Wind drives fast changes of light climate in a large, shallow re-established lake

With ever greater frequency, wetlands and shallow lakes that had been diverted for agriculture are being re- established to reduce nutrient loss and greenhouse gas emission, aswellas to increase biodiversity. Here, we investigate drivers of water column light attenuation (K d ) at multiple time scales and locations in Lake Fil, Denmark, during the ﬁ rst ﬁ ve years after its re-establishment in 2012. We found that K d was generally high (overall mean: 3.4 m − 1 ), with resuspended sediment particles and colored dissolved organic matter being the main contributors. Using daily time series of light attenuation recorded at four stations, we used a generalized additive model to analyze the in ﬂ uence of wind speed and direction on K d . This model explained a high proportion of the variation (R 2 = 0.62, RMSE = 0.74 m − 1 , and MAE = 0.55 m − 1 ) and showed that higher wind speed increased K d on the same day and, with smaller in ﬂ uence, on the next day. Furthermore, we found a signi ﬁ cant in ﬂ uence of wind direction and an interaction between wind speed and wind direction, a combinationthat suggests thatshort-term variationsinlightclimatedepends ontheinterplaybetween wind di- rection and sources of particles. Wind from non-prevailing directions thus in ﬂ uence K d more, as it can activate previously deposited particles. The maximum colonization depths of submerged vegetation occurred at ~2 – 6% of sub-surface light from 2014 to 2016 and peaked at 1.2 m in 2016. The fast, day-to-day variation of K d in Lake Fil reveals the importance of wind on light climate and in turn biological elements such as phytoplankton and submerged macrophyte development in shallow lakes. The implications are essential for the prior planning and management of future lake re-establishment.


