1 INTRODUCTION

While efforts to control eutrophication have been successful in several lakes (Jenny et al., 2020; Jochimsen et al., 2013; Sas, 1989), eutrophication remains a leading global water quality issue (Downing, 2014; Jenny et al., 2020; UNEP-IETC/ILEC, 2001). Wang et al. (2017) estimated that 63% of the 2,000 lakes they studied globally are presently eutrophic. Large lakes are particularly vulnerable through exposure to synergistic effects of various stressors including human population growth, intensifying agricultural production, industrialisation, climate change, and species invasions (Jenny et al., 2020). Accordingly, the intensity of summer near-surface phytoplankton blooms increased in 68% of 71 large lakes (surface area is >100 km²) globally between 1984 and 2012 (Ho et al., 2019). Exacerbated eutrophication often follows the classical trajectory of insufficient reduction in nutrient loads (Lürling & Mucci, 2020) and the rise in inorganic fertiliser application (Fink et al., 2018). Climate change not only amplifies eutrophication along the classical trajectory but also affects lakes through modifying internal processes, such as rearrangements in trophic structures and nutrient regeneration (Dokulil et al., 2010; Jeppesen et al., 2009).

Climate-related increases in nutrient availability within lakes varies along a spectrum from catchment-control dominated to internal-process dominated. On the catchment-controlled end, increasing frequency of intense precipitation and the correspondingly amplified flood dynamics generate large spikes in external nutrient loads in agricultural catchments and also where flood-buffering and nutrient-retaining wetlands have been drained (Havens et al., 2019; Schindler et al., 2012). On the internal-processes end of the spectrum, the warming climate modifies the thermal structure of lakes and decreases mixing by increasing the thermal stability of the water column (Dokulil et al., 2010; Kraemer et al., 2015; Livingstone, 2003). This may lead to rapid hypolimnetic (near-bottom) oxygen depletion and increased phosphorus (P) release from the sediments in both deep, stratified (Pettersson et al., 2003; Salmaso, 2005) and shallow, polymictic lakes (Wilhelm & Adrian, 2008). In the middle of the spectrum (dominated by catchment control at one end, and by internal processes at the other), nutrient availability may increase due to an interplay of catchment-based and in-lake processes (Michalak et al., 2013).

Cyanobacteria blooms are key symptoms of eutrophication. Global expansion of harmful cyanobacteria blooms is attributed to synergistic effects of elevated nutrient loads and climate warming. Raising water temperatures, heat waves, decreased mixing, reduced flushing, and increasing nutrient availability are the main factors promoting global increase in cyanobacteria (O’Neil et al., 2012; Paerl & Paul, 2012). Many of these factors also favour other summer algae. Accordingly, the frequency of large blooms of eukaryotic species is also increasing in lakes and reservoirs, albeit at a more limited spatial scale than cyanobacteria. An example is the increasing bloom frequency of Ceratium furcoides, an invasive dinoflagellate in South America (Cavalcante et al., 2016; Crossetti et al., 2019).

Large phytoplankton blooms often come as ecological surprises. Surprises are generated by the cumulative impacts of multiple disturbances within the characteristic recovery time of an ecosystem (Christensen et al., 2006; Paine et al., 1998). They are unexpected events that deviate from our past experience of the behaviour of complex adaptive systems because of the inherent unpredictability of non-linear dynamics. Abrupt ecosystem changes are predicted to occur more frequently in the changing climate and under the growing anthropogenic pressure (Christensen et al., 2006; Paine et al., 1998). Upon passing a critical threshold, climate change may cause an ecological regime shift defined as an abrupt, elementary, and persistent reconfiguration of the structure, functions, and feedbacks of lake ecosystems (Randsalu-Wendrup et al., 2016). These surprises and regime shifts present a challenge for resource management.

