Ref. No. [UMCES]CBL Variability in the Dynamics of Forage Fish Abundances in Chesapeake Bay: Retrospective Analysis, Models and Synthesis Robert J. Wood* Edward D. Houde University of Maryland Center for Environmental Science Chesapeake Biological Laboratory P.O. Box 38 / One Williams Street Solomons, MD 20688 *Present address: NOAA Chesapeake Bay Office 710 Severn Avenue Annapolis, MD 21403 Introduction Forage fishes form an essential link between plankton productivity and production of economically important fishes and, as such, are a critical element in multispecies fisheries management. They also serve as important components of the diet of fish-eating birds (gulls, terns, cormorants, ospreys). Some species, such as Atlantic menhaden Brevoortia tyrannus, support commercial fisheries. Declines in recruitment of the two most abundant forage species in Chesapeake Bay, Atlantic menhaden and bay anchovy Anchoa mitchilli in the 1990's have drawn attention to possible changes in the trophic state of Chesapeake Bay and concern for the welfare of fish productivity, including requirements for prey by large predator fishes. The Chesapeake Bay Program, has committed to, develop ecosystem-based multi-species management plans for targeted species by 2005 and to revise and implement existing fisheries management plans to incorporate ecological, social and economic considerations, multi-species fisheries management and ecosystem approaches by 2007 (CBP 2000), and the N.O.A.A. Chesapeake Bay Office is currently leading an effort to craft a Fisheries Ecosystem Plan for the Bay. To develop ecosystem approaches for fisheries management, a concerted research, modeling, and retrospective analysis effort is required to provide information about species relationships, trophic interactions, and the impact of interannual climate variability on the Bay ecosystem and its fisheries (Houde et al., 1998). Understanding factors that control abundances of forage fishes, because of their 1
key trophic position, constitutes an important element of multispecies management that will be implemented in Chesapeake Bay. Project Objectives Our objective is to determine the history and extent of fluctuations in forage fish abundances in Chesapeake Bay, and to provide insight into the processes and factors that may be driving historical and contemporary fluctuations in these populations. Retrospective analyses, synthesis of existing data, and modeling are the approaches that we are following to achieve these goals. Specific objectives include: i. document long-term (decadal) shifts in forage-fish abundances and biomasses, with particular emphasis on menhaden and bay anchovy; ii. iii. iv. investigate the causes of variability in Chesapeake Bay forage-fish annual recruitment and abundance; model the interactions between forage species and major predators, estimate the impact of forage species on plankton populations. Methods This report covers the second year of a two-year study. Because research during these two years was continuous, work conducted in the first year is summarized along with detailed information on research conducted during the second year. Details of the first year s research results are provided in the extended abstract presented for the CBSAC 2002 Project Presentation Workshop (http://noaa.chesapeakebay.net/fisheries/ WoodAb2002.pdf). Initial efforts during the first year of this project focused on acquiring and processing forage fish abundance data to construct mulitdecadal, species-specific population abundance time series. While this was done for several forage fish species, analyses and investigations have emphasized Atlantic menhaden and bay anchovy. These numerically dominant species both had exhibited pronounced declines within the Bay in the decade preceding our research. The two-year work plan was directed sequentially towards population assessment, exploratory analyses of the relationship(s) between anchovy and menhaden populations with potentially important extrinsic variables (hydroclimatic factors, predators, and prey), and modeling activities. The statistical and analytical tools used included multivariate and 2
