Layout 1 INTRODUCTION Cultural eutrophication is the most widespread cause of water quality degradation in lakes (Smith, 2003; Jeppesen et al., 2005; Schindler, 2012). Consequently, reversal of eu- trophication is probably one of the most common forms of lake restoration. Restoration of eutrophic lakes started half a century ago, after lake scientists found that eutrophication is caused by excess phosphorus (Schindler, 1977). Based on this finding, the restoration of eutrophic lakes requires the identification of appropriate phosphorus objectives, i.e., the reductions of phosphorus needed to restore healthy lake conditions. Phosphorus objectives are identified using fixed boundary values (cut-offs between different trophic states; Vollenweider and Kerekes, 1982; Nürnberg, 1996) or predictive water quality models (mathematical expres- sions predicting water quality indicators from phosphorus concentrations or phosphorus loadings; Dillon and Rigler, 1974; Vollenweider and Kerekes, 1982). Regard- less of how they are identified, phosphorus objectives are laced with uncertainty. Fixed boundaries are somewhat arbitrary because trophic states tend to overlap along the axis representing phosphorus concentration (Vollenwei- der and Kerekes, 1982). Objectives obtained from pre- dictive water quality models (Dillon and Rigler, 1974; Vollenweider and Kerekes, 1982) may perform poorly outside the calibration domain or when environmental conditions change during restoration. Consequently, phosphorus objectives should be considered as tentative hypotheses, which need periodic reviewing as the restoration progresses and the actual lake response be- comes better known. The aim of this study is to review the current phospho- rus objective for Lake Lugano, a moderately large (49 km2) and deep (288 m) Swiss-Italian lake that is being restored from eutrophication. The occasion for the study is provided by a significant milestone. In 1992, three decades ago, a special issue of the scientific journal Aquatic Sciences pre- sented a scientific roadmap to the restoration of the lake, which became the blueprint for subsequent restoration plans (CIPAIS, 2018). Articles in the issue identified goals to improve water quality and set the phosphorus objectives needed to achieve these goals (Barbieri and Mosello, 1992; Imboden, 1992). The three decades elapsed since the pub- lication of the special issue (and four decades since the be- ginning of the restoration in the early 1980s) offer a useful vantage point to evaluate the restoration’s progress and re- view its objectives. We focus on the progress achieved in reducing phos- phorus concentrations and improving water quality (de- scribed by four variables: chlorophyll a, primary production, fall hypolimnetic oxygen concentration, and areal hypolimnetic oxygen depletion). We selected these water quality variables because they are explicitly tied to restoration goals and three of them (chlorophyll a, pri- mary production, and areal hypolimnetic oxygen deple- tion) can be mathematically modelled. Additionally, because one basin of Lake Lugano is meromictic (North basin) and the other is holomictic (South basin), we fo- cussed on the mixolimnion for the North basin and the en- tire water column for the South basin. We address the following questions: 1. What progress has been achieved so far in reducing phosphorus concentrations and improving water quality? ARTICLE Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake (Lake Lugano, Switzerland and Italy) Fabio Lepori1*, Bianca Lucchini2, Camilla Capelli1, Federica Rotta1 1University of Applied Sciences and Arts of Southern Switzerland, Mendrisio, Switzerland; 2Department for Environmental Science, University of Basel, Basel, Switzerland ABSTRACT The restoration of eutrophic lakes requires the identification of phosphorus objectives, i.e., the phosphorus reductions needed to achieve desired water quality goals. Due to inherent uncertainty, phosphorus objectives need periodic revision as the restoration progresses. We used monitoring data from a deep southern Alpine lake (Lake Lugano, Switzerland and Italy) to assess restoration progress and revise the current phosphorus objective of 30 mg m–3. Because one basin of the lake is meromictic (North basin) and the other is holomictic (South basin), restoration focussed on the mixolimnion for the North basin and the entire water column for the South basin. Time series analyses indicated that, thanks to restoration, phosphorus concentrations in the lake declined to values compliant with the objective (~20-30 mg m–3). In contrast, little progress was observed towards achieving the main water quality goals (chlorophyll a ≤4 mg m–3, primary production ≤150 g C m–2 year–1 and oxygen concentrations ≥4 mg L–1). Using predictive models, we estimated that achieving these goals requires a phosphorus objective of <10 mg m–3, which would bring the lake back to the original oligotrophic state. The concentration of <10 mg m–3 is lower than the objectives predicted for other (mainly northern) deep Alpine lakes. The apparent sensitivity of Lake Lugano, which we attribute to unfavorable hydrodynamic conditions common in lakes south of the Alps (weak mixing and long stratification), calls for particularly attentive phosphorus management. Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 105 2. Based on the results, are the phosphorus objectives set thirty years ago sufficient to achieve the desired im- provement in water quality? 