Introduction
During the past 200 years, more than half of Europe's wetlands have been drained and claimed for agriculture (Moreno-Mateos et al., 2012;Zedler and Kercher, 2005). Intensified drainage and stream channelization have resulted in the loss of wetlands in the intensively cultivated European lowland regions (Biggs et al., 2017;Wine and Laronne, 2020). Large shallow lakes have also been drained; in Denmark alone, more than 200 large lakes (~5 lakes per 1000 km 2 land area), including the country's second and third largest lakes, disappeared between 1780 and 1980 (Hansen, 2014). This trend has changed during recent decades with large investments in both the re-establishment of lakes lost due to anthropological activities and the creation of entirely new lakes.
The long-term loss of wetlands, cultural eutrophication of lakes, and decline of biodiversity have stimulated interest in re-establishing wetlands and lakes and improving their ecological quality. More precise motivations include reducing nutrient export to coastal waters (Audet et al., 2020), improving carbon sequestration (Badiou et al., 2011), enhancing aquatic biodiversity (Mitsch et al., 1998), and creating recreational areas for the public (Hoffmann and Baattrup-Pedersen, 2007). For these reasons, more than 50 larger lakes (>10 ha) have been reestablished in Denmark during the past two decades.
Unfortunately, the early development in many re-established lakes is not monitored systematically. Lakes that are re-established on the large phosphorus pools of formerly cultivated soils may face high internal nutrient loading initially accompanied by the release of colored dissolved organic matter (CDOM) that absorbs light in the water column (Kinsman-Costello et al., 2014;Pant and Reddy, 2003;Steinman and Ogdahl, 2011). Often, this results in turbid water columns dominated by phytoplankton that absorbs most of the incoming photosynthetically active radiation (PAR) before it reaches the sediment surface, thus preventing the development of benthic microalgae and submerged macrophytes. In addition to light absorption by photosynthetic pigments and CDOM, scattering of light by suspended organic and inorganic particles are the main determinants of the water column light climate and, summed, yield the diffuse vertical light attenuation coefficient for PAR (K d , m −1 ; Kirk (1994)). K d is regulated through a complex interplay among biological, physical, and chemical processes occurring both beyond and within the lake. In re-established lakes, much initial CDOM derives from the leaching of flooded terrestrial soils (Kragh et al., 2017). Later, CDOM is mostly derived from terrestrial production in the catchment, which is generally a major cause of high K d in north-temperate lakes (Karlsson et al., 2009). CDOM is slowly degraded in lakes as a result of photochemical degradation and bacterial mineralization (Madsen-Østerbye et al., 2018). Phytoplankton development is dependent on a range of interacting variables that influence the biomass at a given point in time. The access to major nutrients such as nitrogen and phosphorus, light availability, and water temperature are of particular importance (Wetzel, 2001). Phytoplankton plays two roles with regards to K d , as the development directly influences the light climate and is itself regulated by the presence of other light attenuating components (Hellström, 1991). Scattering and absorption of light by both phytoplankton and sediment particles suspended in the water column may constitute a large proportion of K d (Balogh et al., 2009). Resuspension of particles may result from the activities of bottomfeeding fish (Breukelaar et al., 1994) and from near-bottom shear stress that results from wind shear (Reardon et al., 2014). We are especially interested in the latter impact.
In shallow lakes, the contribution of suspended particles in the water column to K d may be particularly high (Kristensen et al., 1992;Luettich et al., 1990). This is promoted by high wind speeds, which increase near-bottom shear stress and thus resuspension of sediment particles (Reardon et al., 2014); higher speeds can mobilize heavier particles and deeper sediment layers (Bailey and Hamilton, 1997). The efficiency with which surface wind shear is transformed to near-bottom shear stress depends on wind speed, water depth, and wind direction (Evans, 1994) and the efficiency with which near-bottom shear stress resuspends particles depends on the sheltering of the lakebed. Islands, embayments, or terrestrial vegetation may create sheltered areas that promote particle deposition and hinder resuspension. In short, wind influences horizontal sediment redistribution, and suspension and settling dynamics are modulated by lake bathymetry and morphometry (Håkanson, 1977;Whitmore et al., 1996).
Models based on high temporal resolution data may provide insights not only into the regulation of light climate at different time-scales, but also into the distribution and species richness of submerged macrophytes in turbid lakes. Areas with high macrophyte cover may play a key role in ecosystem functioning by serving as refuges for zooplankton, reducing sediment nutrient release, enhancing particle sedimentation, and competing with phytoplankton for nutrients and, thereby, increasing water transparency (Hilt et al., 2017;Sand-Jensen and Borum, 1991;Scheffer et al., 1993). Moreover, the colonization and development of submerged macrophytes are strongly dependent on K d , thus creating a positive feedback loop (Hamilton and Mitchell, 1996;Pérez et al., 2013). Improving our understanding of the temporal development and drivers of K d is especially important in re-established lakes where early management actions could improve future ecological quality.
In the five years following the re-establishment of Lake Fil, highfrequency measurements of K d collected at four stations enabled us to investigate the spatio-temporal variability of K d . Due to this lake's shallowness and wind-exposed location, we expected wind to be an important driver of K d . Specifically, our goals were to: 1) determine the inter-and intra-annual variation in K d , 2) quantify the influence of wind on K d , and 3) evaluate the influence of K d on the development of submerged macrophytes.

Study site
The study was conducted in Lake Fil, Western Jutland, Denmark (55.70°N, 8.235°E), three km from the North Sea. Lake Fil had been the third-largest lake in Denmark (2185 ha), but it was drained on several occasions starting in 1852. By 1952 it had been totally claimed for agriculture and was intensively cultivated until 2012. Large amounts of inorganic fertilizers and manure were applied, resulting in phosphorus-rich soils.
In 2012, Lake Fil was re-established at approximately one-third of its original size (889 ha) divided into southern and northern basins of equal size. The catchment area is 104 km 2, with the inlet and outlet located in the southern and northern basins, respectively (Fig. 1). The two basins are separated by a narrow dam and interconnected by a canal through the dam. The lake is shallow, with 1 m in mean depth and 3.5 m in maximum depth. Water retention time averages 70 days (range: 20-180 days). Prior to flooding in autumn 2012, the lakebed was prepared by leaving grain stubble on the fields. These root structures were intended to stabilize the sediment and reduce particle resuspension in and phosphorus release into the water immediately after flooding.

