Introduction

Central European fishponds represent unique, human-made freshwater ecosystems that are important and integral parts of the landscape [1,2,3]. Fishponds used mostly for fish production are artificial equivalents of natural polymictic shallow lakes and represent the predominant form of standing water bodies in Central and Eastern Europe (from < 1 ha to several hundred ha, depth 1–4 m) [2].

Production intensification of the whole agricultural sector in the last century led to both enhanced fish production and significant environmental degradation of fishpond ecosystems [1, 4, 5]. Such environmental degradation then has been manifested in impaired fish production, frequent cyanobacterial blooms, hypoxic events (mostly in deep water layers) with periodic fish kills alongside direct effects on eco-services, such as degraded recreational uses or biodiversity support [2, 5,6,7]. Long-term excessive nutrient loads, both directly from aquaculture management and indirectly from municipal and agricultural runoff and atmospheric nitrogen (N) deposition [8, 9], reinforced recently by climate change [10], have modified biological processes such as primary production, nutrient utilization, respiration, or biogeochemical cycles [7, 11, 12]. Consequently, the current fishponds have become less predictable regarding production processes and thus also less cost-efficient, more stochastic, and greatly unstable systems like the most eutrophic shallow lakes [13, 14].

The above mentioned 'whole-fishpond fertilizing experiment' (held in the twentieth century in Czechia [4]) included the application of both nitrogen (N) and phosphorus (P) in the form of either organic or inorganic fertilizers and revealed P as the nutrient that limited fish production [15, 16]. The positive effects of P addition were reflected in fish production increment and naturally drew attention away from N availability (in determining phytoplankton biomass and community composition) because, at that time in mesotrophic fishponds, P certainly was the limiting nutrient, unlike in hypereutrophic fishponds at present [16]. While light usually limits primary producers in turbid freshwaters, denitrification may cause a significant N loss, driving phytoplankton from P to N limitation in shallow eutrophic lakes and fishponds [17,18,19]. The shift from P to N limitation might be reinforced by internal P loading from anoxic sediments [12] and by P runoff from municipal sewages that often miss an efficient P removal [8]. Moreover, conventional aquaculture management is based on supplementary feeding with cumulative doses of ~ 1 ton of cereals fed per hectare per year and, in some cases, with additional fertilizing with the amount of 1–2 tonnes per hectare per year of organic manure [16, 20]. Phytate-bound P in cereals is not digested fully by the stocked common carp (Cyprinus carpio) [21] and, therefore, further increases internal P loading, while the decomposition of organic matter (manure) supports oxygen deficiencies and possibly denitrification, thus, ultimately N losses [16].

We hypothesized that seasonal variations in nutrient (N and P) dynamics in fishpond water would lead to measurable shifts in the stoichiometric balance from P to N limitation. Specifically, we expected to observe a transition from P limitation at the beginning (April–May) to N limitation at the end of the growing season (August–September) in fishponds across different altitudinal ranges. We further assumed that chlorophyll-a concentrations would exhibit regional and seasonal variability, with higher concentrations correlating with lower dissolved inorganic nitrogen to total phosphorus (DIN:TP) ratios. Additionally, we predicted that fishponds with larger surface areas would support higher chlorophyll-a concentrations due to increased nutrient inputs, particularly in lowland regions.

Material and methods

Data and their processing

We compiled data from 32 fishponds from lowland to highland regions of Czechia (altitudinal range from 170 to 730 m above sea level, a.s.l.; Fig. 1). The compiled water quality datasets came from Czech state government offices and research institutions and were collected for water quality monitoring and the assessment of ecological status. Water quality data were taken bi-weekly or monthly over a period of growing season from April to September (Apr–Sep). A subset of 5 fishponds was sampled each year over a long period from 2007–2019; otherwise, sampling was held for one to three growing seasons (Table S1). The collection of the water quality data followed the same methodology regardless of the fishpond entity and sampling year. The mixed water sample was collected from the epilimnion (approx. 15 cm below the water surface). Water transparency (Secchi depth, cm), dissolved oxygen concentration (DO, mg L−1), water temperature (°C), pH, and conductivity (µS cm−1) were measured in-situ. Chlorophyll-a concentration (µg L−1) as a proxy of phytoplankton biomass, alkalinity (mmol L−1), concentrations of total phosphorus (TP, mg L−1), orthophosphate (PO4-P, mg L−1), total nitrogen (TN, mg L−1), nitrate (NO3-N, mg L−1), and ammonium (NH4-N, mg L−1) were analyzed in accredited laboratories by the standard set of established methodologies and protocols. Detailed methods of the analyses were explained in previous studies [22,23,24]. Dissolved inorganic N (DIN) was calculated by summing the measured values as follows: DIN = NO3-N + NH4-N; nitrite was not taken into consideration due to its instability in water (conversion to nitrate) and concentrations close to detection limits. Data spanning the growing seasons were taken as observation units, herein referred to as pond-year, resulting in 150 pond-year cases (Table S1).