3. If not, what new phosphorus objectives should be set? We addressed these questions by analysing time series of monitoring data (question 1) and using a predictive mod- elling approach (questions 2 and 3). The time-series analy- sis summarizes and updates previously-published work (Lepori, 2019a; Lepori and Capelli, 2021). The predictive modelling is new. MATERIALS AND METHODS Study site Lake Lugano is a natural lake located at the southern edge of the Alps (Switzerland and Italy; Figure 1). The lake is divided by a causeway into two basins, the North basin and the South basin (Figure 1, Table 1). The North basin, which is deep (288 m) and has a long residence time (12.3 years), is nearly meromictic, i.e., it is stratified in an oxic mixolimnion and an anoxic, phosphorus-rich (total phosphorus concentrations: 90-270 mg m–3) moni- molimnion (the terms used to describe lake stratification are explained in Figure S1). Meromixis is not the original mixing regime of the basin, but was probably induced in the mid-1900s by increased precipitation of carbonates due to eutrophication (Barbieri and Mosello, 1992). Since the beginning of systematic monitoring in the early 1980s, the entire water column mixed only twice, in 2005 and 2006, following exceptionally cold winters (Holzner et al., 2009). The depth of the mixolimnion varies from year to year, but usually does not exceed ~100 m (the maxi- mum depth, excluding the 2005-2006 turnovers, was of 102 m in 2000). Typical mixolimnion depths are substan- tially less deep; e.g., the median long-term (1987-2021) depth is 63 m (DACD-SUPSI, 2022). The South basin, which is shallower (95 m) and has a shorter residence time Figure 1. Geographical position of Lake Lugano and the two sampling stations. Table 1. Morphometric characteristics of the two main basins and one sub-basin (Figino, part of the South basin) of Lake Lugano, Switzerland and Italy. North basin* South basin* Figino sub-basin Lake area (km2) 27.5 21.4 8.2 Lake volume (km3) 4.69 1.14 0.47 Mixolimnion depth Maximum: ~100 m - - Median: 63 m ZMAX (m) 288 95 95 ZHYPO (m) 71 (20-100 m layer) 46 (20-63 m layer) 45 *Data from Barbieri and Mosello (1992). ZMAX: maximum depth, ZHYPO: average hypolimnion depth. Non -co mmerc ial us e o nly F. Lepori et al.106 (1.5 years), is essentially holomictic, although it has skipped complete mixing following milder than usual winters (unpublished results). In both basins, turnovers occur once a year in late winter, usually in January-March. Thermal stratification starts after the late-winter turnover and ends in fall (September-November), when mixing to approximately 20 m of depth erodes the thermocline and allows gas and nutrient exchange between the epilimnion and the hypolimnion. In pre-industrial times Lake Lugano was oligotrophic (Niessen et al., 1992). In the 20th century, the lake under- went rapid eutrophication, reaching eutrophic state by the 1970s (Barbieri and Mosello, 1992). The first plans to re- store the lake were set out in the early 1980s by the Ad- ministration of Canton Ticino (Switzerland; D. A., 1982). Since then, the management of the lake has been coordi- nated by an international commission (International Com- mission for the Protection of Swiss-Italian waters, Italian acronym CIPAIS; www.cipais.org) established in the late 1970s to manage transboundary Swiss-Italian waters. The commission proposes restoration measures to the Swiss and Italian governments and coordinates their implemen- tation to ensure uniform application across the border. The restoration, which is ongoing, is pursued mainly through improved sewage management (Lepori and Roberts, 2017). To date, the restoration has achieved a substantial reduction in phosphorus external loadings and lake con- centrations (Lepori and Roberts, 2017; Lepori, 2019a). A recent application of a phosphorus budget model (Lepori, 2019a) indicates that in Lake Lugano the phosphorus con- centrations are nearly in equilibrium with the current ex- ternal loadings. Current water-quality goals and phosphorus objectives In the special issue of Aquatic Sciences, Imboden (1992) presented two water quality goals and a phosphorus objective based on Swiss policy. The water quality goals for eutrophic lakes were to (1) reduce primary production to ≤150 g C m–2 year–1 and (2) increase dissolved oxygen concentrations to ≥4 mg L–1 (anytime and anywhere, except in meromictic lakes). Imboden (1992) does not present a rationale for the primary production goal of 150 g C m–2 year–1, although the suggested value is close to 110 g C m–2 year–1, which marks the boundary between oligotrophy and mesotrophy according to Wetzel (1975). The oxygen goal of 4 mg L–1 is an estimate of the lowest tolerable oxy- gen concentrations for fish egg survival (Müller et al., 2012). The phosphorus objective was to reduce spring total phosphorus concentrations to 30 mg m–3 (Imboden 1992). Again, the rationale for the 30 mg m–3 objective is unex- plained, although it is close to the 25 mg m–3 cut-off be- tween mesotrophy and eutrophy proposed by OECD (Vollenweider and Kerekes, 1982). The international commission managing the lake (CIPAIS) adopted and operationalized these goals and the phosphorus objective, with some integrations. An impor- tant integration was the addition of a chlorophyll a goal (yearly average within 2.5-4 mg m–3; CIPAIS, 2018). Ad- ditionally, the CIPAIS defined the operational phosphorus, chlorophyll a, primary production and oxygen indicators necessary to measure the progress towards the respective objective or goals, which we listed in Table 2. For the North basin, the indicators focus on the 0-100 m layer, implying that the 0-100 m and 100-288 m layers are con- sidered to be isolated. This choice appears to follow Bar- bieri and Mosello’s (1992) suggestion that, for modelling purposes, the North basin is best described by a two-box model where the 0-100 m layer is isolated from the 100- 288 m layer. The oxygen goal was applied to the 20-100 m (North basin) or 20-95 m (South basin) layers (roughly corresponding to the hypolimnia of these basins, assum- ing a mixolimnion depth of 100 m), because hypolimnia present the greatest risk of becoming de-oxygenated. To simplify the analysis of hypolimnetic oxygen, we fo- cussed on the concentrations at the end of the stratification period, when concentrations tend to reach the annual min- imum (Müller et al., 2019). If oxygen concentrations meet Table 2. Main objectives, goals and operational indicators for the restoration of Lake Lugano, Switzerland and Italy. Objectives/goals Indicators Comments Phosphorus