Data collection
Field measurements were initiated during summer 2013 (seven months after re-establishment). At four open water sites (Fig. 1, St. 1-St. 4), measurements of incoming illuminance (lux) were collected throughout the year, except in cold winter months when the lake surface was frozen. Each station was equipped with two to five sensors (HOBO UA-002-64, Onset Computers) displaced vertically 50-100 cm apart through the water column. Daily water level (2 mm precision) was determined by analyzing differences in pressure recorded by two data loggers, one submerged and the other in the air (HOBO U 20-001-04, Onset Computers). High-frequency (10 min) light measurements were used to determine daily values of K d . For each day and station, incoming illuminance was summed and converted from lux to incoming irradiance (I, PAR 400-700 nm) using a linear model without an intercept (I = lux⋅0.00975, n = 699, p < 0.001, R 2 = 0.996, Fig. S1) fitted from concurrent lux and I measurements on land (HOBO S-LIA-M003, Onset Computers). When daily incoming irradiance (I z ) was available at three or more depths (z), K d was calculated as the slope of linear model fits (Eq. (1)) and from Eq. (2) when available at only two depths (I 1 at z 1 and I 2 at z 2 ; Kirk (1994)).
Some low-quality readings were observed, mainly due to excessive fouling of some sensors during certain periods. We used four criteria to identify and then remove low-quality observations: total daily illuminance <500 lx, <10% difference in incoming illuminance between top and bottom sensors (when applying Eq. (2)), observations with R 2 values <0.5 (when applying Eq. (1)), and K d values <0 m −1 . In all, 10% of days with measurements were discarded based on the four criteria or only having readings from one sensor.
High-frequency (10 min) meteorological data measured at 2 m above the ground were obtained from a nearby (18 km) meteorological station operated by the Danish Meteorological Institute (DMI, 2021) to calculate daily wind speed and direction averages.
Phytoplankton chlorophyll a fluorescence was measured continuously in the southern basin ("Chl. a" site in Fig. 1) using a CYCLOPS-7 fluorescence probe (Turner Design, CA, USA) placed 30 cm below the water surface. The fluorescence signal was calibrated against water samples analyzed for chlorophyll a by extraction in ethanol and measurement on a Shimadzu UV-1800 spectrophotometer (Jespersen and Christoffersen, 1987).
The maximum colonization depths of submerged macrophytes were determined from visual inspection of side-scan imagery collected during July and August in all years. We collected side-scan imagery by traversing the lake in parallel paths with up to 50 m distance intervals with a boat using a Lowrance HDS-12 Gen3 sonar (Navico, Egersund, Norway) equipped with a Lowrance Hybrid Dual Imaging Skimmer Transducer operating at 200 KHz and a StructureScan HD transducer. The collected sidescan imagery was assembled in Reefmaster 2.0 (ReefMaster Software Ltd., United Kingdom). Using coordinates of the observed submerged macrophyte patches, the maximum colonization depths were calculated from the average water level from April till August every year.
Water samples were collected every other week from the canal connecting the two basins ( Fig. 1). Nutrient concentrations were  tion, 1971). For light absorbance measurements, water was filtered through pre-combusted GF/F filters (0.7 μm), and absorbance at 440 nm (A 440 ) was measured in 1 cm cuvettes (L = 0.01 m) on a spectrophotometer (UV-1800 Shimadzu) as a proxy of CDOM. The light attenuation coefficient (a 440 , m −1 ) at 440 nm was determined (Eq. 3) after subtracting the absorbance at 750 nm (A 750 ).
In examining the relative contribution of the optical components to K d , we used empirical estimates of specific light absorption for CDOM, chlorophyll a, and water itself. For CDOM, we used the value from Balogh et al. (2009), which is based on measurements in shallow lakes; for chlorophyll a, we used the mean value of a literature survey (Krause-Jensen and Sand-Jensen, 1998), and for water, we used the value reported by Smith and Baker (1978). After subtraction of the attenuation due to CDOM (K CDOM ), chlorophyll a (K Chl. a ), and water (K water ) components from K d , the remainder is the estimated attenuation due to organic and inorganic particles (K particles ) in the water column (Eq. (4)). We determined the contribution of each component to K d only for the southern basin, where chlorophyll a was measured.