Recently renewed eutrophication makes continuously polymictic Lake Balaton a suitable case to examine how the synergistic impact of climate change and lake management may cause ecological surprises and regime shifts. Balaton is a large (surface area is 596 km²), shallow (mean depth is 3.7 m at +1.20 m water level [WL] above datum level), recreational lake in Hungary (Figure S1). It was one of the exceptional lakes in the global dataset of Ho et al. (2019), in which a sustained decrease in bloom intensity was observed between 1984 and 2012. Indeed, substantial reduction in nutrient loads from the 1980s resulted in recovery from historic eutrophication (Istvánovics et al., 2007). In late summer 2019, however, a record-setting mixed dinoflagellate-cyanobacterium bloom developed in the southwestern area of the lake. Peak concentration of chlorophyll a (Chl) exceeded 300 mg/m³, significantly higher than the maxima during historical eutrophic condition of approximately 200 mg/m³ (Herodek, 1986). After 25 years of good water quality, this record-setting bloom fell well and truly into the category of an ‘ecological surprise’, with potentially severe consequences for the regional economy that relies on tourism.

To explore the immediate as well as the long-term background of the record-setting bloom, we used high-frequency sensor data and long-term traditional data. We started by examining some common factors thought to promote proliferation of summer algae, particularly cyanobacteria globally. We found that these factors could not explain the record-setting bloom, which required an extraordinarily high internal P load to develop. Since anoxic P release rates are known to be much higher than oxic release rates (Lijklema, 1980), this led us to Hypothesis 1: a shift from the usual oxic to anoxic P release from the sediments occurred upon depletion of dissolved oxygen (DO) under insufficient vertical mixing. To test this hypothesis, we asked: relative to previous years throughout the past decade with much smaller (or even absent) summer blooms, was the DO status or any of its drivers extreme in 2019? Only subtle differences could be detected at the shallow near-shore site where long-term sensor data were available. To explore likely DO conditions in the open water, we set up a one-dimensional DO model driven by eddy diffusivity. Eddy diffusivity is calculated by a hydrodynamical model, forced by local high-frequency hydrometeorological data. To explain the unusual bloom composition, we hypothesised that the synergistic impact of a multidecadal increase in summer temperature (climate) and a series of management actions might trigger an ecological regime shift, including changes in phytoplankton structure (Hypothesis 2). Finally, we explore how renewed eutrophication could be managed.

中文翻译：

---

混合型巴拉顿湖（匈牙利）创纪录的藻类水华：气候变化和（错误）管理的协同影响

1 简介

虽然控制富营养化的努力在多个湖泊中取得了成功（Jenny 等人，  2020 年；Jochimsen 等人，  2013 年；Sas，  1989 年），但富营养化仍然是全球主要的水质问题（Downing，  2014 年；Jenny 等人，  2020 年） ；UNEP-IETC/ILEC，  2001 年）。王等人。（2017 年）估计，他们在全球研究的 2,000 个湖泊中有 63% 目前是富营养化的。大型湖泊特别容易受到各种压力因素的协同影响，包括人口增长、农业生产集约化、工业化、气候变化和物种入侵（Jenny et al.,  2020）。因此，1984 年至 2012 年间，全球 71 个大型湖泊（表面积大于 100 km ²）中有 68% 的夏季近地表浮游植物水华强度增加（Ho 等人，  2019 年）。富营养化加剧通常遵循营养负荷减少不足（Lürling 和 Mucci，  2020 年）和无机肥料施用量增加（Fink 等人，  2018 年）的经典轨迹。气候变化不仅沿经典轨迹放大富营养化，而且通过改变内部过程影响湖泊，例如营养结构的重新排列和养分再生（Dokulil 等，  2010；Jeppesen 等，  2009）。

湖泊内与气候相关的养分可用性增加的范围从流域控制为主到内部过程为主。在流域控制端，强降水频率的增加和相应放大的洪水动态在农业集水区以及洪水缓冲和养分保留湿地已经排干的地方产生了巨大的外部养分负荷峰值（Havens 等人，  2019 年；辛德勒等人，  2012 年）。在光谱的内部过程末端，气候变暖改变了湖泊的热结构，并通过增加水柱的热稳定性来减少混合（Dokulil 等人，  2010 年；Kraemer 等人，  2015 年；Livingstone，  2003 年））。这可能会导致深层、分层（Pettersson 等人，  2003 年；Salmaso，  2005 年）和浅层、多聚体湖泊（Wilhelm 和 Adrian，  2008 年）。在范围的中间（一端由流域控制控制，另一端由内部过程控制），由于基于流域的过程和湖内过程的相互作用，养分的有效性可能会增加（Michalak 等，  2013 年）。