concentrations ≤30 mg m–3 Annual average (volume weighted) Assumes that the 0-100 m and concentrations of total phosphorus 100-288 m layers are isolated in the 0-100 m layer for the North basin (Barbieri and Mosello, 1992) and the 0-95 m layer for the South basin Chlorophyll a between 2.5 and 4 mg m–3 Annual average concentration The 0-20 m layer represents the euphotic zone in the 0-20 m layer Primary production ≤150 g C m–2 year–1 Cumulative annual value in the 0-20 m layer The 0-20 m layer represents the euphotic zone Oxygen concentrations ≥4 mg L–1 anytime Average (volume weighted) dissolved The 20-100 m and 20-95 m layer and anywhere in the lake (except for the concentration across the 20-100 m approximately correspond to the respective monimolimnion of the North basin) layer for the North basin and the 20-95 m hypolimnia (Figure S1). layer for the South basin Therefore, we refer to these concentrations as hypolimnetic concentrations Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 107 the 4 mg L–1 goal in these layers and at this moment, they would meet or exceed this goal nearly anytime and any- where in the lake (except for the monimolimnion in the North basin). Finally, for Lake Lugano, the CIPAIS has identified the qualitative goal of restoring a mesotrophic state (CIPAIS, 2018), a goal consistent with the Swiss law requirement that nutrient content in standing waters “should allow at most an average production of biomass” (Swiss Waters Protection Ordinance, 1998). Data sources We used data from monitoring sites in Gandria (near the deepest point of the North basin) and Figino (near the deepest point of the South basin; Figure 1, Table 1). These stations will be referred to as North basin and South basin. We included data from the period 1984-2021, for which systematic monitoring data are available for both stations. The time series of phosphorus, chlorophyll a, primary production, and areal hypolimnetic oxygen depletion were presented in specific papers, which provide detailed method information (Lepori, 2019a; Lepori and Capelli, 2021). In short, the average yearly total phosphorus con- centration in the lake (P, in mg m–3) was calculated from samples collected monthly at different depths throughout the water column. The monthly concentrations were vol- ume-weighted (North basin: 0-100 m, South basin: 0-95 m) and averaged within years. Chlorophyll a (CHL, in mg m–3) was measured by col- lecting water samples at different depths across the eu- photic layer (0-20 m) at time intervals ranging from two weeks to one month, depending on the season. CHL was measured spectrophotometrically from each sample after extraction in ethanol (Rand et al., 1975). Yearly average values were calculated by averaging values across depths and months (fortnightly CHL values were first averaged within months). Primary Production (PPr, in g C m–2 year–1) was calcu- lated using the 14C radiotracer method from monthly sam- ples. Yearly values were calculated from monthly values according to Gächter (1972) until 2014 and according to a mathematical procedure based on Finger et al. (2013) there- after (a comparison of the approaches indicated that the re- sults are comparable, unpublished results). Fall hypolimnetic oxygen concentration (HOFALL, in mg L–1) is the hypolimnetic volume weighted oxygen concen- tration measured at the end of the stratification period in the fall. In keeping with CIPAIS’ oxygen indicator (Table 2), the hypolimnion was operationally defined as the 20- 100 m layer for the North basin and the 20-95 m layer for the South basin. The end of stratification was identified by the interruption of the nearly linear hypolimnetic oxygen decline observed in spring-summer. Areal hypolimnetic oxygen depletion (AHOD, in g m– 2 day–1) is the linear rate of areal hypolimnetic oxygen content decline during the stratification period (areal oxy- gen content = volume-weighted hypolimnetic oxygen concentration × average hypolimnion depth). The strati- fied period was defined as the period starting with the late winter turnover, identified by a hypolimnetic oxygen con- centration peak, and ending with fall partial mixing, iden- tified as described above. AHOD (g m–2 day–1) was calculated using the equation: AHOD = AHOLW – AHOFALL TSTRAT where: AHOLW = areal hypolimnetic oxygen content (g m–2) after the late winter turnover AHOFALL = areal hypolimnetic oxygen content (g m–2) at the end of stratification TSTRAT = duration of stratification (days) For the North basin (Gandria’s station), hypolimnetic content was calculated using the volume and surface of the entire North basin (Table 1). For the South basin (Figino’s station), we used the volume and surface of the sub-basin of Figino, which represents approximately 40% of the en- tire South basin (Table 1). Time series analysis Temporal trends were analysed using Mann-Kendall tests. In cases of significant trends, we used moving av- erages (step = 3 years) to highlight temporal trends. In ad- dition, we used the averages of the first and last three years of the study period (1984-1986 and 2019-2021) to quantify the magnitude of the changes in percentage. Trend analyses were performed using the Excel template application MAKESENS 1.0 (Salmi et al., 2002). The re- sults report the Z statistic, the associated P-value, and the Sen’s slope (Q), i.e., the linear rates of change estimated using Sen’s method (Sen, 1968). Predictive models We predicted CHL (mg m–3) and PPr (g C m–2 year–1) from P by using the following OECD models (Vollenwei- der and Kerekes, 1982): CHL = 0.28×P0.96 PPr = 512 × P 28.1+ P where: P = average yearly total phosphorus concentration (g m–3) in the 0-100 m layer (North basin) or 0-95 m layer (South basin) Non -co mmerc ial us e o nly F. Lepori et al.108 The P-CHL model is empirical, whereas the P-PPR model attempts to model self-shading by phytoplankton and phosphorus limitation using a saturation function. We predicted AHOD from P by using a model based on Müller et al. (2019). Müller et al. (2019) examined the