Until about 20 years ago, budgetary constraints and a short-term approach to management led to an emphasis on research and monitoring projects of short duration, with results that were of immediate use in decision-making processes. In addition, given that in most academic institutions the professional advancement of researchers was (and is) dependent on their publication records, investigators tended to avoid long-term commitments to data collection that had no obvious goals in terms of testable hypotheses and subsequent papers to be published in primary journals (Wolfe et al. 1987). For these reasons, there was a shortage of monitoring and research programmes that entailed sampling from the same sites over long periods in accordance with comparable analytical methods. However, researchers have long been aware of the importance of long-term studies in ecology for the creation of testable hypotheses regarding environmental changes (Franklin 1989, Coull 1985, Wolfe et al. 1987) and comparison of current environmental quality with that of the past. In the last few decades, the importance of maintaining long-term data series and the challenges this involves have been addressed with the creation of Long-Term Ecosystem Research (LTER) sites and networks, now well-established at the international (LTER-International), regional (LTER- Europe) and national (e.g. LTER-Italy) levels (Mirtl et al. 2018). What makes LTER networks unique is that they make it possible to look at the bigger ecological picture, enhancing knowledge by enabling the innovative combination of different types of information and the integration of long-term data series with the large-scale analysis of patterns (Haase et al. 2016, 2018).

The Lagoon of Venice (LoV) is probably one of the most extensively studied ecosystems in the world (Tagliapietra et al. 2011). However, until recently, most of the research was concerned with specific processes or was conducted in response to environmental emergencies (Solidoro et al. 2004). Moreover, data were collected by separate local agencies in accordance with different methodological approaches, meaning that they were not always comparable.

Since 1997, CNR-ISMAR has been collecting long-term hydrological and biological data in the LoV (Bianchi et al. 2003, Bernardi Aubry et al. 2013). Moreover, since 2007 the LoV has been part of LTER-Italy, LTER-Europe and LTER-International (Bertoni 2012). In recent years, multiannual monitoring programmes have been implemented at regular intervals by the Italian Regional Environmental Protection Agencies (ARPAs) in fulfilment of the requirements of the European Water Framework Directive (WFD 2000/60/EC; Borja 2005, European Union 2000). The main concern of the WFD, which came into effect in 2000, is the assessment of the ecological status of water bodies (coastal waters, transitional waters, inland surface waters and ground waters) with a view to their protection, improvement and sustainable use.

Nutrient levels and phytoplankton biomass have also been included as water quality indicators for assessing the evolution of the trophic state of transitional waters, which is a key ecological issue and has profound implications for environmental management. From the 1950s to the 1980s, as a result of the growing use of fertilizers in the farmland surrounding the LoV and the increasingly industrial character of its drainage basin, nitrogen inputs to the lagoon doubled (from about 4500 t yr^–1 to 9000 t yr^–1; Solidoro et al. 2010) and phosphorus inputs increased nearly threefold (from about 500 t yr^–1 to 1300 t yr^–1; Solidoro et al. 2010). The construction of wastewater treatment plants and the total ban on the use of phosphorus in the manufacturing of detergents in 1989 resulted in phosphorus inputs falling to less than half the levels recorded in 1950, with further decreases over the last 20 years. In contrast, a fall in nitrogen loads has only been observed since the 1990s. This has been driven by a number of factors, including the introduction of nitrification/denitrification processes in the major treatment plants, the closure of fertilizer factories and the adoption of better fertilization practices in agriculture (Solidoro et al. 2010). Consequently, nitrogen levels have returned to 1950 levels. As for phytoplankton biomass, after peaking in the 1970s (190 μg L^–1 in 1978), chlorophyll a concentrations fell to <25 μg L^–1 in the 1980s. Despite a small increase from 2000 to 2003, even lower values were recorded in subsequent years (Solidoro et al. 2010, Bernardi Aubry et al. 2013).

