Variable climatic conditions dominate recent phytoplankton dynamics in Chesapeake Bay

Variable climatic conditions strongly influence phytoplankton dynamics in estuaries globally. Our study area is Chesapeake Bay, a highly productive ecosystem providing natural resources, transportation, and recreation for nearly 16 million people inhabiting a 165,000-km2 watershed. Since World War II, nutrient over-enrichment has led to multiple ecosystem impairments caused by increased phytoplankton biomass as chlorophyll-a (chl-a). Doubled nitrogen (N) loadings from 1945–1980 led to increased chl-a, reduced water clarity, and low dissolved oxygen (DO), while decreased N loadings from 1981–2012 suggest modest improvement. The recent 30+ years are characterized by high inter-annual variability of chl-a, coinciding with irregular dry and wet periods, complicating the detection of long-term trends. Here, we synthesize time-series data for historical and recent N loadings (TN, NO2 + NO3), chl-a, floral composition, and net primary productivity (NPP) to distinguish secular changes caused by nutrient over-enrichment from spatio-temporal variability imposed by climatic conditions. Wet years showed higher chl-a, higher diatom abundance, and increased NPP, while dry years showed lower chl-a, lower diatom abundance, and decreased NPP. Our findings support a conceptual model wherein variable climatic conditions dominate recent phytoplankton dynamics against a backdrop of nutrient over-enrichment, emphasizing the need to separate these effects to gauge progress toward improving water quality in estuaries.

. Chesapeake Bay showing major rivers, cities, salinity zones, and sampling stations for pigments and cell counts. Study-site map was generated using the software package Surfer v. 8 (Golden Software; URL http://www.golden software.com) customized with Photoshop v. CS6 (Adobe; URL http://www.adobe. com).
long-term data assembled from hydrological sampling, shipboard monitoring, and aircraft remote sensing. These findings have implications for understanding phytoplankton dynamics in estuarine ecosystems subject to contemporaneous forcing by nutrient over-enrichment and climatic conditions.

Results
Freshwater flow, nitrogen loading. Annual freshwater flow from the Susquehanna River (SRF), the largest river entering Chesapeake Bay, was highly variable for calendar years from 1945-2015, reflecting irregularly spaced dry and wet years (Fig. 2). Successive decades of low and high annual SRF prevailed in the 1960s and 1970s, corresponding to protracted drought and flood conditions, respectively. Since the 1970s, flow extremes have been frequent, exemplified by several high-flow years after the mid-1990s, e.g., 1996,2003,2004, and 2011, all well above the 75 th percentile of annual SRF from 1945-2015 of 40.7 (×10 9 ) m 3 yr −1 (Fig. 2). Despite high inter-annual variability of annual SRF, there was no significant monotonic trend from 1945-2015 based on a Mann Kendall test (p > 0.778). Observed TN and NO 2 + NO 3 loadings (Fig. 3, filled dots, SE bars) increased significantly from 1945-2012, approximately doubling by the early 1980s, and characterized by high inter-annual variability over the last 30+ years.
Inter-annual variability of TN and NO 2 + NO 3 loadings (see Methods for data sources) resembled that of annual SRF as concentrations measured at the gaging station varied much less than freshwater flow 2 . We accounted for this effect of climatic conditions by developing generalized additive models (GAM, see Methods) to obtain flow-adjusted TN and NO 2 + NO 3 loadings (Fig. 3, solid lines). Additional details of the statistical approach to generate flow-adjusted outputs for loadings and water-quality properties are given by Harding et al. 2 . Annotations on the historical trajectories of TN and NO 2 + NO 3 loadings depicted in Fig. 3(a) correspond to the post-World War II period (1) wherein flow-adjusted TN loadings from GAM increased from 28 to 64 (×10 6 ) kg yr −1 from 1945-1980 (+ 121%), leveling off around 1980 (2), and decreased loadings (3) from 66 to 53 (×10 6 ) kg yr −1 from 1981-2012 (− 19.2%). NO 2 + NO 3 loadings increased from 21 to 41 (×10 6 ) kg yr −1 from 1945-1980 (+ 89.8%), and decreased from 41 to 39 (×10 6 ) kg yr −1 from 1981-2012 (− 5.3%). These flow-adjusted TN and NO 2 + NO 3 loadings confirmed steep increases that culminated in nutrient over-enrichment, followed by gradual decreases after the implementation of nutrient-management strategies in the early 1980s.