relationship between AHOD and areal phosphorus load- ings in a sample of European lakes (mostly Swiss lakes located north of the Alps, including transboundary lakes and Lake Annecy, France). They found a threshold pattern where AHOD increases linearly with increasing phospho- rus loadings to a threshold of 0.5 g P m–2, after which AHOD reaches a plateau of 1.1 g m–2 day–1. We developed a P-AHOD model using data annexed to the paper. Since the paper provided spring turnover concentrations, PSP, we converted spring turnover concentrations to annual aver- age concentrations using P = 0.9PSP (Chapra and Tara- pchak, 1976; for Lake Lugano, over the period 1987-2015, empirical conversion coefficients were esti- mated to be 0.91 for the North Basin and 0.88 for the South basin; unpublished results). Next, we fitted a piece- wise regression to predict AHOD from P (Figure S2). The piecewise regression yielded accurate predictions of AHOD using a sloping line below a breakpoint concen- tration of 17.68 mg m–3 and a constant concentration (plateau) above it: if P < 17.68 mg m–3, AHOD = 0.363+0.0396P if P ≥ 17.68 mg m–3, AHOD = 1.06 This model assumes that AHOD is largely due to min- eralization of primary production settling from the euphotic zone. The breakpoint separates an area of phosphorus sat- uration (>18 mg m–3) from an area of phosphorus limita- tion, where PPr increases linearly with P. The model is based on temporally averaged data (each lake is represented by a mean AHOD calculated over a period of 6-36 consec- utive years, depending on lake), so it is not designed to cap- ture year-to-year variation. For each model we assessed accuracy using the Mean Absolute Percentage Error (MAPE) and model bias using the Mean Absolute Error (MAE). Phosphorus objectives Phosphorus objectives can be calculated for CHL, PPr and HO goals using the models described above. Phos- phorus objectives to attain the CHL goal of 4 mg m–3 and the PPr goal of 150 g C m–2 year–1 were calculated directly using OECD’s P-CHL and P-PPr models. Phosphorus ob- jectives to attain the HO goal of 4 mg L–1 were calculated indirectly, by linking the HO goal to a matching AHOD goal, then using the P-AHOD model to calculate the phos- phorus concentration required to attain the AHOD goal (Müller et al., 2019). The link between the HO goal and the corresponding AHOD goal is provided by the theoret- ical maximum tolerable AHOD (AHODTOL, g m–2 day–1), the value that could prevent the development of hypolim- netic concentrations below a certain concentration goal. AHODTOL can be calculated using the equation (Müller et al., 2019): AHODTOL = ZHYPO TSTRAT × (HOLW – HOGOAL) where: ZHYPO = average hypolimnion depth (m) TSTRAT = duration of stratification (days; see AHOD model above) HOLW = hypolimnetic oxygen concentration during the late winter turnover (mg L–1) HOGOAL = hypolimnetic oxygen concentration goal (4 mg L–1) The values of TSRAT were obtained from our oxygen concentration data (the beginning and end of stratification were identified as indicated above). The values of ZHYPO were obtained from morphometric data (Table 1). For the South basin, we used a ZHYPO of 45 m, which corresponds to the 20-95 m layer. For the North basin, we defined the hypolimnion in two ways: as the 20-100 m layer (100 m is the maximum mixolimnion depth) and as the 20-63 m layer (63 m is the long-term median mixolimnion depth based on monitoring data, see above). We refer to these layers (ZHYPO = 71 m and 46 m, respectively) as the ‘max- imum’ and ‘typical’ North basin hypolimnia. However, the phosphorus objectives obtained assuming a hy- polimnion depth of 71 m were unrealistic (e.g., according to the results, even the observed P values would not cause hypolimnetic concentrations below 4 mg L–1, which is un- supported by the data). Therefore, we report only the phosphorus objectives obtained assuming a typical hy- polimnion depth of 46 m. Values of HOLW were obtained from analyses of deep oligo-mesotrophic lakes located South of the Alps (lakes Maggiore, Como, and Garda). In these lakes, the water column oxygen content at late-winter turnovers varies between approximately 6-10 mg L–1 depending on the depth of mixing (deeper mixing is associated with higher oxygen concentrations; Rogora et al., 2018). For this study we used a HOLW of 8 mg L–1 to represent typical years, a HOLW of 7 mg L–1 to represent years of lower- than-average mixing, and a HOLW of 6 mg L–1 to repre- sent years of weak mixing. Measured HOLW values for Lake Lugano are currently lower (data not shown), but we assume that HOLW would reach values in the 6-8 mg L–1 range following restoration to a mesotrophic or olig- otrophic state. Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 109 RESULTS Restoration outcome so far During the study period, Phosphorus concentrations (P) declined from 76 to 21 mg m–3 in the North basin (a 72% decline) and from 114 to 28 mg m–3 in the South basin (a 76% decline, Table 3; Figures 2A-B). Chloro- phyll a concentration (CHL) declined from 11.8 to 6.9 mg m–3 in the North basin (a 42% decline) and from 13.8 to 7.3 mg m–3 in the South basin (a 47% decline; Figures 2C- D). Primary Production (PPr) and Fall Hypolimnetic Oxy- gen Concentrations (HOFALL) showed no trends (Figures 2E-H). Primary production fluctuated around average val- ues of 323 g C m–2 year–1 in the North basin and 368 g C m–2 year–1 in the South basin; HOFALL fluctuated around average values of 3.4 mg L–1 in the North basin and 2.3 mg L–1 in the South basin. The Areal Hypolimnetic Oxy- gen Depletion (AHOD) showed no trend in the North basin (average = 0.8 g m–2 day–1), whereas AHOD dis- played a declining trend in the South basin (from 1.2 to 0.9 g m–2 day–1 over the study period, a 25% decline; Fig- ures 2I-J). PREDICTIVE MODELS The P-CHL model showed poor accuracy (MAPE: North basin = 90%, South basin = 48%; Figures 3A-B) and a tendency to overestimate the observed values [MAE (10th percentile, 90th percentile): North basin = 5.4 mg m–3 (-1.0 mg m–3, 10.9 mg m–3), South basin = 4.2 