Fig. 1
figure 1

The location of fishponds under study. For a complete list of the fishponds under study, see Table S1 in Supplementary materials

Full size image

All monitored fishponds were polymictic, eutrophic to hypereutrophic waterbodies (area 0.23–520 ha; depth 0.5–5 m) with managed fish stock and limited littoral zone with macrophytes. The fishponds were used for rearing 1-year-old common carp (20–35 g body weight) to market size (ca. 2 kg) in a one- or two-year production cycle. For more detail on carp rearing in fishponds, see [5, 25]. The initial biomass of common carp was approximately 50–450 kg ha−1. In addition, predatory and supplemental fish such as pikeperch (Sander lucioperca), catfish (Silurus glanis), pike (Esox lucius), roach (Rutilus rutilus), rudd (Scardinius erythrophthalmus), grass carp (Ctenopharyngodon idella), and silver carp (Hypophthalmichthys molitrix) were stocked, but in comparably lower initial biomass accounting 10–150 kg ha−1. The undesirable fish, largely topmouth gudgeon (Pseudorabora parva) and Prussian carp (Carassius gibelio) were present in densities around 10% of total fish biomass. The supplemental feed for the improvement of carp growth performance was provided during the growing season: estimated doses of feed used in April, May, June, July, August, and September, respectively, comprised 5, 10, 20, 30, 20, and 15% of total feed fed (farmers' records). The feed supplied in fishponds were usually cereals, often locally produced wheat (Triticum sp.), triticale (Triticum × Secale), and corn (Zea mays).

Outline of the methods and statistical analysis

Both statistical analyses and visualizations were performed in R software [26]. We employed the DIN:TP ratio to assess whether fishponds were N-limited, P-limited, or co-limited during the growing season. This approach allowed us to investigate the seasonal patterns of nitrogen (N) and phosphorus (P) limitation across a substantial regional sample of fishponds. All fishponds presented in this study were used for common carp production.

We aggregated the data to predict chlorophyll-a concentrations from P and N concentrations and analyzed patterns of limitation in the following steps: (1) we calculated monthly mean values of TN, TP, DIN, chlorophyll-a (proxy of phytoplankton biomass), TN:TP, DIN:TP for each year at each fishpond; (2) we calculated the growing season means for each year at each fishpond from the monthly means calculated in the first step; (3) we took the means over the years for each fishpond calculated in the second step to generate the single mean value of each variable for each fishpond. The data on water quality variables were missing in some cases (i.e., low detection value); therefore, in such cases, the fishpond was excluded from the analysis (see below).

The seasonal patterns of limitation

To quantify seasonal patterns of nutrient limitations, we classified each fishpond into one of three altitudinal range groups based on altitude and fish production [27]:

  1. (a)

    Lowlands: < 199 m a.s.l., annual fish production ∼ 800 kg ha−1, the warmest region with the longest growing season.

  2. (b)

    Midlands: 200–449 m a.s.l., annual fish production ∼ 500 kg ha−1, the region with moderate growing seasons.

  3. (c)

    Highlands: > 450 m a.s.l., annual fish production ∼ 450 kg ha−1, the region with overall lower temperatures and, thus, the shortest growing season.

Each pond-month in each altitudinal category was classified into one of the three limitation categories based on Ptacnik et al. [28]:

  1. (i)

    likely P-limited: mean DIN:TP > 5;

  2. (ii)

    likely N-limited: mean DIN:TP < 2;

  3. (iii)

    likely N and P co-limited: mean DIN:TP between 2 and 5.