蓝藻大量繁殖是富营养化的关键症状。有害蓝藻水华的全球扩张归因于营养负荷增加和气候变暖的协同效应。提高水温、热浪、减少混合、减少冲水和增加养分可用性是促进全球蓝藻增加的主要因素（O'Neil 等人，  2012 年；Paerl & Paul，  2012 年）。其中许多因素也有利于其他夏季藻类。因此，湖泊和水库中真核生物大量繁殖的频率也在增加，尽管其空间尺度比蓝藻更有限。一个例子是Ceratium furcoides的开花频率增加，这是一种在南美洲的侵入性甲藻（Cavalcante 等人， 2016 年；Crossetti 等人，  2019 年）。

大型浮游植物的大量繁殖通常会带来生态惊喜。惊喜是由生态系统特征恢复时间内多重干扰的累积影响产生的（Christensen 等人，  2006 年；Paine 等人，  1998 年）。由于非线性动力学固有的不可预测性，它们是偏离我们过去对复杂自适应系统行为的经验的意外事件。在气候变化和人为压力不断增加的情况下，预计生态系统的突然变化会更频繁地发生（Christensen 等人，  2006 年；Paine 等人，  1998 年））。超过临界阈值后，气候变化可能导致生态系统转变，定义为湖泊生态系统结构、功能和反馈的突然、基本和持续的重新配置（Randsalu-Wendrup 等，  2016）。这些意外和制度转变对资源管理提出了挑战。

最近重新出现的富营养化使得持续多变的巴拉顿湖成为研究气候变化和湖泊管理的协同影响如何导致生态意外和政权更迭的合适案例。Balaton 是匈牙利的一个大型（表面积为 596 km ²）、浅水（平均深度为 3.7 m，水位高于基准面 +1.20 m [WL]）、休闲湖（图 S1）。它是 Ho 等人的全球数据集中的特殊湖泊之一。（2019 年），其中观察到 1984 年至 2012 年间开花强度持续下降。事实上，自 1980 年代以来养分负荷的大幅减少导致历史上的富营养化恢复（Istvánovics 等人，  2007 年））。然而，在 2019 年夏末，湖西南部地区出现了创纪录的甲藻-蓝藻混合水华。叶绿素a (Chl) 的峰值浓度超过 300 mg/m ³，显着高于历史富营养化条件下约 200 mg/m ³的最大值（Herodek，  1986 年）。经过 25 年的良好水质后，这种创纪录的水华真正落入“生态惊喜”的范畴，对依赖旅游业的区域经济造成潜在的严重后果。

为了探索创纪录的大爆发的近期和长期背景，我们使用了高频传感器数据和长期传统数据。我们首先研究了一些被认为会促进夏季藻类，特别是全球蓝藻增殖的常见因素。我们发现这些因素无法解释创纪录的绽放，这需要极高的内部 P 负荷才能发展。由于已知缺氧 P 释放率远高于好氧释放率 (Lijklema,  1980)，这导致我们提出假设 1：在垂直混合不足的情况下，溶解氧 (DO) 耗尽时，沉积物中的磷释放从通常的好氧转变为缺氧。为了验证这一假设，我们问道：相对于过去十年中夏季开花少得多（甚至没有）的前几年，DO 状态或其任何驱动因素在 2019 年是否极端？在可以获得长期传感器数据的浅近岸地点，只能检测到细微的差异。为了探索开放水域中可能的 DO 条件，我们建立了一个由涡流扩散率驱动的一维 DO 模型。涡流扩散率由当地高频水文气象数据强制的水动力模型计算。为了解释不寻常的绽放成分，我们假设夏季温度（气候）的数十年增加和一系列管理行动的协同影响可能会引发生态状况的转变，包括浮游植物结构的变化（假设 2）。最后，我们探讨了如何管理新的富营养化。