mg m–3 (0.2 mg m–3, 11.0 mg m–3]). In comparison, the P-PPr model showed moderate accuracy (MAPE: North basin= 23%, South basin=17%; Figures 3B-C) and no bias [MAE (10th percentile, 90th percentile): North basin = 31 g C m–2 year–1 (-76 g C m–2 year–1, 125 g C m–2 year–1), South basin = -18 g C m–2 year–1 (-109 g C m–2 year–1, 79 g C m–2 year–1)]. Concerning the AHOD model, the phospho- rus concentrations included in the study were always above the threshold of 18 mg m–3 (Figures 3E-F). There- fore, the model predicted the plateau AHOD value of 1.1 g m–2 day–1 for all years, in both lake basins. The data sup- ported the existence of a plateau in the North basin, whereas in the South basin AHOD decreased slightly with decreasing P. Model accuracy was moderate to modest (MAPE: North basin = 49%, South basin = 31%). The model tended to overestimate the observed AHOD values [MAE (10th percentile, 90th percentile): North basin = 0.3 g m–2 day–1 (0.0 g m–2 day–1, 0.5 g m–2 day–1), South basin = 0.2 g m–2 day–1 (0.0 g m–2 day–1, 0.4 g m–2 day–1)]. The P-AHOD graphs (Figures 3E-F) suggest that the model estimates were closer to the upper limit of the observed AHOD values than to the average values. Additional analyses supported this idea, showing that the predicted threshold of 1.1 g m–2 day–1 matched precisely the 90th per- centile of the observed AHOD values in both basins. PHOSPHORUS OBJECTIVES Based on the P-PPr model, achieving the CHL goal of 4 mg m–3 would require P values of 19 mg m–3, while the PPr goal of 150 g C m–2 year–1 would require P values of 12 mg m–3. The median duration of stratification TSTRAT was 210 days in the South basin and 202 days in the North basin. Using these values, achieving the oxygen goal of 4 mg L–1 would require AHOD values of 0.9 g m–2 day–1 during typical years (HOLW = 8 mg L–1), AHOD values of 0.6-0.7 g m–2 day–1 during years with lower-than-average mixing (HOLW = 7 mg L–1), and AHOD values of 0.4-0.5 g m–2 day–1 during years of weak mixing (HOLW = 6 mg L–1; double values are for the South basin and the North basin, respectively). According to the P-AHOD model, these AHOD values correspond to P of 14 mg m–3, 6-9 mg m–3, and 1-3 mg m–3. Table 3. Results of the time series analysis: Mann-Kendall tests (Z statistic and associate significance) and Sen’s slopes (Q). Time series Unit Test Z Significance Q P (N.B.) mg m–3 -4.53 *** -1.4 P (S.B.) mg m–3 -7.78 *** -2.1 CHL mg m–3 -2.93 *** -0.10 CHL mg m–3 -4.80 ** -0.16 PPr (N.B.) g C m–2 year–1 0.50 0.6 PPr (S.B.) g C m–2 year–1 -0.43 -0.4 HOFALL (N.B.) g L–1 -1.13 0.0 HOFALL (S.B.) g L–1 -0.13 0.0 AHOD (N.B.) g m–2 day–1 -0.55 0.00 AHOD (S.B.) g m–2 day–1 -3.29 *** -0.01 Significance: ** = P <0.01, *** = P <0.001. P, annual Phosphorus concentration; PPr, Primary Production; HOFALL, Fall Hypolimnetic Oxygen Concen- tration; AHOD, Areal Hypolimnetic Oxygen Depletion; N.B., North Basin; S.B., South Basin. Non -co mmerc ial us e o nly F. Lepori et al.110 Figure 2. Temporal trends of annual Phosphorus concentrations (P), Chlorophyll a (CHL), Primary Production (PPr), Fall Hypolimnetic Oxygen Concentration (HOFALL), and Areal Hypolimnetic Oxygen Depletion (AHOD) in the two main basins of Lake Lugano (North basin and South basin). Restoration objectives and goals are indicated where applicable (Table 2). Moving averages (step = 3) were used to highlight significant trends (Table 3). Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 111 DISCUSSION Thanks to the restoration program, during the study pe- riod, average Phosphorus concentrations (P) declined from high values typical of eutrophic lakes (76 and 114 mg m–3) to moderate values (20-30 mg m–3) well within the restora- tion objective of 30 mg m–3. Even so, the phosphorus de- cline did not result in the expected improvements in water quality. Only Chlorophyll a (CHL) declined consistently in both lake basins, although the concentrations at the end of the study period were still nearly twice as high (1.7-1.8×) as the restoration goal. Values of Primary Production (PPr) and Fall Hypolimnetic Oxygen Concentrations (HOFALL) not only remained far from the respective goals, but also did not change. Therefore, phosphorus concentrations in the 20-30 mg m–3 range, including the restoration objective of 30 mg m–3, are not sufficient to attain the CIPAIS’ water quality goals in Lake Lugano. The decline of CHL may at first glance appear consis- tent with the restoration’s underlying tenet that phytoplank- ton biomass is phosphorus limited. However, increased phosphorus limitation was unlikely, because primary pro- duction showed no decline (see below). We suggest that the decline in CHL was caused, in part, by an abrupt ecosystem change observed in 1989, marked by a sudden increase in large-bodied zooplankton and (presumably) greater grazing pressure on phytoplankton (triggers may have included a mild winter and a decline of zooplanktivorous fish, but the causal relationships remain unsolved; Lepori and Roberts, 2017; Lepori, 2019b). This regime shift was not the only cause of the CHL decline, because CHL appeared to decline (less markedly) even after 1989, but at this time we cannot advance any explanations for this pattern. The lack of significant changes in PPr and HOFALL could have been expected. Based on existing empirical models, P influences PPr and AHOD at low concentrations, but the effect saturates at moderate to high concentrations. Accord- ing to the OECD model used in this study, the effect of P on PPr saturates above concentrations of approximately 20- 30 mg m–3. Other studies suggest lower thresholds. Data from Lake Washington, USA, indicate that P limits PPr dur- ing the productive season up to concentrations of 20 mg m–3, beyond which PPr reaches a plateau (Smith, 1979). Similarly, according to Müller et al. (2019), AHOD reaches a plateau beyond a phosphorus loading of 0.54 g P m−2 (areal loading during the productive season), which corre- sponds to a yearly concentration of approximately 18 mg m–3 (Figure S2). These examples suggest that a