Statistical analysis
We used generalized additive models (GAM) to quantify the influence of wind speed and direction on K d . GAM enable different types of smoothing functions to be fitted to predictor variables and include correlation structures accounting for the residuals' temporal dependence, often prominent in time series data (Pedersen et al., 2019). Results from initial experiments using linear mixed models violated statistical assumptions and were deemed unsatisfactory. Therefore, we used a GAM approach that has proven suitable for modeling time series data by means of different residual error structures and the inclusion of nonlinear relationships (Simpson, 2018;Yang and Moyer, 2020). The use of GAM and smoothing functions can also be justified from the expected relationship between K d and wind speed, i.e., variation in K d is expected to follow, but not instantaneously, sustained changes in wind speed. We used data collected from calendar-days 80-295, the main growing season (late March-late October; Middelboe and Markager, 1997; van Dijk et al., 1992), where the overlap of data across years was largest, resulting in 3037 observations over the five years. The initial model fits included: station, wind speed on the same day (t 0 ) and 1-3 days before (t −1 , t −2 and t −3 ), wind direction, the interaction between wind speed at t 0 and wind direction, and inter-and intra-annual time trends. From initial GAM fits, we assessed the significance of smoothing and included non-significant smooth terms as linear predictors. 'Station' was included as a random effect, wind speed at t 0 was used as thin plate regression splines, wind direction as cyclic cubic regression splines, and their interaction as a tensor interaction smooth. Seasonal and inter-annual effects were included using a day of the year and year smooth terms, respectively, and the inclusion of a tensor interaction smooth between the two was also tested. Initial GAM fits revealed high residual autocorrelation. Therefore, we compared the performance of different correlation structures using the Akaike information criterion (AIC; Burnham and Anderson (2002)). The continuous-time covariateorder-one autocorrelation structure was found to provide the best fit. This was followed by using AIC to compare multiple candidate models containing different combinations of predictors. Model diagnostics were performed using autocorrelation and partial autocorrelation functions, along with residual plots. GAM were fitted using the mgcv Rpackage (Wood, 2017). All statistical analyses were performed using the R programming language (R Core Team, 2021), and all scripts for the analysis are available from an online repository (https://github. com/KennethTM/FilsoLight).

Spatio-temporal variability of light attenuation
Water column K d exhibited a large variation (range: 0.2-8.9 m −1 ) and averaged 3.4 m −1 over the five years (Fig. 2). The mean water depth where only 10% of incoming irradiance remained (z 10% ) was 0.8 m. Sudden increases in K d generally occurred simultaneously at all (3.2 m −1 ), but was consistently higher at St. 4 (3.9 m −1 ), thus increasing with distance from the inlet.

