Introduction

Ecologists have long recognized the value of long-term datasets. Indeed, the systematic and repeated collection of data through time has allowed scientists to better understand the dynamics of ecological systems, especially when change is slow, episodic, nonlinear, or nonstationary (Strayer et al., 1986; Magnuson, 1990; Lindenmayer and Likens, 2010). Insights gained from long-term studies led to numerous calls in the 1980s and 1990s for more longer-term studies. Collectively, individuals, agencies, and institutions have initiated annual sampling programs in recognition of the value these data provide (Magnuson, 1990; Stow et al., 1998; Morecroft et al., 2009; Abbot and Le Maitre, 2010; Peters, 2010; Vanderbilt and Gaiser, 2017). Although many long-term study programs have been developed over the last several decades, no study has evaluated whether the contribution of long-term studies to the scientific literature has changed over time.

Limnology is among the oldest ecological subdisciplines (Egerton, 2014). Since its inception, inland waters have been used as sentinels of environmental change (Pham et al., 2008; Williamson et al., 2009; Schindler, 2009; Adrian et al., 2009) as lakes and reservoirs and their food webs integrate and reflect natural and human-mediated changes to their surroundings (Coe and Foley, 2001; Adrian et al., 2009; Detmer et al., 2017). Limnologists were important early advocates for the collection of long-term data (Likens, 1983). For these reasons, limnological studies of natural lakes and reservoirs are ideal for evaluating trends in the scientific literature.

Here, we address the question of whether limnologists have responded to early calls for studies of greater longevity, allowing us some insight into whether the value that was proposed is still perceived. We analyzed trends in limnological studies of natural lakes and reservoirs over a 40-year timespan and hypothesized (1) that studies would increase in duration in response to calls for longer studies, (2) trends in increased duration would be consistent among journals, and (3) increases in study duration would be inversely related to study complexity (studies with greater complexity in number of metrics would be less likely to have greater duration). Our analysis spans the development and implementation of long-term ecological monitoring programs in the 1980s and 1990s (e.g., Likens, 1983, 1989; Strayer et al., 1986; Magnuson, 1990). For this study, we distinguished man-made reservoirs from natural lakes because they have increased greatly in relative frequency over the same period as the development of long-term monitoring programs (Wisser et al., 2013) as well as categorically differ from natural lakes in ontogeny, age, and function (Hutchinson, 1957; Saulnier-Talbot and Lavoie, 2018).

Methods

We selected scientific journals with publication histories spanning the evaluation period (1975–2015), broad readerships, and consistent publication of widely distributed research: Canadian Journal of Fisheries and Aquatic Sciences (est. 1901), Ecology (est. 1920), Freshwater Biology (est. 1971), Hydrobiologia (est. 1948), Limnology and Oceanography (est. 1956), and Transactions of the American Fisheries Society (est. 1872). Across the 40 years from 1975 to 2015, we evaluated time points at 5-year intervals. Specifically, the publication years included in the study were 1975, 1980, 1985, 1990, 1995, 2000, 2005, and 2015. Journal selection was based on the following criteria: (1) established before 1975, (2) had an H-index greater than 90 according to the Scimago Journal Ranking system (SJR; https://www.scimagojr.com) in 2018, (3) was categorized as an aquatic sciences journal by Scimago and was focused on freshwater aquatic sciences (e.g., not Aquatic Toxicology, Fish and Shellfish Immunology) and (4) was released in the English language. As an outgroup that regularly includes limnologically related work and has a long publication history, the journal Ecology was selected.

Publication selection in journals in selected years was completed based on an initial review of the title, then abstract, then full study to extract data of interest. Individual publications were included if they provided information on system-wide limnological processes within one or more lakes or reservoirs (system-wide sampling was defined here as possessing a sampling strategy intended to represent that body of water). Marine, wetland, river, mesocosm, and laboratory studies were excluded. Additionally, for each publication, the start and end date needed to be explicitly stated or duration could otherwise be determined. The waterbody or waterbodies were explicitly described to confirm status in terms of natural lake or reservoir. From each study, the type of ecosystem (lake or reservoir), study duration (year), and response metrics evaluated (biotic or abiotic, independent or response variables) were recorded. The number of study metrics from a list of 10 (Table 1) was scored for each study. The number of response metrics was recorded (1, 2, 3, 4, 5 +) to serve as a simple index of research complexity (akin to richness from a community ecology perspective). Study metrics were curated from the tables of contents of several limnological texts as well as informed by the surveyed literature included in the present study. We aimed for a balance between comprehensiveness and practicality in presenting a concise list of 10 metrics, acknowledging the potential for further categorization within specific groups.