phosphorus concentration of approximately 20 mg m–3 marks the tran- sition point between phosphorus control and phosphorus saturation in lakes. During the study period, phosphorus concentrations in Lake Lugano were nearly always >20 mg m–3, i.e., in the domain where phosphorus has little or no influence on PPr or AHOD. Our attempts to predict CHL, PPr and AHOD using models produced mixed results. CHL was not adequately predicted by using the OECD P-CHL model. The model bias and inaccuracy were not entirely unexpected, as other studies have remarked that global P-CHL models may perform poorly when applied to individual lakes (e.g., Smith and Shapiro, 1981). In Lake Lugano, the model’s tendency to overestimate CHL might have been caused by high grazing pressure by zooplankton (Lepori, 2019b), which could reduce phytoplankton biomass yields for unit phosphorus. For these reasons, we suggest that global P- CHL models, in their current form, provide an unreliable approach for setting lake restoration goals. The P-PPr model performed comparatively better. Despite the mod- erate accuracy, which may indicate methodological dif- ferences, measurement error, and influence by additional factors (Tadonléké et al., 2009) in the calibration data set, the model yielded unbiased predictions, which could be used to set the phosphorus concentrations needed to keep the medium-term PPr average (e.g., 4-5 years) near the restoration goal. The performance of the P-AHOD model was some- what below expectations, because the model had good predictive ability for the calibration data set (Figure S2). An explanation for the model tendency to overestimate the observed values in this study is that the initial oxygen concentration (HOLW) in Lake Lugano probably differed from those of the calibration set [Müller et al. (2019) men- tion 10 mg L–1 as a typical HOLW value, whereas we ob- served average values of 5-6 mg L–1]. This difference would bias the predictions because, in deep Alpine lakes, AHOD tends to be positively associated with HOLW (higher HOLW results in faster hypolimnetic oxygen de- pletion; Rogora et al., 2018, Lepori and Capelli, 2021). Therefore, in Lake Lugano, lower HOLW values could have caused lower AHOD values relative to lakes in the calibration set with similar phosphorus concentrations. In addition, as expected, the model did not predict year-to- year variation, which was substantial, probably because of parallel variation in mixing depth and HOLW (Lepori and Capelli, 2021). Despite these drawbacks, the P-AHOD model provided reasonable predictions of upper-range AHOD values. A bet- ter prediction of upper-range (rather than average) values can be explained by the tendency of higher AHOD values to be associated with higher HOLW concentrations, close to concentrations found in the calibration domain (not shown), for which the model should yield accurate results. There- fore, we suggest that the P-AHOD model may be useful to estimate the phosphorus concentrations that would keep AHOD within the maximum tolerable value (AHODTOL) most of the time (if the model predicted average AHOD values, the phosphorus concentrations predicted to keep AHOD at the maximum tolerable value would suffice only Non -co mmerc ial us e o nly F. Lepori et al.112 on average). In addition, the model bias may diminish as restoration progresses and HOLW increases in the lake. However, we also suggest that, in the future, the model should be further developed to improve its mechanistic basis (e.g., a AHOD model should additionally account for mixing) and predictive accuracy. Given that the current phosphorus objective (30 mg m–3) is not sufficient to attain the water quality goals, what new phosphorus objectives should be set for the lake? Ex- cluding CHL due the uncertainty of the P-CHL model, our results indicate that attaining the restoration PPr and HOFALL goals would require a reduction of P to 12-14 mg m–3 or lower. A reduction to 12-14 mg m–3 would help at- tain the PPr and HOFALL goals in typical years, assuming HOLW values close to 8 mg L–1. However, a phosphorus reduction to 12-15 mg m–3 would not be enough to attain the oxygen goal in years when mixing during turnovers is below average or weak (causing HOLW values of 6-7 mg L–1), or when stratification lasts more than ~200 days. In these years, only reduction to phosphorus concentrations below 10 mg m–3 would keep hypolimnetic oxygen con- centrations above 4 mg L–1 most of the time, as recom- mended by the restoration programme (the model predicts values between 1-9 mg m–3, but given the relatively low Figure 3. Observed (markers) and predicted (dashed line) Chlorophyll a (CHL), Primary Production (PPr), and Areal Hypolimnetic Oxygen Depletion (AHOD) versus Phosphorus concentrations (P). Gray lines connect data from consecutive years. Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 113 model accuracy we lump these values under the single category ‘<10 mg m–3’). The phosphorus concentration of 10 mg m–3 is considered a boundary between mesotrophic and oligotrophic conditions (Vollenweider and Kerekes, 1982). Therefore, our predictions suggest that meeting the oxygen goal will require a return to oligotrophic condi- tions, i.e., all the way to the reference state of the lake. We were surprised by the low AHOD values (≤0.4-0.9 g m–2 year–1) and low phosphors concentrations (≤14 mg m–3) that were predicted to reach the oxygen goal in Lake Lugano. According to Müller et al. (2019), deep lakes are resistant to hypolimnetic de-oxygenation. They estimated that, in lakes with ZHYPO greater than 32 m, even a plateau AHOD of 1.06 g m−2 day−1 would keep hypolimnetic oxy- gen concentrations at 4 mg L–1 at the end of stratification. Based on the P-AHOD model, in these lakes, even the highest phosphorus concentrations observed in Lake Lugano (~110 mg m–3) should not cause critical hypolim- netic de-oxygenation. In contrast, although the hy- polimnion of Lake Lugano is deep (ZHYPO of 45-46 m), the low tolerable AHOD and P values found in this study indicate high