We decided to rely on a single ratio because drawing upon recent literature that investigates nutrient dynamics in various lake types, the DIN:TP ratio emerges as a valuable indicator for predicting phytoplankton limitations in polymictic nutrient-rich lakes [29]. Fishponds indeed resemble polymictic lakes by their mean and maximal depths, mixing processes and trophic state [16]. While DIN represents the available inorganic N forms for primary production, TP reflects the biomass of phytoplankton. On their own, DIN or PO4-P, and TN or TP concentrations are unreliable indicators of N or P limitation, respectively, as phytoplankton can have considerable internal stores of nutrients [30]. Moreover, PO4-P (or dissolved reactive P) concentrations in fishponds are often below detection limits [16].

The dynamics of phytoplankton biomass

We analyzed the data using generalized linear mixed models (GLMMs) with Gaussian distribution and log10(x) transformations to explore which variables predict the concentration chlorophyll-a best. We used the DIN:TP ratio and altitude as explanatory variables in interaction with the seasonality (April–September), as we expected different trends during the growing season, but we also tested the models without interaction. Based on previous studies [7, 15, 16, 27], we expected chlorophyll-a concentrations would vary with a pond surface area, so we used this variable in all candidate models. A preliminary inspection of our data further suggested that chlorophyll-a did not have a linear trend with explanatory variables, and we thus used pond surface area, DIN:TP ratio, and altitude as a second-order polynomial. This led to a set of 17 candidate models differing in the use of fishpond surface area, DIN:TP, altitude, and their second-order polynomial, and with or without interaction with a season as explanatory variables (Table S2). The year of sampling and locality were used as random intercepts in the GLMMs. We then used a model selection approach to build a set of candidate models and used the corrected Akaike information criterion (AICc) to rank these models and identify the most parsimonious ones [31].

We used the function glmmTMB from the package glmmTMB to analyze the GLMMs [32]. We built up all possible candidate models, including the null model and compared their AICc. In all analyses, we retained all plausible models with ΔAICc ≤ 2 [31]. We checked model residuals using the package DHARMa. If necessary, we transformed [33] and summarised them using the packages bbmle for ΔAICc, weight and df [34] and package performance for marginal R2 [35]. We visualized the most parsimonious model in the package ggeffects [36] using ggeffect function.

Results

We examined 32 fishponds in Czechia across different regional and temporal scales sampled over the growing seasons (Apr–Sep), resulting in 150 pond-year cases (Table S1). There was no single critical factor in the dynamics of available nutrients (N and P) and chlorophyll-a concentrations. Variables such as pond surface area and altitude contributed to the seasonal variability of both chlorophyll-a concentration and DIN:TP ratio (Table 1).

Table 1 Seasonal variation in environmental variables and stoichiometric ratio (median and range) of all fishponds under study
Full size table

The seasonal patterns of limitations

We found clear differences between altitudinal categories in the proportion of fishponds predicted to be P-limited, N-limited, and co-limited (Fig. 2). In the lowland region, approximately 80% of fishponds were predicted to be N-limited, and this proportion was constant across all months. P limitation occurred only in August, while co-limitation was observed in all months except June, averaging 17%. Fishponds in midlands were P-limited primarily from April to June. The proportion of P-limited fishponds declined towards the end of the season from 34% to zero, while the number of N-limited fishponds increased from 54% in April to 90% in August–September. Fishponds in highlands experienced strong seasonal shifts in limitation from P to N. The proportion of P-limited fishponds declined strongly from 85% in April to 24% in August–September, while the proportion of N-limited fishponds increased from 3% in April to 52% in August–September. The co-limitation state averaged for ~ 22% in all months.

Fig. 2
figure 2

Seasonal limitation patterns in fishponds located in three different altitudinal ranges. The bar plots show the proportion of fishponds whose phytoplankton biomass is predicted to be limited by N or P or co-limited by both nutrients. Altitudinal range groups are based on the length of the growing season, temperature, and annual fish production

Full size image

The dynamics of phytoplankton biomass

The concentration of chlorophyll-a showed high variability both on regional and seasonal scales. Based on the most parsimonious model (marginal R2 = 39.7%, Table S2), chlorophyll-a concentrations increased with the pond surface area peaking at around 31 ha, followed by a slight decrease (Fig. 3A, Table S2). Chlorophyll-a exhibited different trends during the growing season in regards to DIN:TP and altitude over the growing season. Chlorophyl-a remained stable during the beginning of the season (Apr–May) for the DIN:TP ratio but became skewed towards a lower DIN:TP ratio at the end of the season (Aug–Sept) (Fig. 3B). Additionally, concentrations of chlorophyll-a tended to be higher at lower altitudes and decreased with increasing altitude. The lowest concentrations of chlorophyll-a were observed at the beginning and the highest at the end of the growing season (Fig. 3C).