The aim of this paper is to document the analysis of some of the vast amount of hydrological and biological data collected on transparency, temperature, salinity, dissolved oxygen, inorganic dissolved nutrients and chlorophyll a in the framework of LTER-Italy. Here, we present some findings emerging from the statistical analysis of a twenty-year (1998-2017) dataset based on the LoV, focusing on the evolution of trophic state and phytoplankton biomass. It is not the purpose of this article to deal with spatial and seasonal variability, already discussed in previous articles (Bianchi et al. 2003, Bernardi Aubry et al. 2013); our main aim is to identify significant trends in the LoV by considering multiannual series.

Data series based on samplings performed at regular intervals across large lagoon areas for prolonged periods in accordance with shared methods are a uniquely useful resource that can be used to both generate and test hypotheses regarding the impact of environmental changes.

Over the last 20 years, the temperature of the lagoon waters has increased (Table 2). In addition, it has been shown (Amos et al. 2017) that the waters of the LoV are warming at a faster rate than both the global average and the adjacent sea. Specifically, the increase during winter months appears to be directly linked to tourism (University of Southampton 2012) as a result of the ‘urban heat island effect’, a microclimatic phenomenon occurring in correspondence with urban centres, in which the increased number of tourists leads to more human activities, more anthropogenic heat emissions, causing local overheating (Zhang and Wu 2015).

With regard to salinity, identifying long-term changes in the LoV is difficult. Precipitation patterns and the (far from even) ebb and flow of the tides result in wide variability that is unrelated to changes in the salinity of the water itself. The increase in this parameter shown by our data is not significant (Tables 2 and 3), in accordance with the steady-state condition of the LoV detected by Zirino et al. (2014).

Increasing water transparency in the period 1998-2007 has been attributed to declining river discharge into the LoV (Bernardi Aubry et al. 2013). More recently (2011-2014), the reduction in clam farming and the resurgence of macroalgae and seagrasses have also been cited as causes (Sfriso et al. 2019). The overall increase in transparency during the period considered in this study is shown in Tables 2 and 3.

Mean relative oxygen increased significantly and since 2012 has never fallen below 110%, although these levels are still way below those observed in the late 1980s and early 1990s. Indeed, that period saw average saturation levels of up to 250% in the central (Sfriso et al. 2019) and northern lagoons (Bianchi et al. 2000) due to the massive development of macro- and microalgae.

The biggest changes in nutrients over the years have been found in ammonium, concentrations of which have decreased significantly since 2007. What we observed is in agreement with trends in the LoV reported for the first few years of the 21^st century by Harleman (2005), Solidoro et al. (2010) and Bernardi Aubry et al. (2013). Specifically, Sfriso and Buosi (2018) showed that ammonium levels in the waters around the industrial zone of Porto Marghera fell by two orders of magnitude over a fifty-year period (1962-1964, 3800 µM; 1970-1972, 1200 µM; 1976-1977, 200 µM; 2011, 10-20 µM). Nitrates also fell, although not to the same degree as ammonium. This fall can be attributed at least partly to the effects of Council Directive 91/676/EC, which provides the identification of “Nitrate Vulnerable Zones” in which the prohibition of spreading farm waste was introduced beyond a maximum limit of 170 kg nitrogen/hectare/year and established codes of good agricultural practice to be adopted by farmers. The directive was implemented by the subsequent Italian legislation through Legislative Decree no. 152 of 11 May 1999 and the Ministry Decree of 7 April 2006. The fall in ammonium and nitrate concentrations explained the significant decline in dissolved inorganic nitrogen levels detected by the present study (Table 2) and other authors (Bernardi Aubry et al. 2013, Sfriso and Buosi 2018). In the late 1980s, phosphorus concentrations fell sharply as a result of the ban on the use of phosphorus in detergents. Our data, which run from 1998, show that phosphates have continued to decline, albeit not significantly (Table 2). As in other lagoons, the N/P ratio has become much higher in the LoV than the canonical 16:1 (Pérez-Ruzafa et al. 2019): in the last three years of this study (2015-2017) it varied between 50 and 70, while in previous years (1998-2014) it was always above 100; its decrease is the result of the significant decrease in nitrogen inputs.

The decline of nutrients in lagoon waters does not appear to be linked to a decrease in rainfall and river discharge over the period studied (Fig. 3). In fact, the analysis of the rainfall time series recorded by 17 meteorological stations located in the drainage basin of the lagoon did not highlight any significant trend (Table 2).