Variables influencing light attenuation
During the first summer in Lake Fil, the water was much clearer at the inlet than at the outlet: mean a 440 increased rapidly from the inlet (1.7 m −1 ) to the dam-connection between the two basins (3.5 m −1 ) and on to the outlet (6.3 m −1 ). This was likely due to the accumulated release of CDOM from the soil and crop stubble that comprised the newly flooded lakebed (Fig. 3). The difference between inlet and outlet a 440 decreased over the following years. Seasonal variations in a 440 were not pronounced, though slightly lower values were observed during summer (mean: 3.1 m −1 ) compared to autumn, winter, and spring (mean: 3.5 m −1 ). Pronounced increases in CDOM at the inlet could occur in late summer and autumn following precipitation events from the catchment.
Winds measured at a nearby meteorological station were relatively strong, with mean daily wind speeds of 7.2 m s −1 (range: 1.3-20.7 m s −1 ). Wind speeds were highest during winter and lowest during summer. Westerly winds were prevailing, and the wind was rarely from the north (Fig. 4C-D). The morphometry of Lake Fil makes it very exposed to wind from the most prevailing wind directions.

Modeling light attenuation
To select a generalized additive model for further analysis, we evaluated a range of candidate models with different combinations of predictor variables. The top three models performed very similarly (within 2.60 AIC units), and they differed only in the number of lagged wind speed terms (Table 1). We selected the rank one model for further analysis. This model contained the fewest terms of the three best ranking models, namely two wind speed terms (at the day t 0 and the day t −1 before light measurement), wind direction and, a wind direction-wind speed interaction term, and inter-and intra-annual temporal terms. This GAM accounted for variations in K d with an R 2 of 0.62, rootmean-squared-error (RMSE) of 0.74 m −1 , mean-absolute-error (MAE) of 0.55 m −1 . The predicted vs. the observed values were spread evenly around the 1:1 line (Fig. 5).
Based on this GAM, it was possible to quantify the influence of wind speed and direction on K d . Wind speed at t 0 had the strongest influence on K d , which increased rapidly as wind speeds exceeded 5 m s −1 (Fig. 6A). K d also depended on wind speed on the previous day, t −1 (Fig. 6B). However, the effect of wind speed was modulated by wind direction: K d was larger when winds came from northerly directions, while the prevailing westerly winds were less influential (Fig. 6C-D). This pattern reflects the interplay between source and process limitation, i.e., the availability of sediment particles that are susceptible to resuspension at a given wind speed. For example, the difference in predicted K d at a wind speed of 5 m s −1 from the prevailing (median wind direction of 227 degrees) and north (0 degrees) is only 0.04 m −1 but this difference increases drastically to 0.5 and 1.3 m −1 at wind speeds of 10 and 15 m s −1 . The influence of intra-annual variation (Fig. 6E) is strongest during summer and is likely due to the combined influences of phytoplankton and reduced lake water volume. The inter-annual trend (Fig. 6F) shows a decrease in K d during the first three years, followed by increases in 2016 and 2017.

Partitioning of light attenuation components
Using the measured a 440 and chlorophyll a concentration in the southern basin, we determined the relative contribution of each component to K d . K d was generally dominated by the contribution to light attenuation from particles (mean: 52.2%) and CDOM (mean: 39.8%) with a minor periodic contribution from chlorophyll a (mean: 6.8%) and water (mean: 1%; Fig. S3). Most of the time, phytoplankton chlorophyll a was not a major contributor to K d , but during short-lived blooms generally lasting a few weeks, chlorophyll a reached about 35 μg L −1 corresponding to an attenuation of~0.52 m −1 and increased the contribution accordingly. The particles that dominate K d are most likely suspended in the water column as a consequence of windinduced resuspension.

Distribution of submerged macrophytes
Submerged macrophytes rapidly colonized Lake Fil, extending to 0.5 m the first summer and steadily increasing during the following three years, reaching 1.2 m in 2016. Their maximum colonization depth generally followed that of light availability (Fig. 7). The water depth with 10% of incoming irradiance (z 10% = 2.3·K d −1 ) increased during the lake's first four years and dropped in 2017, reducing the depth distribution of submerged macrophytes. During 2015-2016, the submerged macrophytes reached a depth where approximately 2-6% of incoming radiation was present. The observed maximum colonization depth and z 10% trends were consistent with the inter-annual trend term estimated with the GAM (Fig. 6 F), with increasing K d from 2016 to 2017.