Fig. 3
figure 3

Predicted effects of pond area (A), DIN:TP ratio (B), and altitude (C) on the chlorophyll-a concentrations (log10 transformed) in fishponds as identified by the most parsimonious model. Non-focal continuous values are fixed at the mean value in the dataset. Dots represent individual data, while lines and ribbons denote model predictions with a 95% confidence interval. Colours represent different months (only in B and C). Vertical dotted lines (only in B) indicate N or P limitations as follows: the left side to the first vertical line represents N-limited ponds (DIN:TP < 2), the central part (2–5) indicates co-limitation, and the right side from the second vertical line represent P-limited ponds (DIN:TP > 5)

Full size image

Discussion

On a global scale, fishpond aquaculture remains very important yet challenged by many factors, such as price, sustainability, ecosystem demands, and resource utilization [37]. Fishpond farming practices in Czechia have not changed since the 1990s, although eutrophication and climate change have progressively devaluated the ecological functioning of these production systems [7, 10, 15, 21]. In this study, we investigated local and seasonal dynamics of available macronutrients and chlorophyll-a as the proxy of phytoplankton biomass in fishponds under conventional aquaculture management involving supplementary feeding and, in some cases, manuring. The local and seasonal variation in nutrient relationships is poorly understood, which may hinder eutrophication management and, ultimately, effective fish production [38,39,40,41].

Our results clearly demonstrated seasonal shifts in nutrient limitation (Fig. 2), which also Dolman et al. [29] reported for German shallow (polymictic) eutrophic lakes. The effect of seasonal variations in both concentrations and ratios of N and P might be particularly evident in shallow lakes and ponds, where P concentrations peaked during summer while N concentrations, particularly dissolved mineral forms, were considerably lower [16, 29, 42, 43]. These opposing trends were likely due to increased summer denitrification and sediment P release [16, 19, 44]. Our study also revealed significant seasonal variations in limiting nutrients, with more frequent N limitation in summer, as recently observed in shallow polymictic lakes [29]. Such patterns might differ in deeper lakes, particularly those stratified and oligotrophic, and/or in regions with high atmospheric N deposition (at higher altitudes), which tend to be P-limited [45, 46]. Indeed, this pattern partly might explain the higher proportion of P-limited fishponds at higher altitudes in our study (Fig. 2).

Graeber et al. [47] have recently analyzed consistent stoichiometric relationships between TN, TP, and chlorophyll-a across shallow lakes that are often tautologous, i.e., autocorrelated as in this study (see Fig. S1). Cyanobacteria, which usually dominate phytoplankton assemblages in the second half of the growing season, are effective in N2 utilization, hence increasing the TN concentration. Thus, N is more abundant in phytoplankton biomass than in bioavailable forms [16, 24]. This could challenge the precision of single nutrients in reflecting N limitation. While Graeber et al. [47] assumed that NO3-N likely accumulate in shallow lakes over long-term periods, our results clearly suggest that such an assumption may not apply for hypereutrophic and/or eutrophic shallow waterbodies with surplus organic matter and common anoxia [12, 16, 20, 24]. Therefore, the DIN:TP ratio seems to be a better predictor of chlorophyll-a than the TN:TP in shallow hypereutrophic lakes. In addition, the DIN:TP ratio exhibited a positive relationship with water transparency in this study (see Fig. S1). The high water transparency likely indicated the P limitation, resulting in reduced chlorophyll-a concentrations, in particular in highlands or, overall, at the beginning of the growing season (April–May).