univariate statistical methods (e.g. linear regression techniques, principal components analysis), spatial mapping, and simulation and bioenergetics modeling. Results and Discussion Bay Anchovy Abundance trends & potential causal mechanisms Abundance indices indicate that bay anchovy experienced a decline from high abundance years in the mid-1980's during the 1990 s. Lowest abundance occurred in 1994 and, in some locations, signs of recovery were evident in the late 1990 s. Important conclusions from this analysis were that the estuary s bay anchovy population has declined in recent decades and that the mechanism(s) influencing bay anchovy annual abundance appear to act, or at least are detectable, on a Bay-wide scale (Figure 1). Because the scale of variability is consistent with climate-driven processes, covariability between synoptic-scale climate variability and bay anchovy abundance patterns was investigated. This investigation revealed that the population s spatial distribution patterns appear to be strongly influenced by spring hydroclimatic conditions. Specifically, atmospheric circulation patterns favoring high riverflow throughout the watershed and low Chesapeake Bay salinities, resulted in peak mid-summer biomass of bay anchovy being located further down-bay relative to dry years (Figure 2) Because peak spawning occurs in July, this mid-summer variability in spatial distribution patterns is potentially important in determining interannual bay anchovy survival and recruitment patterns. These findings prompted further investigation into the potential mechanisms through which hydroclimatic variability may affect reproduction, recruitment, and survival of the bay anchovy population within Chesapeake Bay. Because hydroclimatic fluctuations are most likely to influence the population indirectly, through various ecosystem processes, our research emphasized interannual variability in hydrographic conditions and focused on two trophically important species for the bay anchovy: the copepod Acartia tonsa (a primary bay anchovy prey), and the ctenophore Mnemiopsis leidyi (predator of anchovy and A. tonsa). Comparing the Bay and its monitored tributaries, bay anchovy and mesozooplankton data (from the Chesapeake Bay Program) indicated that anchovy abundance historically has varied synchronously with that of a preferred summer prey, the copepod Acartia tonsa. Associations between these two species demonstrate strong correspondence throughout Chesapeake Bay and its tributaries (Figure 3). Building on the relationship between freshwater flow (i.e., salinity distribution) and the latitudinal distribution of pre-spawning adult anchovy, and considering that recruitment strength in fishes is often associated with 3
survival of their early life stages, bioenergetics modeling was applied to determine if A. tonsa variability could limit the growth, and therefore survival, of larval bay anchovies. Following methods and using parameters established in Houde and Madon (1995), a basic bioenergetics equation (Grodzinski et al., 1975) was used to determine whether the Consumption = Metabolic loss + Waste + Growth C = C max x p x f(t) C max = a x (Weight of fish) b Cmax = max feeding rate standardized to grams of fish (g*g-1*d-1) a = intercept of the weight dependence function b = weight dependence coefficient p = proportion of Cmax that actually achieved by the fish f(t) = water temperature dependence function for warm water fishes downward trend in A. tonsa could be a factor limiting anchovy annual recruitment. While most parameters of the equation can be derived or reasonably estimated using standard methods or assumptions, there is little information about the value of p. Solving for consumption, both p and C are unknowns, however since our objective is to determine whether consumption demand of the larval anchovy population could be limiting growth and survival, p can be set to 1. This approach can determine whether anchovy larvae, consuming at their maximum rate at all times, would deplete levels of A. tonsa (assuming it is their only prey, see Figure 4) sufficiently to limit larval growth. Using water temperature (T), A. tonsa and anchovy larvae densities collected during July by the Chesapeake Bay Program and the TIES research program for years 1995-1999, annual maps were produced of potential bay anchovy larval consumption, A. tonsa availability and the net difference or consumption-availability balance (CAB). These maps revealed that lower CAB occurred in the lower Chesapeake Bay (where spawning occurs) during poor bay anchovy recruitment years (1995 & 1996) than in years of strong recruitment (1997-1999). Further negative CAB values never occurred in years of strong recruitment (Figure 5). Using average annual CAB values derived from these data, linear regression (Figure 6) revealed that CAB adds significant explanatory power to predictions of annual anchovy recruitment (October abundance) from a given year s spawning biomass (Spring abundance). The correspondence between decline of bay anchovy and A. tonsa populations, combined with information from the bioenergetics model, suggests that a bottom-up, predator-prey linkage could control bay anchovy population abundance. However, it is also possible that the copepod and anchovy are responding to some unidentified factor that controls their abundances. To investigate this possibility, station-specific water temperature, salinity, 4
