b s : 1 8• Iest:I L X P I+—. + + + + + + + + +: +variable: 4 secchi scale max = 8obs:I 1 9 0 8|est: I P X L I+ + + + + + + + + + +variable: 5 areal hod scale max = .75obs:I 19 Iest: I L X P I+ + + -+ + + + + + + +St Catherineobs P and productivity slightly lower than best land use estimatesLS closer than PMmeta/hypolimnetic algal populations

case:Shel+ +— + + +- + + + + + +variable: 1 spring p scale max = 35obs:I* jest: | L X P |+ + +• + + + + + + + +variable: 2 summer chl-a scale max = 25obs:I *|est: i L X P |+ + + + + + + + + + +variable: 3 max chl-a scale max = 50obs:I *|est:I L X P |+ + + + + + +_ + + + +variable: 4 secchi scale max = 8obs:19 8 |est:I P X L I+ + + + + + + + + + +variable: 5 areal hod scale max = .75obs: | |est:I* |+ + +- +— + + + + + + +ShelbumeP and productivity under-predicticted by about 6xwide variance between PM and LS, PM closerneed better estimate of tilled agric landmost (at least 75%) is sedimentary and poorly drained (hsg D)export coefficients substantially higher than those used

case:Star+ + + + + + + + + + +variable: 1 spring p scale max = 35o b s : 1 9 8 0 1 Iest:| L * I+ + + +—• + + +—• + + + +variable: 2 summer chl-a • scale max = 25obs:I 9 0 1est:I L* I+ + + +—. + + + + + +_—.—+variable: 3 max chl-a scale max = 50obs:I 9 0 1est:I L * I+ + + + + + + + + + +variable: 4 secchi scale max = 8obs:I * Iest: I *L I+ •+ + + + + + + + + +variable: 5 areal hod scale max = .75obs: IIest:I* |+ + + —+ + + +- + + + +Staragreement with average spring P ok, but large variance inobserved values (check)chlorophyll substantially under-predictedoutlier on observed chl-a vs obs. spring P plotobserved Chl-a/Spring P • 1.44substantial increases in chl-a in early summer may be dueto seasonal loading componentare you sure there are no shoreline resorts/public facilitiessummer P measurements would be usefulobserved chlorophyll vs. secchi suggests high non-algalturdibity/color

caserWino+ + + :- + + + + + + + — +variable: 1 spring p scale max = 35obs: I 9 8 1 0 7 Iest:I L XP . I+ + + + + + + + + + +variable: 2 summer chl-a scale max = 25obs: I |est: j L * ' I+ + + + H + h + + + Hvariable: 3 max chl-a scale max = 50obs: I |est: I L XP |+ +— + +-* + + + + + + +variable: 4 secchi scale max = 8obs:I |est: I * L |+ (• -—+ + + + + + + + +variable: 5 areal hod scale max = .75obs:I |est:I* |+ + + + + —+ —+ + + + +Winonamodel under-predicts spring P valuesPM closer than LSabout 40% sedimentary - export coef. under-estimated

APPENDIX EAnalyses of VarianceObserved Lake Quality

anova for variable 2 psprLakeBomoseenCarmiCedarCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarWinonan555554•555655545mean2.69692.993672.840792.497652.52392.994092.627732.456813.391813.298552.729832.464244.723912.618763.25569std dev.165688.171697.317351.30592.157471.192948.308131.170699.194939.366193.281696.219791.290235.589563.278044variance.0274525.02948.100712.0935869.024797.0372289.094945.0291381.0380011.134097.0793524.0483079J1B433£2^C347iS5-~^.0773087sourceamong groupswithin groupstotalhdof145973sum sq23.83064.7341928.5648mean sq1.70218.0802406F= 21.2135H»edLci. '"anova for variable 4 chlaLakeBomoseenCarmiCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarn4415453544422mean1.681423.081431.840551.440862.346911.271941.301142.352462.252991.826321.158084.327632.98435std dev.294419.1850770.240302.359321.246462.318475.216424.315637.217274'.31115.346513.213745variance.0866827.03425340.0577452.129112.0607436.101427.0468392.0996266.0472082.0968145.120071.0456867sourceamong groupswithin groupstotalhdof123446sum sq27.85132.5110230.3623mean sq2.32094.0738534F= 31.4263

anova for variable 5 chlmaxLakeBomoseenCanaiCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarn2214242232122mean2.696814.174272.442352.13463.074881.967811.791763.603933.018292.752671.48164.946723.61709std dev.557524.02174910.119213.596278.2362420.544547.0390259.3437470.090394.170527variance.3108334.73022E-040.0142117.355547.05581030.2965321.52302E-03.11816208.17108E-03.0290794sourceamong groupswithin groupstotalhdof121628sum sq24.3041.3319425.636mean sq2.02534.0832462F= 24.3295anova for variable 6 secchiLakeBomoseenCarmiCurtisElmoreFairfieldHarveysHortoniaIroquoisMoreyParkerSt CatherShelburneStarn4415453554422mean1.53389.6093751.3351.173691.048031.894891.59472.9755171.55171.333661.85382-.717742-.164252std dev.148826.1223910.0690249.304382.0558181.0626105.205933.235082.125137.164057.510626.0832851variance.0221491.014979504.76444E-03.09264853.11566E-033.92008E-03.0424086.0552638.0156592.0269146.2607396.93641E-03sourceamong groupswithin groupstotalhdof123547sum sq18.56551.2147819.7802mean sq1.54712.0347081F= 44.5752

anova for variable 7 hodLakeBomoseenCarmiFairfieldHarveysHortoniaIroquoisMoreyParkerSt Cathern221234312mean-.967584-1.43499-.798508-.994252-.892184-.790481-.682095-.916291-.904555std dev0.17770600.227016.152638.08178130.0524022variance0.031579500.0515365.02329836.68818E-0302.74599E-03sourceamong groupswithin groupstotalhdof81119sum sq.798504.2206691.01917mean sq.099813.0200608F- 4.97552anova for variable 3 tdoLakeBomoseenCarmiFairfieldHarveysHortoniaIroquoisMoreyParkerSt Cathern221234312mean4.733423.858024.327226.193854.666324.10833.860154.516345.07956std dev0.17770800.227008.152644.08177510.0524078variance0.0315800.0515328.02330026.68716E-0302.74658E-03sourceamong groupswithin groupstotalhdof81119sum sq9.28119.2206739.50186mean sq1.16015.0200611F= 57.8306