Table 1 Percent of components in studies from different years in each lacustrine (lake or reservoir) ecosystem type
Full size table

Cumulative study duration, which we define as the sum of all study durations for a given publication year, was examined across publication years. By accounting for both number and duration of studies, we assessed a type of research productivity that is not reflected within any individual paper (i.e., an emergent property). We also surveyed temporal overlap of studies. For a given study, we subtracted the study duration from the publication year giving us an approximation of the year the study started. This simple approach allowed us to visually examine how studies overlap across the publication years.

In analyses focused on changes distribution over time, such as study duration over time, regression methods may give an inaccurate representation of the relationship between dependent and independent variables. Multiple trends within a dataset such as this can cause unequal variation in a probability distribution (Cade and Noon, 2003). Such distortions can be minimized by using quantile regression to estimate relations between variables across a frequency distribution with unequal variation (Koenker and Bassett, 1978; Cade and Noon, 2003). Here, we use meta-quantile regression as a robust means of tracking publication trends in scientific literature (Yu et al., 2016). We quantified trends (H0: no trend) in study duration across publication years for lakes and reservoirs for each quantile (τ, 0.0–1.0). For our purposes, we consider ‘long-term’ to be those studies in the 90th percentile and above (i.e., 0.9 quantile). We also examined whether changes in study durations of long-term studies (i.e., studies falling in the 0.9 quantile) were consistent across journals. Finally, we explored whether long-term studies began publishing more response metrics thru time.

Results

A total of 6,362 publications were initially identified over the five publication years, of which 801 studies met our criteria: natural lakes (661) and reservoirs (140). Over the 40-year period simple linear regression indicated that there was no change in the annual number of publications fitting our criteria over time for natural lakes (P > 0.05) or reservoirs (P > 0.05).

Study durations within the 0.1 to 0.4 quantiles showed no increase in length over the evaluation period for natural lakes (Fig. 1; τ = 0.1 to 0.4, P > 0.05). Notably, most studies of lakes and reservoirs were one year or less in duration. Slopes for quantiles 0.5–0.9 were significant for natural lakes and indicated a positive shift in study duration for the upper portion of the probability distribution. Study durations increased (P < 0.01) by factors of 2 (τ = 0.5), 3 (τ = 0.6), 4 (τ = 0.7), 9 (τ = 0.8) and 6 (τ = 0.9) across upper quantiles. The small number of reservoir studies precluded analysis of slopes due to a lack of statistical power, but the data suggest a positive trend for the 0.9 quantile slope and a lack of response for other quantiles (Fig. 1). Across studies of natural lakes in 2015, nearly 22% exceeded 10 years and 12% exceeded 20 years in duration. In contrast, of the 2015 reservoir studies, only one study (6%) exceeded 10 years.

Fig. 1
figure 1

Trends in quantiles in study durations (y-axis) of natural lakes (left panel) and reservoirs (middle panel) as a function of publication year (x-axis). Quantile regression slopes across the evaluation period for natural lakes and reservoirs (right panel). Asterisks denote a significant slope (quantile regression; P < 0.05) for natural lakes. Note: line and symbol colors correspond across panels

Full size image

The summation of all study durations in a given publication year provided further insight in publication trends. As study durations increased from 1975 to 2015, we observed greater temporal overlap among decades (Fig. 2). For example, long-term studies published in recent publication years had datasets that extended further back in time, covering more periods of time. Across publication years, we further observed that total years of research published increased 351% for studies of natural lakes (Fig. 3). Because of inter-decadal increases in study duration, cumulative years of published research were highest in 2015, even though 2015 had fewer studies meeting our criteria than 1995 or 2005 (Fig. 3). For reservoirs, changes through time were more variable, as total cumulative research productivity was 171% greater in 2005 and 24% lower in 2015, when compared to 1975. Lakes and reservoirs are combined in subsequent analyses because there were so few studies of reservoirs.