and dissolved oxygen, and both the abundance of bay anchovy and the ctenophore Mnemiopsis leidyi (an abundant predator of bay anchovy larvae) were simultaneously analyzed using principal components analysis. For this analysis, ctenophore data was available only for Chesapeake Bay Program plankton monitoring stations in Maryland waters. Principal component analyses revealed that the primary pattern of covariance among these species July densities over the years 1985-2000 is with the existence of a bay anchovy - A. tonsa - M. leidyi trophic triangle. The analysis also indicated that salinity, a physiologically relevant variable for all three species, may play a role controlling abundances in areas north of the Potomac River. (Figures 7-8). Because the abundance of the M. leidyi is thought to be largely controlled by the stinging nettle Chrysaora quinquecirrha, and reproduction in this species is sensitive to salinity (Purcell et al., 1999), we are currently investigating whether salinity changes may affect forage fish by directly controlling the Bay s stinging nettle population. Atlantic menhaden The Atlantic menhaden s early life history in the Bay is poorly known, especially the ecology and dynamics of late larval and early juvenile stages. Because Atlantic menhaden is a commercially important species, coastal Atlantic spawning population estimates are available. Coupled with the long-term Maryland DNR-derived indices of annual YOY recruitment to Maryland tributaries and the upper Bay, these estimates allowed construction of a simple Ricker stock-recruitment relationship describing the possible dependency of annual Chesapeake Bay recruitment on coast-wide spawning stock biomass (Figure 9). The difference between annual recruitment levels predicted by the Ricker relationship and the actual survey recruitment index was then calculated. These residuals were assumed to be driven by extrinsic (to the population itself) factors and were used (as the dependent variable) to investigate the statistical relationship between interannual climate variability and the variability in menhaden recruitment that is unrelated to spawning population levels. A temporal synoptic index (TSI) was constructed to describe interannual climate variability during the winter-spring transitional months of March-May. These months span the period of late-larval-stage Atlantic menhaden migration from oceanic spawning grounds to Chesapeake Bay nursery areas, a period in which menhaden recruitment strength may be established. Frequently used in climatological studies, temporal synoptic classification describes and quantifies typical weather patterns using a two-step multivariate procedure involving principal components and cluster analyses. Sea-level pressure was the input variable for this study because the goal was to investigate the role of large- scale climate patterns in influencing recruitment of Atlantic menhaden to Chesapeake Bay. The resulting temporal synoptic classification defined characteristic atmospheric circulation patterns that occur during the late winter and early spring. Further, the resulting temporal synoptic index placed each daily observation (March-May, 1966-1998) into one of these patterns. 5
Ten TSI atmospheric circulation patterns were identified. The number of days each of these ten TSI circulation patterns occurred within each of the three spring months were tabulated for each year of record, resulting in thirty time series. Of these thirty, only one pattern, the Azores-Bermuda high (ABH) pressure system, was selected by the regression tree algorithm for inclusion in the predictive model. During winter, this pattern is centered over the Azores before it begins a westward migration as spring approaches. During summer the ABH typically dominates the Chesapeake region when it is centered over Bermuda. Because this model described a simple linear relationship between the ABH and menhaden recruitment to the Bay, correlation analysis was used to estimate the strength of the relationship. Nearly half (44%) of the variation in the 1966-1997 Ricker-residual time series of menhaden recruitment levels was accounted for by the number of days in March that the ABH pattern dominated the Chesapeake Bay region s weather, as described by the TSI (Figure 10a). In comparison, the best Atlantic menhaden recruitment model constructed using individual weather variables (temperature, river discharge, wind direction, and wind velocity) included only March temperature (Figure 10b)and accounted for only half (22%) of the recruitment variability attributed to the ABH (Figure 10c). Mechanisms driving this ABH-menhaden recruitment relationship are not known, but ongoing investigations indicate that the abundance of dominant