Missing pieces in the full annual cycle of fish ecology: a systematic review of the phenology of freshwater fish research

In recent decades, fish ecologists have become increasingly aware of the need for spatially comprehensive sampling. However, a corresponding reflection on the temporal aspects of research has been lacking. We quantified the seasonal timing and extent of freshwater fish research. Since reviewing all prior work was not feasible, we considered two different subsets. First, we compiled the last 30 years of ecological research on juvenile Pacific salmon and trout (Oncorhynchus spp.) (n = 371 studies). In addition to the aggregate, we compared groups classified by subject matter. Next, to evaluate whether riverscape ecology has embraced space at the expense of time, we compiled research across taxa in which fish were enumerated in a spatially continuous fashion (n = 46). We found that ecological Oncorhynchus spp. research was biased towards summer (40% occurred during June-August) and the month of June in particular, at the expense of winter work (only 13% occurred during December-February). Riverscape studies were also biased toward summer (47% of studies) and against winter (11%). It was less common for studies to encompass multiple seasons (43% of ecological Oncorhynchus spp. studies and 54% of riverscape studies) and most were shorter than 4 months (73% of ecological Oncorhynchus spp. studies and 81% of riverscape studies). These temporal biases may cause researchers to overemphasize ecological phenomena observed during summer and limit our ability to recognize seasonal interactions such as carry-over effects or compensatory responses. Full year and winter studies likely hold valuable insights for conservation and management.

4 63 probably most commonly studied in the form of population dynamics, i.e., fluctuations in 64 abundance typically described at an annual resolution. However, many important processes that 65 may scale up to affect population dynamics (e.g. growth) play out at intra-annual timescales and 66 relate to seasonality.

67
It is often recognized that short-term datasets can be inadequate because they fail to 68 capture historical levels of productivity (i.e. the shifting baseline) or reveal coarser scale 69 temporal patterning such as regime shifts [12]. Likewise, for cyclically patterned temporal 74 Riverine systems may exhibit extreme seasonal variation, with water temperatures ranging 20°C 75 or more and flows varying 100-fold. This strongly affects not only fish and other aquatic 76 organisms, but also the feasibility of field sampling. While a temperature logger can effectively 77 collect data every day of the year, the cost and logistical challenges of sampling fish vary 78 tremendously and can strongly govern when biological data are collected. Extrapolating from 79 data that pertain to specific points in time can lead to misleading interpretations regarding how 80 fish behave, the production capacity for ecosystems, and what locations or habitat types are 81 important [15,16]. This is particularly problematic in the study of mobile organisms that undergo 82 substantial physiological and ecological changes throughout their lifetimes, such as Pacific 83 salmonids. The objective of this paper is to characterize the temporal attributes of fish ecology 84 research to elucidate potential data gaps and guide future research.
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 6 108 studies that employed the use of spatially continuous data (or nearly so) that covered a high 109 spatial extent so that multi-scale patterns could be revealed [6] as opposed to the more typical 110 method of using of a handful of points that are extrapolated to represent large extents. We 111 focused on three temporal aspects of research: 1) what months and seasons juvenile salmonid 112 ecology research occurs, 2) the duration of studies, and 3) whether seasons were studied 113 individually or if seasonal interactions were examined.

114
To examine our first case study of Oncorhynchus research, we reviewed 13 journals that 115 commonly publish fisheries ecology research as opposed to human consumption of fish research. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 7 131 occurring in estuarine or marine environments, studies that collected physical or biological 132 habitat data but did not actually sample fish, reviews, or models not validated with field data.
133 This resulted in 371 articles examined in this study (S1 Fig). 134 To identify temporal patterns across fish species, we identified "riverscape" studies that 135 utilized spatially continuous sampling [6]. The term "riverscape" has been applied inconsistently 136 but is often used to refer to sampling employing large spatial extent. We use the term to include 137 large spatial extent sampling as well as sampling in line with the argument by  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 8 154 30 for spring. Seasons were not defined by solstice or equinox to stay consistent with 155 presence/absence within a single month. Studies may encompass more than one month, therefore 156 the number of data points for these analyses are greater than the number of studies included in 157 the review. Second, we quantified the frequency of the number of meteorological seasons (1-4) 158 that were included in these studies to analyze temporal extent and consideration of inter-seasonal 159 interactions (i.e., carry-over effects).