In the study area, concentrations of chlorophyll a decreased significantly during the period 1998-2017, remaining below 10 µg L^–1 except in the year 2001. This decrease could be linked to both the decrease in nitrogen inputs and the competition for nutrients between phytoplankton and the other primary producers (macroalgae and seagrasses) that have recolonized the lagoon in recent years. This recolonization has been so significant that it can even be detected by satellite, as seen in the maps of macrophyte cover in the sampling area in spring 1998, 1999, 2016 and 2017 (Fig. 6). In the first two years (1998 and 1999) the coverage was very poor, while in the last two years the sampling stations and/or neighbouring areas show dense coverage. The marked presence of these primary producers may explain the increase in dissolved oxygen in recent years, together with the significant reduction in phytoplankton biomass and increased transparency.

The seasonal Kendall-τ test revealed significant temporal trends for some parameters (Table 2). Further statistical analysis performed using the regime shift method enabled us to observe that for temperature, transparency, dissolved oxygen and ammonium the change was abrupt, while the other parameters showing significant τ changed more gradually. For transparency and oxygen, the shift was influenced by the recolonization of the lagoon by macroalgae and seagrasses in recent years. The temperature shift after 2011 was corroborated by the average annual air temperature in the Veneto district, which in the period 2011-2017 was always above 11.5°C, the average for the last 25 years (Dell’Acqua 2017, Fig. 7). Both the reduced and the oxidized forms of nitrogen decreased in the period considered, though the decrease in ammonium was more abrupt than that in nitrates. Overall, the decrease in dissolved inorganic nitrogen was most affected by the gradual decrease in its oxidized form. All these shifts need to be verified in the long term in order to understand whether they are exceptional or reveal a new regime.

The last few decades have seen substantial changes in the LoV’s morphological and hydrodynamic characteristics, together with significant variations in the amounts of pollutants and nutrients entering its waters (Solidoro et al. 2010, Bernardi Aubry et al. 2013, Sfriso et al. 2019). In the 1970s and 80s, when nutrient loads were at their highest, there was huge growth of Ulvaceae, and phytoplankton blooms were frequent, especially in the central basin, accompanied by a decline in seagrasses. Macroalgae remained dominant until the early 1990s and then began to fall. Thus, by the time the data series referred to this study began, Ulvaceae biomass had already peaked. At first (1997-2000), phytoplankton were still abundant, microalgal blooms were frequent and the presence of macroalgae and seagrasses in the sampling areas was limited. In subsequent years, there was a decrease in phytoplankton abundance and bloom frequency while macrophyte biomass increased. The actions promoted by the laws to limit the impacts caused by the supply of eutrophying substances (e.g. the Ronchi-Costa Decree of 23 April 1998 setting limits on the discharge of total nitrogen and total phosphorus at 10 µg L^–1 and 1 µg L^–1, respectively, and the Ministry Decree of 2 February 1999 setting the maximum annual allowable nutrient loads at 3000 t for total nitrogen and 300 t for total phosphorus) caused nutrient levels in lagoon waters to decline and hypo- and hyperoxia to become less frequent, although there was a brief and limited hypoxia event in July 2013 that caused fish mortality (Bastianini et al. 2013).

Our data thus highlight an improvement in the ecological condition of the LoV, as also evidenced by the monitoring of biological elements in accordance with the Water Framework Directive 2000/60/EC (Regione Veneto et al. 2015). It is important to monitor the increasing temperature of the LoV’s water in response to climate change, since this is likely to impact dissolved oxygen concentrations, lagoon organism physiology and the timing of lagoon processes (Pérez-Ruzafa et al. 2019). Furthermore, the construction and functioning of MOSE, the system of gates designed to close the seaward inlets during exceptionally high tides in order to protect Venice from flooding, is expected to modify the exchanges and fluxes of waters, nutrients and suspended matter between the lagoon and the sea, consequently affecting primary producers and upper trophic levels. Maintenance of this temporal data series, combined with the use of satellite maps, will be a useful tool for evaluating future changes to the LoV ecosystem.