mesozooplankton prey (primary food for postlarval and small juvenile menhaden) in the low-salinity, upper Bay and tributaries peaks earlier in the year when the ABH dominates in March. This observation is important because postlarval menhaden migrate, or are transported, to low-salinity areas of the estuary during this period. The observed statistical relationship might not represent a predator-prey causal link between menhaden and zooplankton. Because high pressure systems inhibit storminess net down-bay flow would be expected to be less intense when the ABH was dominant. Further, the location of the ABH relative to Chesapeake Bay, combined with the clockwise surface wind field of high pressure systems, results in south to southeasterly winds. Menhaden are spawned near the Gulf Stream front during winter and spend their first 60-90 days in the coastal ocean and it is possible that, together, these conditions favor onshore and subsequent up-bay transport of larval menhaden from their coastal spawning grounds to their oligohaline nursery areas, thereby enhancing recruitment to Chesapeake Bay. Future analyses will investigate this possibility by modeling the transport trajectories presented by these conditions to larval menhaden acting both as passive particles and larvae capable of vertical migration. Future Research In the final months of this project, emphasis will be on synthesis of information, in addition to further development of TSI and its potential application to understanding how hydroclimatic variability controls bay anchovy and menhaden population trends. We will 6
continue to develop bioenergetics modeling approaches to determine interannual and regional variability in potential consumption, growth, and contribution to predators of bay anchovy. As in previous years, results of research will be presented at national scientific meetings. The research holds promise for constructing predictive models to guide singlespecies and ecosystem-based fisheries management. Further, results will be evaluated in the context of assessing Chesapeake Bay health, under the assumption that an adequate forage base is essential in a healthy ecosystem. The forage fish research contributes importantly to our participation in the regional effort to construct a Chesapeake Bay trophic model using the Ecopath with Ecosim modeling package. When this model is developed (Spring 2003), we will apply it to quantify the role of forage fishes in the ecosystem. References CBP. 2000. 2000 Chesapeake Bay Agreement. http://www.chesapeakebay.net/ agreement.htm. Grodzinski, W., R., Z. Klekowski, et al. (1975). IBP Handbook #24: Methods for Ecological Bioenergetics. Oxford, Blackwell Scientific Publications. Houde, E. D., M. J. Fogarty and T. J. Miller (1998). Prospects for Multispecies Fisheries Management in Chesapeake Bay, Solomons, MD, Chesapeake Bay Program STAC. Purcell, J. E., J. R. White, et al. (1999). "Temperature, salinity and food effects on asexual reproduction and abundance of the scyphozoan Chrysaora quinquecirrha." Marine Ecology Progress Series 180: 187-196. MEPS 180:187-196 (1999) 7
Bay Anchovy (MDNR seine) index values -2-1 0 1 2 3 4 Potomac River (n=5) -2-1 0 1 2 3 4 Upper Bay (n=4) 1970 1980 1990 2000 1970 1980 1990 2000 index values -2-1 0 1 2 3 4 Choptank River (n=1) -2-1 0 1 2 3 4 Nanticoke River (n=5) 1970 1980 1990 2000 1970 1980 1990 2000 Figure 1. Example bay anchovy annual abundance index time series for four different sampling regions within Chesapeake Bay. Time series for each region were standardized to z-score values to facilitate direct comparison. Within station variance is indicated by the point whiskers except for the Choptank River series, which features only 1 station (n=1). 8
Mid-Atlantic storm track Latitude 37.0 37.5 38.0 38.5 39.0 39.5 High pressure Maximum 397 1998-77.5-77.0-76.5-76.0 Northern storm track High Pressure Latitude 37.0 37.5 38.0 38.5 39.0 39.5 Maximum 449 1999-77.5-77.0-76.5-76.0 Figure 2. Comparing the distribution of the bay anchovy population (right) immediately prior to spawning (indicated by annual spring sampling cruises of the TIES project) with the hydroclimatic regime (left) present during the March-May period as indicated by average sea level pressure over these months in two highly contrasting years (1998-top and 1999- bottom). All years of the TIES project (1995-2000) displayed this pattern. 9
July Copepod abundance Annual anchovy abundance Annual abundance Annual abundance 3 2 1 0 1 2 2 1 0-1 -2 Upper Bay 1980 1985 1990 1995 2000 Lower Bay 4.5 4.0 3.5 3.0 2.5 5 4 3 2 1-2 -1 0 1 2 3 2 1 0-1 -2 Potomac River 1980 1985 1990 1995 2000 James River 4.0 3.5 3.0 2.5 4.5 4.0 3.5 3.0 2.5 1980 1985 1990 1995 2000 1980 1985 1990 1995 2000 Figure 3. Plots of bay anchovy annual abundance time series (points) and July copepod abundance (lines) for four different sampling regions within Chesapeake Bay. Upper Bay and Potomac river anchovy data are derived from the Maryland DNR while Lower Bay and James River data were provided by the Virginia Institute of Marine Science s Juvenile finfish trawl survey. Copepod abundance data provided by the Chesapeake Bay Program. 10