The light climate in re-established lakes
Re-established lake ecosystems are often highly dynamic, and their development is difficult to predict. Just three years after reestablishment, Lake Fil supported a very high species richness, including 40 different species of aquatic macrophytes (Baastrup-Spohr et al., 2016). This high richness was maintained seven years after reestablishment. As re-established lakes age beyond 20 years, more scattered observations suggest that macrophyte richness may decrease (Sø et al., 2020). This apparent trend underscores that monitoring the early development of new lakes is essential in order to understand the ecology and biodiversity and improve the design and early management depending on the aims of the lake restoration (Audet et al., 2020).
Prior to the re-establishment of Lake Fil, we expected that high nutrient levels would enable phytoplankton to dominate in the new lake. We were mistaken. As a consequence of the lake's wind exposure, low water depths, and short water retention times, large quantities of nutrients and organic matter were rapidly flushed out of the system (Kragh et al., 2017). Rapid nutrient loss has also been observed in other restored wetlands (Kinsman-Costello et al., 2014;Steinman and Ogdahl, 2011). In the case of Lake Fil, nutrient loss was further promoted by windinduced wave action and physical disturbance of surface sediment, resulting in resuspension of sediment particles and, thereby, further nutrient release (Hamilton and Mitchell, 1996;Søndergaard et al., 2003). In combination with CDOM, these resuspended particles are the main contributors to K d in Lake Fil.

Contributions of CDOM and chlorophyll a to K d
The highest CDOM absorption coefficients in both basins were observed in Lake Fil's first summer (2013) with decreasing values in 2014-2017. The higher a 440 values were likely a result of lakebed organic carbon and grain-stubble leaching humic compounds to the water after flooding (Zak et al., 2018). This dynamic countered the effects of grain stubble root structures that, as intended, stabilized lakebed sediment and initially resulted in lower wind-induced sediment Table 1 Model selection table for the top-four candidate GAM models. Only some terms are included in the table as the smoothed terms, time (inter-and intra-annual variation), wind speed and wind direction are included in all the best candidate models. For each model df (number of parameters), AIC, ΔAIC, and AIC, weights are reported. For parametric terms (wind speed lag t −1 to t −3 ), the effect size is also included. resuspension (Kragh et al., 2017). The gradual release of sediment CDOM continued throughout the first five years, with CDOM generally increasing from lake inlet to outlet, suggesting that in-lake sources are still contributing to this organic matter pool. However, the largest source of seasonal variation in CDOM appears to be pulse events at the inlet determined by precipitation patterns and catchment processes (Kragh et al., 2020). The CDOM contribution to K d is variable, but generally changes on longer time-scales (seasons) than that of chlorophyll a or suspended particles.
Although chlorophyll a is expected to be high when phosphate and nitrogen are as concentrated as they are at Lake Fil (Schallenberg and Burns, 2004), the low observed chlorophyll a levels are likely due to light limitation caused by high CDOM and particle concentrations in the water column (Jones et al., 1996;Karlsson et al., 2009). Time periods (at the scale of weeks) with low wind speeds and, consequently, lower light attenuation by resuspended particles, are likely required for larger phytoplankton blooms to develop.