160
To explore whether temporal aspects of sampling differed among research areas, we 161 classified each study into three focal areas: 1) fish-habitat interactions and the impact of habitat 162 units and types on juvenile salmonid biology or behavior, 2) trophic ecology including fish diet, 163 foraging, and food web structure, and 3) spatial distribution including movement and landscape-164 scale distribution. Studies examining fish growth and survival were often presented by 165 researchers as a function of some aspect of one of the three focal areas identified and were 166 classified accordingly. The temporal distribution and extent of sampling effort was then 167 quantified both collectively and by research category. Each study was only classified into one of 168 the three focal areas based on the main objective of the study. Studies that did not fall into one of 169 these four main categories were classified as "Other" and included in overall analysis but not the 170 subset analyses.

171
172 Statistical methods 173 We tested for temporal biases in temporal distribution and extent using Pearson X 2 -tests in R 174 4.0.2. Equal values would indicate that no bias exists, supporting the null hypothesis. While the 175 test is objective, we acknowledge that the interpretation is subjective due to the assumptions that . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 9 176 all months and seasons are equally important and present equal stresses, limitations, or 177 opportunities for growth, fitness, and survival for juvenile salmonids. 178 We also acknowledge that seasonality varies with latitude, elevation, and position in 179 watershed, so the ecological conditions associated with a particular month or season may vary 180 among locations (and thus among the studies in our paper). Thus, the implications of the 181 temporal biases we observed may be somewhat context dependent.

Monthly temporal distribution of studies
185 At a monthly resolution across all ecological topics within juvenile Oncorhynchus spp. studies, 186 we found that the most frequently represented month was 3-6 times more common than the least 187 frequently represented month (Fig 1). December was the least represented month across all 188 topics, while the summer months of June, July, and August were most common among topics.
189 The month of June had a significantly higher proportion of studies than the month of December 190 at 14% and 3%, respectively. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 11 221 studies containing data from 4 months or less (Fig 3). Less than 2-8% of studies across all topics 222 encompassed data from all 12 months of the year.  (Fig 3). Only 43% of all studies collected data from multiple seasons and 243 73% of studies were shorter than 4 months. Again, there has been little change in the temporal . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 12 244 extent of research efforts with the proportion of single-season studies remaining significantly 245 higher than multi-season or year-round studies (Fig 4). (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 13 267 more than one month or season. Seasons were defined meteorologically, but as whole months.
268 Summer is defined as the months June, July, and August; Autumn is defined as the months 269 September, October, and November; Winter is defined as the months December, January, and 270 February; Spring is defined as the months March, April, and May. 306 not achieve substantial growth in many cases, there is evidence that winter fish growth may 307 exceed growth observed during other seasons for some fish [25]. Understanding winter habitat 308 use and foraging ecology could help improve our ability to increase overwinter survival.

309
The lack of winter research contrasted with the overabundance of summer studies. While 310 emphasis on summer has benefits, such as an improved understanding of salmonid ecology 311 during periods of climate stress, relying on summer-biased data could pose problems for . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 15 312 conservation and management by violating assumptions of models. For example, species 313 distribution models (SDM) are increasingly used in climate change adaptation and rely on the 314 assumptions that a species occurs in all suitable habitats and that a species only occupies a 315 portion of that suitable habitat due to constraining factors such as competition or predation [26].
316 Developing such models from temporally biased data would be valid only if the focal species 317 were sedentary and their habitat use did not vary over time. However, it's rarely possible to 318 confirm that a species meets these criteria without having temporally representative data (i.e., 319 you can't dismiss the possibility of winter habitat shifts without data on winter habitat use). (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint 358 meteorological seasons represented only 6-10% of total research. Research is heavily skewed 359 toward shorter, single season studies: 73% of all studies capturing 4 months or less of data and 360 57% of studies focused on a single season in isolation. Within riverscape studies, 81% of 361 research occurred during 4 or fewer calendar months. These patterns are similar to patterns 362 observed in the phenology of mammal, bird, reptile and amphibian research [17]. While there is 363 increasing recognition of the value of long-term study [40], this usually means having multiple 364 years or decades of data collection. Our review shows that there is also a lack of temporal extent 365 in terms of the annual cycle. Lacking extent at this timescale leads to two issues. First, we are 366 likely to temporally extrapolate and draw conclusions based on a subset of the year (as discussed 367 above) and second, we will often lack the ability to identify interactions between different time 368 periods, or carry-over effects [17]. 377 There is evidence that ephemeral food subsidy pulses, such as salmon eggs during the adult 378 spawning season, can positively influence juvenile salmon growth rate and energy density as 379 long as 6 months after this ephemeral resource pulse has disappeared [43]. Whether juvenile 380 salmonids grow large enough to consume eggs depends on their emergence timing and early . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted November 24, 2020. ; https://doi.org/10.1101/2020.11.24.395665 doi: bioRxiv preprint