sensitivity to deep-water de-oxygenation. This sensitivity is probably linked to the lake’s location south of the Alps. In this region, the relatively milder cli- mate causes lake hydrodynamic conditions that are unfa- vorable to deep-water oxygenation. First, as already mentioned, milder winters can cause weak mixing during turnovers, which result in HOLW lower than 8 mg L–1. Sec- ond, in southern Alpine lakes the stratification period is longer (more than 200 days according to our data, whereas Müller et al., 2019 suggest a typical value of 180 days for their sample of lakes). Based on the AHODTOL equation used in this study, it is easy to see how the low initial oxygen concentrations and long stratification peri- ods make hypolimnetic de-oxygenation likely even at rel- atively low AHOD values. High sensitivity to de-oxygenation has important im- plications for the management of Lake Lugano and other deep southern Alpine lakes. For Lake Lugano and other mesotrophic or eutrophic lakes (e.g., Lake Iseo), only drastic phosphorus reductions could restore adequate oxy- genation in hypolimnetic waters. For oligotrophic lakes (lakes Maggiore and Garda), even small enrichment in phosphorus could lead to hypolimnetic de-oxygenation. This vulnerability will be heightened by climate change, which is expected to weaken mixing and cause longer stratification periods (Fenocchi et al., 2018). In the future, phosphorus objectives that prevent hypolimnetic de-oxy- genation may become progressively harder to attain. Thus phosphorus management and policy should be particu- larly attentive to deep southern Alpine lakes, especially as we progress towards a warmer climate. A caveat to our conclusions is that, for the North basin, the restoration focusses on the mixolimnion, as- suming isolation from the deeper phosphorus-rich, anoxic hypolimnion. This is, however, a simplistic assumption, because an analysis of internal loadings (Lepori, 2019a) indicated that during the exceptional turnovers of 2005- 2006, and the decade leading to these events, high internal loadings of phosphorus were transferred from the moni- molimnion to the mixolimnion. After the turnovers, the phosphorus upwelling subsided, and phosphorus concen- trations in the mixolimnion quickly returned to pre- turnover levels. The upwelling was caused jointly by an increase in the monimolimnion temperature and a de- crease in calcite precipitation, which weakened the den- sity difference between monimolimnion and mixolimnion (Holzner et al., 2009). Similar upwelling events will prob- ably repeat in the future. Ironically, the restoration may facilitate these events by reducing calcite precipitation in- duced by elevated primary production. Therefore, over the next decades, the recovery of the North basin will be punctuated by events of mixing and re-eutrophication. The magnitude and frequency of these events will be dic- tated by the long-term fate of the meromixis of the basin, which, lacking specific research, we are currently unable to predict. CONCLUSIONS Based on this study we propose the following recom- mendations for the management of Lake Lugano: • The water layers that may be realistically restored in the near future are the 0-94 m layer in the South basin (whole water column) and the 0-63 m layer in the North basin. • Predictions of the phosphorus objectives should be based on P-PPr and P-AHOD models. The P-CHL model used in this study is too uncertain to set phos- phorus objectives. • P should be reduced to at least 12-14 mg m–3, possibly to lower concentrations (<10 mg m–3) to ensure that the oxygen goal of 4 mg L–1 is met most of the time. Due to uncertainties in our predictions, and the high costs involved in reducing P, the reduction should proceed in increments, followed by further revisions of the P ob- jectives. An objective of approximately 15 mg m–3 may be used as an intermediate first increment. • A budget model can be used to estimate the correspon- ding reduction in external loadings needed to achieve the P objective (Lepori, 2019a). Developing better AHOD models that improve accu- racy, reduce bias, and account for variation in physical mixing during turnovers would be particularly helpful to future revisions of phosphorus objectives. As AHOD models seem to be regionally-specific, new AHOD mod- els should be calibrated based on the characteristics of deep southern Alpine lakes. Non -co mmerc ial us e o nly F. Lepori et al.114 Corresponding author: Fabio Lepori, Institute of Earth Sciences, University of Applied Sciences and Arts of Southern Switzerland, Campus Mendrisio, Via Flora Ruchat-Roncati 15, 6850 Mendrisio, Switzerland. E-mail: fabio.lepori@supsi.ch Authors’ contributions: All the authors made a substantive intel- lectual contribution. All the authors have read and approved the final version of the manuscript and agreed to be accountable for all aspects of the work. Conflict of interest: The authors declare no potential conflict of in- terest. Funding: The monitoring data used in the paper were collected for research programs funded by the International Commission for the Protection of Italian-Swiss Waters (CIPAIS). Availability of data and materials: The data are archived in data- bases maintained by University of Applied Sciences of Southern Switzerland (contact FL). Key words: Eutrophication, oligotrophication, primary production, water oxygenation, mixing. Acknowledgments: We gratefully acknowledge all the staff that contributed to the production of the data since the program’s be- ginning. Received: 20 November 2022. Accepted: 20 December 2022. Publisher’s note: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher. ©Copyright: the Author(s), 2022 Licensee PAGEPress, Italy Advances in Oceanography and Limnology, 2022; 13:11061 DOI: 10.4081/aiol.2022.11061 This work is licensed under a Creative Commons Attribution- NonCommercial 4.0 International License (CC BY-NC 4.0). REFERENCES Barbieri A, Mosello R, 1992. Chemistry and trophic evolution of Lake Lugano in relation to nutrient budget. Aquatic Sciences 54:219-37. Chapra SC, Tarapchak SJ, 1976. A chlorophyll a model and its relationship to phosphorus loading plots for lakes. Water Resources Research 12:1260-4. CIPAIS [Commissione Internazionale per la Protezione delle Acque Italo-Svizzere] 2018. Piano d’Azione 2019-2027. Turin, Italy. Available at: https://www.cipais.org/upload_files/ azione_2019-2027.pdf D.A., Dipartimento Ambiente [del Cantone Ticino], 1982. Il Lago Ceresio. Stato delle sue acque, obiettivi, misure d’intervento. [Lake Lugano: State of its waters, restoration schemes and goals]. Bellinzona, Switzerland, 85 pp. DACD-SUPSI. 