The influence of wind induced resuspension
The selected GAM shows that wind speed was an important driver of short-term (day-to-day) variation in K d and is likely an important process in other wind-exposed shallow lakes. The model showed that higher wind speed resulted in higher K d on both that day and, to a lesser degree, the following day. This is expected, as increasing wind speeds result in higher near-bottom shear stress, which enables particles of increased size and deeper position in sediments to be suspended in the water column (Valipour et al., 2017). Observations in other systems confirm that wind is the major contributor to near-bottom shear stress during resuspension events (Reardon et al., 2014). Also, high wind events have carry-over effects that continue to influence K d for hours or even days after the event. The abrupt increase in K d during windy periods was evident at all stations and, because meteorological data can be used to forecast wind speed, it is also possible to predict K d (Hamilton and Mitchell, 1996).
Importantly, the model also shows that wind direction modulates the influence of wind speed on K d . This phenomenon has also been observed in other shallow lakes (Bengtsson et al., 1990;Lövstedt and Bengtsson, 2008). It is likely a consequence of differences in sediment distribution and composition, which determine the quantity of resuspendable particles at a given wind speed. Sediment composition is influenced by the prevailing wind directions, lake morphometry, and bathymetry (Whitmore et al., 1996). Wind direction from nonprevailing directions may thus resuspend recently deposited particles or erode the softer sediments more easily. In the case of the shallow Lake Fil, a light fraction of the sediment particles is likely involved, repeatedly, in the majority of resuspension events (Luettich et al., 1990). All of these conditions can be ameliorated during the design of to-bere-established lakes. By considering historical meteorological data during the planning phase, export or settlement of organic particles can be improved by choices taken regarding lake morphometry and bathymetry. Such actions could also include establishing islands or pits that enable particle settlement and thus facilitate more transparent water (Sø et al., 2021).

Influence of light climate on submerged macrophyte distribution
Light availability is essential for the colonization and distribution of submerged macrophytes. Furthermore, the development of submerged macrophytes provides important feedback effects on K d and ecosystem functioning (Jeppesen, 1998;Sand-Jensen and Borum, 1991). This presents an interesting interplay between wind and macrophyte development. When light and nutrient requirements are met, wind-induced resuspension may promote development by reducing organic matter and improving texture in sediment (Barko and Smart, 1986). Once established, submerged macrophyte stands decrease sediment resuspension by reducing flow velocities and stabilizing sediments (Sand-Jensen, 1998). On the other hand, high mechanical stress in wind-exposed lakes may restrict the development of submerged macrophyte stands. Most important, however, is that temporal changes in light availability enable macrophyte colonization and growth. Rooney and Kalff (2000) show that biomass distribution is affected by inter-annual variations. Good light conditions early in the growing season resulted in increasingly deep macrophyte colonization. While short-term changes in K d are driven directly by changes in wind speed and direction, high wind speeds during late winter and early spring may also hamper early colonization and development of submerged macrophytes in Lake Fil. This may be reinforced when high CDOM levels appear in spring, as they did in 2017 in Lake Fil. Nonetheless, the reestablished Lake Fil supports one of the highest species richness of aquatic plants in Denmark (Baastrup-Spohr et al., 2016). This is likely due to the lake's location in a species-rich region and the existence of a dynamic coastal zone, where submerged macrophytes can colonize in extensive shallow waters that are kept free of tall emergent swamp and reed vegetation by extensive water level fluctuations (50-100 cm annually), wave exposure, and cattle grazing (Baastrup-Spohr et al., 2016).

Conclusions
The intense monitoring in the re-established Lake Fil highlights the importance of wind-induced resuspension on day-to-day variations in light climate in shallow lakes. Light attenuation in Lake Fil was sufficient to suppress the development of phytoplankton, despite high nutrient levels, and the main contributors to light attenuation were resuspended particles and CDOM. High wind speeds increased light attenuation, though this effect was modulated by wind direction, as wind shifts and wind from the non-prevailing directions increased light attenuation more than wind from the prevailing directions. This indicates that a small proportion of the lake's particles is regularly redistributed by wind, which can result in fast changes in light climate. During the first five years, K d decreased and submerged macrophytes grew increasingly deep. Submerged macrophytes did not respond instantaneously to changes in the light climate due to the time delay for growth. This was evident in the first summer following re-establishment (2013) where macrophytes grew shallower than expected from the light climate, and in 2017 where growth was observed deeper than what the prevailing light climate predicted.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.