Larval anchovy prey occurrence >11mm (n=102) Barnacle nauplius Cladoceran Copepod copepodite/adult Copepod egg Copepod nauplius Ostracod Polychaete juvenile Rotifer Unidentified 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Figure 4. Larval anchovy prey frequency histogram for larval anchovy >11mm. Data derived from the TIES sampling program and provided by Auth and Houde 11
1995 Mean July prey density (Bay-wide mean = 0.03 g/m^3) 1995 Larval anchovy consumption (Bay-wide mean = 0.007 g/day) 1995 Mean consumption-availability balance (Bay-wide mean = 0.016 g/m^3) 37.0 37.5 38.0 38.5 39.0 39.5-76.6-76.4-76.2-76.0 Figure 5a. Maps of larval anchovy prey (A. tonsa adults + copepedites) densities (top left), maximum consumption (top right), and the consumption-availability balance (bottom), for 1995, a poor bay anchovy recruitment year. Values are linearly related to symbol surface area and negative values are represented with squares. 12
1998 Mean July prey density (Bay-wide mean = 0.08 g/m^3) 1998 Larval anchovy consumption (Bay-wide mean = 0.005 g/day) 1998 Mean consumption-availability balance (Bay-wide mean = 0.065 g/m^3) 37.0 37.5 38.0 38.5 39.0 39.5-76.6-76.4-76.2-76.0 Figure 5b. Maps as described in figure 5a, but for 1998, a strong bay anchovy recruitment year. 13
0 100 200 300 Oct. anchovy abundance July anchovy larvae Oct. abundance predicted by July larvae & prey deficit 1995 1996 1997 1998 1999 Years Figure 6. Line plots illustrating that a substantial increase in explanatory power is achieved (61% versus 83% of variance explained) when adding July CAB as an explanatory variable to a linear regression modeling Bay-wide anchovy abundance in October using only July larval abundance. 14
Upper Bay (CB3.3c) Choptank (ET5.2) Potomac (LE2.2) -0.4 0.2 anchovy A.tonsa ctenoph surf.temp surf.salin -0.6 0.0 anchovy A.tonsa ctenoph surf.temp surf.salin -0.4 0.2 anchovy A.tonsa ctenoph surf.temp surf.salin Upper-mid Bay (CB4.3) Mid-Bay (CB5.2) -0.6 0.0 A.tonsa ctenoph surf.temp surf.salin -0.6 0.2 A.tonsa ctenoph surf.temp surf.salin Figure 7. Principle component one loadings for site-specific analyses of July bay anchovy, Acartia tonsa, and Mnemiopsis leidyi (ctenophore) abundances, and temperature and salinity (integrated over surface-most 3 meters). Note: for 1985-2000, ctenophore data was collected in Maryland waters only and mainstem bay anchovy abundance is available only for the upper Bay. 15
1.5 1.0 Upper Bay (CB3.3c) Upper-mid Bay (CB4.3c) Mid Bay (CB5.2c) Choptank (ET5.2) Potomac (LE2.2) Lower Bay anchovy abundance 0.5 PC1 Score 0.0-0.5-1.0 1985 1990 1995 2000 Year Figure 8. Principle component one associated with the loadings depicted in Figure 7. For comparison, a bay anchovy abundance index is presented for lower Chesapeake Bay (Virginia mainstem). 16
Recruitment 0 10000 20000 30000 40000 50000 Annual NMFS Menhaden SSB vs. MDNR YOY Ricker SSB-recruitment curve 0 20 40 60 80 100 Annual NMFS Menhaden SSB estimates Figure 9. Ricker spawner-recruit relationship used to isolate and extract recruitment variability unrelated to spawning stock biomass. An annual abundance index derived from the Maryland DNR seine survey was used as the recruitment time series. Spawning stock biomass (SSB) data pertain to the Atlantic coastal stock and provided by the National Marine Fisheries Service. 17
a) b) Ricker curve residuals Ricker curve residuals -20000 0 30000 Individual weather variables model (one predictor chosen: March temperature) Menhaden recruitment model fitted recruitment r-squared = 0.22 65 70 75 80 85 90 95 Year (19xx) Synoptic circulation model (one predictor chosen: March Azores-Bermuda High frequency) -20000 0 20000 Menhaden recruitment Model fitted recruitment r-squared = 0.44 65 70 75 80 85 90 95 Year (19xx) 10 m/s c) Gulf moisture & storm track Azores-Bermuda High warm subtropical air advected northward Figure 10. Climate-recruitment models of menhaden recruitment (Ricker model residuals-see Figure 8). The best, statistically significant model derived from a potential predictor pool of individual weather variables (a) used only March air temperature explained only half (22%) of the variance accounted for by a model constructed with atmospheric circulation patterns (44%). Of thirty potential daily frequency time series of spring circulation patterns, only the March frequency of the Azores-Bermuda High (c) was used by the latter (b) model. 18