Fig. 2
figure 2

Across publication years (x-axis), the distribution (box-and-whisker plot) of study durations of natural lakes and reservoirs broadened from the year of publication. In natural lakes, temporal overlap was greatest, as evidenced by larger boxes and greater numbers of outliers (black circles)

Full size image
Fig. 3
figure 3

Cumulative years of study (y-axis) of natural lakes and reservoirs across publication years (x-axis; 1975–2015, 10-year. increments). Within each column, studies are sorted from shortest to longest duration. Numbers above each column denote the number of published studies

Full size image

The pattern of long-term studies becoming longer over four decades was observed across multiple scientific journals (Fig. 4). For example, study durations in the 0.9 quantile exceeded 10 years in 5 of 6 journals, 20 years in 4 of 6 journals, 30 years in 3 of 6 journals, and 40 years in 2 of 6 journals for the 2015 publication year (Fig. 4). The number of metrics a study included was related to study duration. For studies reporting 1 or 2 metrics, durations increased greatly over the evaluation period (Fig. 5a). For studies with 3, 4, and 5 + metrics, study durations also increased, but less rapidly (Fig. 5a). Interestingly, study durations between 1975 and 2015 were inversely correlated with the number of metrics reported within a study. For more complex studies (5 + response metrics) in the 0.9 quantile, study durations increased by about only 5 years (Fig. 5b). In contrast, studies in the 0.9 quantile reporting 1 and 2 metrics were 35–38 years longer (Fig. 5b).

Fig. 4
figure 4

A Examination of the 0.9 quantile (τ) assessing changes in study duration (y-axis) of natural lakes and reservoirs (pooled) as a function of publication year (x-axis) for each journal. B Changes between 1975 and 2015 in the study duration of the 0.9 quantile (τ) journal. Differences were arranged by increasing value. Note: line and symbol colors correspond across panels

Full size image
Fig. 5
figure 5

A The 0.9 quantile (τ) assessing changes in study duration (y-axis) of natural lakes and reservoirs (pooled) as a function of publication year (x-axis) based on the number of response metrics reported in each study. B Changes between 1975 and 2015 in the study duration of the 0.9 quantile (y-axis) as a function of the number of response metrics in published studies (x-axis). Note: line and symbol colors correspond across panels

Full size image

Discussion

This study focused on limnological assessments of natural lakes and reservoirs. Our intent was to determine whether limnological studies have changed through time with respect to duration, response metrics, and dissemination of findings in academic journals. Our meta-analysis revealed several notable patterns which collectively suggest that lake focused studies have changed over time, but reservoir studies have not changed in frequency or duration as expected. Such emergent trends in the literature indicate that limnological studies have changed in several ways. Below, we describe these patterns which collectively reveal that scientific studies of lakes have changed in terms of implementation and publication. Notably, we found that long-term studies increased in duration while short-term studies did not suggesting that study duration is likely dependent on the goal of the study and that there are either constraints or limited gains to increased duration for some studies, or both.

Our analysis revealed that long-term limnological studies of lakes have gotten longer over four decades. In contrast, studies of reservoirs did not show a strong trend for greater study duration. For lakes, we found that studies comprising the upper quantiles (τ = 0.5 to 0.9) increased in duration across publication years, as indicated by positive slopes. Differences in slopes further reveal that the longest of published studies became longer at a faster rate. Similar patterns were not detected for reservoirs because fewer studies met our criteria. Nonetheless, the slope of the highest quantile (τ = 0.9) of reservoir suggests a similar pattern, albeit weaker. Findings also revealed that short-term studies were common and did not change with respect to study duration. This lack of change may simply reflect the fact that many studies do not require long-term data (e.g., multiple years). Indeed, many ecological questions can be explored without long-term datasets (e.g., spatial distributions or predator–prey interactions). Moreover, many seasonal or annual datasets originate from student theses or other research projects that are ultimately constrained by funding or other factors.