2022. Ricerche sull’evoluzione del Lago di Lugano. Aspetti limnologici. Programma triennale 2019- 2021. Campagna 2021 e Rapporto triennale. Commissione Internazionale per la Protezione delle Acque Italo-Svizzere (Ed.); 93 pp. Available at: file:///C:/Users/fabio.lepori/ Downloads/1668585187S1-RL-CIPAIS_Rapporto_ Limnologia_2021_finale_Lugano%20(1).pdf Dillon PJ, Rigler FH, 1974. A test of a simple nutrient budget model predicting the phosphorus concentration in lake water. Journal of the Fisheries Board of Canada 31:1771-8. Fenocchi A, Rogora M, Sibilla S, Ciampittiello M, Dresti C, 2018. Forecasting the evolution in the mixing regime of a deep subalpine lake under climate change scenarios through numerical modelling (Lake Maggiore, Northern Italy/Southern Switzerland). Climate Dynamics 51: 3521-36. Gächter R, 1972. Die Bestimmung der Tagesraten der planktischen Primärproduktion—Modelle und In-situ- Messungen. Schweizerische Zeitschrift für Hydrologie 34:211-44. Holzner CP, Aeschbach-Hertig W, Simona M, Veronesi M, Imboden DM, Kipfer R, 2009. Exceptional mixing events in meromictic Lake Lugano (Switzerland/Italy), studied using environmental tracers. Limnology and Oceanography 54:1113-24. Imboden DM, 1992. Possibilities and limitations of lake restoration: Conclusions for Lake Lugano. Aquatic Sciences 54:381-90. Jeppesen E, Søndergaard M, Jensen JP, Havens KE, Anneville O, Carvalho L, et al., 2005. Lake responses to reduced nutrient loading–an analysis of contemporary long�term data from 35 case studies. Freshwater Biology 50:1747-71. Lepori F, 2019a. Il risanamento del Lago di Lugano: tendenze pluridecennali dei carichi e delle concentrazioni di fosforo. Bollettino della Società Ticinese di Scienze Naturali 107:13-19. Lepori F. 2019b. Effects of zooplankton structure and phosphorus concentration on phytoplankton biomass in a freshwater pelagic food chain. Fundamental and Applied Limnology/ Archiv für Hydrobiologie 192:305-17. Lepori F, Capelli C, 2021. Effects of phosphorus control on primary productivity and deep-water oxygenation: insights from Lake Lugano (Switzerland and Italy). Hydrobiologia 848:613-29. Lepori F, Roberts JJ, 2017. Effects of internal phosphorus loadings and food-web structure on the recovery of a deep lake from eutrophication. Journal of Great Lakes Research 43:255-64. Müller B, Bryant LD, Matzinger A, Wüest A, 2012. Hypolimnetic oxygen depletion in eutrophic lakes. Environmental Science and Technology 46:9964-71. Müller B, Steinsberger T, Schwefel R, Gächter R, Sturm M, Wüest A, 2019. Oxygen consumption in seasonally stratified lakes decreases only below a marginal phosphorus threshold. Scientific Reports 9:1-7. Niessen F, Wick L, Bonani G, Chondrogianni C, Siegenthaler C, 1992. Aquatic system response to climatic and human changes: Productivity, bottom water oxygen status, and sapropel formation in Lake Lugano over the last 10 000 years. Aquatic Sciences 54:257-76. Nürnberg GK, 1996. Trophic state of clear and colored, soft-and Non -co mmerc ial us e o nly Mesotrophy is not enough: Re-assessing phosphorus objectives for the restoration of a deep Alpine lake 115 hardwater lakes with special consideration of nutrients, anoxia, phytoplankton and fish. Lake and Reservoir Management 12:432-47. Rand MC, Greenberg AE, Taras MJ, 1975. Standard methods for the examination of water and wastewater (14th edition). American Public Health Association, American Water Works Association and Water Control Federation, New York, USA. 1193 pp. Rogora M, Buzzi F, Dresti C, Leoni B, Lepori F, Mosello R, Patelli M, Salmaso N, 2018. Climatic effects on vertical mixing and deep-water oxygen content in the subalpine lakes in Italy. Hydrobiologia 824:33-50. Salmi T, Maatta A, Anttila P, Airola TR, Amnell T, 2002. Detecting trends of annual values of atmospheric pollutants by the Mann-Kendal test and Sen’s slope estimates — the Excel template application MAKESENS. Available at: https://www.researchgate.net/publication/259356944_Detecti ng_Trends_of_Annual_Values_of_Atmospheric_Pollutants_ by_the_Mann-Kendall_Test_and_Sen’s_Solpe_Estimates_ the_Excel_Template_Application_MAKESENS Schindler DW, 1977. Evolution of phosphorus limitation in lakes. Science 195:260-2. Schindler DW, 2012. The dilemma of controlling cultural eutrophication of lakes. Proceedings of the Royal Society B: Biological Sciences 279:4322-33. Sen PK, 1968. Estimates of the regression coefficient based on Kendall’s tau. Journal of the American statistical association 63:1379-89. Smith VH. 1979. Nutrient dependence of primary productivity in lakes 1. Limnology and Oceanography. 24:1051-64. Smith VH, 2003. Eutrophication of freshwater and coastal marine ecosystems a global problem. Environmental Science and Pollution Research 10:126-39. Smith VH, Shapiro J, 1981. Chlorophyll-phosphorus relations in individual lakes. Their importance to lake restoration strategies. Environmental Science & Technology: 15:444-51. Tadonléké RD, Lazzarotto J., Anneville O, Druart JC, 2009. Phytoplankton productivity increased in Lake Geneva despite phosphorus loading reduction. Journal of Plankton Research 31:1179-94. Vollenweider RA, Kerekes J, 1982. Eutrophication of waters. Monitoring, assessment and control. Organization for Economic Co-Operation and Development (OECD), Paris, France. Available at: http://lakes.chebucto.org/TPMODELS/ OECD/OECD1982.pdf Wetzel RG, 1975. Limnology. Saunders, Philadelphia, USA. 743 pp. Non -co mmerc ial us e o nly