Our results suggested that fishponds might be stressed by seasonal and local stoichiometric imbalances of available macronutrients. The regional scope, namely altitude, affected the TN and TP contents present in fishponds. The fishponds situated at higher altitudes tend to exploit areas characterized by diminished agricultural management intensity, lower temperature, and lesser population density compared to those in lowlands [48]. These conditions could lead to a reduced nutrient influx, resulting in lower total nutrient budgets [10]. This was also supported by our correlation analysis (see Fig. S1 in supplementary materials), where altitude exhibited a strong correlation with TN, TP, conductivity, and alkalinity. Our results showed that chlorophyll-a concentrations were affected not only by nutrient dynamics but also by altitudinal range and pond surface area (Fig. 3). Indeed, the altitudinal range could reflect the nutrient budget as discussed above. The pond surface area most likely reflected the intensity of fish farming. Our correlation plot revealed that the pond surface areas were highly associated with fish production. Fishponds with dense fish stock were likely to get higher nutrient inputs through feeding and manuring [15] and/or through fish nutrient regeneration (excretion and faecal losses) [49]. This nutrient excess might ultimately support primary production, leading to higher phytoplankton biomass. Moreover, additional nutrient pollution from allochthonous sources [8] could support further phytoplankton growth in synergy with the cascading effect of planktivorous fish, which suppressed the keystone daphnids [16, 23, 50]. In contrast, nursery ponds of smaller size were characterized by lower fish stock biomass, thus supporting the presence of submerged and emergent macrophytes [22]. The macrophytes could retain nutrients and effectively compete with phytoplankton for available nutrients, which ultimately resulted in lower phytoplankton biomass (Fig. 3).

Some fishponds under study received nutrients from manuring and feeding with cereals [5, 20]. Nutrient inputs from feed and manuring can significantly influence pond ecosystem functioning, as recently demonstrated by previous studies [16, 49, 51]. For instance, cereals represent ca. 2 kg undigested P in each ton fed per hectare of pond [6, 21]. Common carp likely excreted the undigested phytate in cereals, which was apparently utilized by dense microbial communities in fishponds [52, 53]. At the same time, intensive community respiration usually causes oxygen depletions [12] with consequent N losses via denitrification [17, 19, 20]. Contrary to N-fixing cyanobacteria, algal species are more sensitive to an N shortage and unable to grow in the absence of inorganic N [54]. Such a stoichiometric imbalance in inorganic nutrients may severely bottom-up affect phytoplankton primary production, zooplankton secondary production, and, ultimately, fish production [7, 16]. In many cases, the oxygen depletions could even be endorsed by the addition of manure at the beginning and/or during the growing season to enhance the productivity of the fishpond. However, reliable data on feeding and manuring practices usually were not available.

Until present, planktonic food-web interactions in fishponds have been rather understudied and little attention has been paid to their stoichiometry [16, 52] and its further manipulation to achieve the desired goals, i.e., good water quality and fish production. Any changes in the current fishpond management practices may not necessarily result in an improvement of the ecological state due to a certain hysteresis from the past long-term 'whole-pond fertilizing experiment' with a P legacy in the fishpond sediments. The P legacy endorsed by carp bioturbation certainly will further increase the internal P loading [55]. Both total nutrient loading and seston N:P ratio in fishponds are different today from those in the past as warranted by changing nutrient status, stoichiometry, and climate [10, 16, 24, 56].

Nowadays, eutrophication management in shallow eutrophic lakes often supports the reduction of P and N in tandem to achieve good ecological status [57]. Nonetheless, fishponds had been built for fish production purposes; therefore, they are supposed to be, to some limit, rather eutrophic and nutrient-rich water bodies. We suggest not focusing on the overall reduction of N and/or P but rather on better nutrient and/or resource use efficiency through, for example, innovative feeding strategies [49, 58] or balancing pond stoichiometry [16, 39, 40, 59, 60]. Nutrient and organic matter inputs are promising avenues to balance DIN:TP in water. An offset of the DIN:TP ratio to higher values could shift phytoplankton composition from cyanobacteria to green algae and/or cryptomonads, the more nutritious food for keystone species like daphnids [16, 24, 54] and ultimately reflect in both fish production and overall quality of a pond ecosystem. Due to the present hypereutrophic state of Czech fishponds and the changing climate [10, 16], pond ecosystems deserve more efficient nutrient cycling, avoiding P accumulation, and maintaining long-term ecosystem health. This study highlights the importance of understanding local and seasonal nutrient dynamics that would ultimately aid in innovative management strategies for future fishpond farming.