As long-term studies became longer, findings were disseminated across multiple journal outlets. For example, studies published in Limnology & Oceanography exhibited the greatest increase (~ 45 years) in study duration within the 0.9 quantile. Although magnitudes differed among journals, long-term studies became longer in most journals. One exception was our outgroup, Ecology. Although it is unclear why the outgroup would have the least increase over time in study duration, one possibility is the focus and funding sources of labs contributing to general ecology journals versus those focused strictly on aquatic ecosystems may influence where studies are submitted.

We also observed that the number of response metrics reported within a study was inversely related to study duration such that short-term studies typically reported more response metrics and vice versa. Studies that examined fewer response metrics (1–2) were more likely to increase in study duration relative to those that assessed more (3–5 +) response metrics. Findings indicate that multifaceted, long-term studies are still rare within journals included here. There may be potential barriers for publication of complex, long-term studies that stem from the high costs associated with the collection, processing, storage, and dissemination of multiple response metrics over long periods of time.

We expected the number of reservoir studies to increase relative to natural lakes across publication years. However, no such pattern was observed. Despite changes in the abundance of reservoirs, studies of natural lakes consistently outnumbered reservoirs through time. The number of reservoirs increased rapidly during the early portion of our examination (e.g., 1975–1995) and slowed in the latter portion of the study period (Wisser et al. 2013). Inherent physical characteristics of reservoirs resulting from their construction (i.e., deep near the dam and shallow near river inputs) make them unique systems compared to lakes (e.g., Thornton et al., 1990; Wetzel, 2001; Dodds and Whiles, 2010) and therefore inappropriate for use as analogs. For example, recent long-term studies of the response of thermal and oxygen structure in reservoirs to climate change indicate that resiliency to climate change is different in reservoirs and that increased size of reservoir does not dampen the effects of climate change like it does in natural lake (Zhang et al., 2014, 2015; Detmer et al., 2022). Further complicating their use as analogs, processes driving the function of lakes and reservoirs can differ, as reservoirs are frequently constructed where natural lakes are rare (e.g., US Great Plains and Arid Southwest regions) in order to store water (e.g., potable water, irrigation), to control flooding, and for economic development (e.g., recreational fisheries, public recreation; Thornton et al., 1990; Wetzel, 2001). Given that global changes to the water cycle are not expected to be homogeneous (Vörösmarty et al., 2000; Huntington et al. 2006), spatially dependent changes to precipitation may affect natural lakes and reservoirs differently and these patterns may only become apparent through long-term studies. Knowledge and inferences from long-term studies of lakes or reservoirs may reflect unique social and environmental conditions that may hinder the translation of findings from lakes to reservoirs and vice versa (Thornton, 1990; Hayes et al., 2017).

It is likely journals included in the present study reflect trends in the field over time because they broadly cover the limnological literature, are widely read and have long histories. They may not, however, perfectly reflect all trends. Limitations imposed by our selection criteria and number of journals selected may have inadvertently misrepresented some results. In the Journal of Lake and Reservoir Management, for example, since 1990 (journal inception in 1984, but skipped releasing an issue in 1985), there has been no change in the ratio of lake to reservoir studies following the protocol described above, but the number of reservoir studies was more common and the ratio was closer to 1 to 1. Additionally, and perhaps unsurprisingly, over this long time period, some limnological fields and journals have refocused on new challenges and moved in new directions, altering the composition of the field and journal articles. For example, increasing global concern about frequency and drivers of harmful algal blooms is likely to have influenced studies related to biogeochemistry, phytoplankton, or human dimensions. In Table 1, we see those three groups becoming more prevalent in the literature that we reviewed and a decline in the frequency of studies on vertebrates and macroinvertebrates.

Although there are many drivers of study duration, the increased frequency and duration of long-term data sets highlight their value and contribution to the literature and allow the field to declare that it is answering calls for the collection and dissemination of long-term datasets (e.g., Likens, 1983, 1989; Strayer et al., 1986; Magnuson, 1990). There are several mechanisms that likely influenced increases in long-term data sets like increased funding, establishment of field stations focused on long-term monitoring, and recognition of the unique value of long-term data sets and simultaneously the resiliency of change in short-term studies suggests a different set of drivers such as duration of grants, time to graduate degree and question focus. Continued improvement in increased relative contribution of multi-year and multi-decadal studies, particularly those focused on several metrics and reservoirs, is therefore necessary to make management decisions that are robust and can stand the test of time.