Phytoplankton biomass as chlorophyll-a (chl-a). Phytoplankton biomass as mean annual surface chl-a
increased in oligohaline (OH), mesohaline (MH), and polyhaline (PH) salinity zones from 1950-2015, reflecting long-term nutrient over-enrichment of the bay punctuated by successive drought -flood periods in the 1960s and 1970s (Fig. 4, solid dots, SE bars). Mean annual surface chl-a in the OH salinity zone was ~3.5 mg m −3 in the early 1950s, increased steeply to 4.5 to 9.3 mg m −3 in the 1960s, and reached maximum concentrations of 18 to 23 mg m −3 by the early 1970s. After the mid-1970s, surface chl-a in the OH salinity zone decreased by half, declining from >20 mg m −3 to relatively constant, recent concentrations that range from 9.1 to 11 mg m −3 from 2006-2015 (Fig. 4a).
Mean annual surface chl-a in the MH salinity zone increased from 0.78 to 0.89 mg m −3 in 1950-51 to 1.9 to 6.0 mg m −3 in the mid-1960s, reaching 3.5 to 10 mg m −3 by the mid-to late-1970s (Fig. 4b). The upward trajectory of surface chl-a in the MH salinity zone resembled that for the OH salinity zone early in the time series, except chl-a has continued to increase gradually in the MH salinity zone after the 1980s, showing inter-annual variability from 8.3 to 13 mg m −3 .
Consistent with the increases of mean annual surface chl-a in OH and MH salinity zones, concentrations in the PH salinity zone increased from 0.39 to 0.83 mg m −3 in 1950-51 to 1.4 to 2.0 mg m −3 by the late-1960s, and ranged from 4.1 to 8.1 mg m −3 in the late-1970s (Fig. 4c). After the 1980s, surface chl-a in the PH salinity zone has shown high inter-annual variability, ranging broadly from 4.0 to 11 mg m −3 through the mid-2000s. Recent values of mean annual surface chl-a have remained fairly constant, ranging from 5.4 to 7.2 mg m −3 from 2006-2015.
Time-series data for mean annual surface chl-a showed inter-annual variability resembling that of annual SRF and unadjusted TN and NO 2 + NO 3 loadings (compare Figs. 2-4), prompting us to account for climatic conditions using GAM to derive flow-adjusted surface chl-a. GAM for surface chl-a in OH and MH salinity zones ( Fig. 4a,b, solid lines) explained >70% of the variances with R 2 = 0.702 and 0.814, respectively (Table 1). GAM for surface chl-a in the PH salinity zone explained >85% of the variance with R 2 = 0.864 ( Fig. 4c; Table 1). Comparison of observed, fitted, and flow-adjusted values of mean annual surface chl-a in OH, MH, and PH salinity zones confirmed a strong effect of climatic conditions on phytoplankton biomass.
Long-term trends from 1950-2015 computed from flow-adjusted chl-a ( Fig. 4) showed >100% increase in the OH salinity zone, and >1000% increases in MH and PH salinity zones. Decadal trends revealed the increase of flow-adjusted chl-a in the OH salinity zone reversed in the 1970s, while upward trends continued in MH and PH salinity zones after 1980, although the trajectories were shallower than from 1950-1980. Harding et al. 2 provided additional details on long-term trends of water-quality properties based on flow-adjusted data.
Spatio-temporal distributions of surface chl-a based on data from in-situ monitoring and aircraft remote sensing of ocean color were distinct for dry and wet years (Fig. 5a,b). Surface chl-a from in-situ monitoring was higher in MH and PH salinity zones during wet years than the LTA or dry years, contrasted with the OH salinity zone where wet years had lower surface chl-a (Fig. 5a). Remotely sensed chl-a in spring showed a similar response as in-situ chl-a to climatic conditions, with higher remotely sensed chl-a during wet years than the LTA or dry years (Fig. 5b) 18 . Restricted aircraft operations limited retrievals for the OH salinity zone, accounting for limited matches with in-situ chl-a for the uppermost bay (Fig. 5a,b).

Floral composition.
Climatic conditions affected floral composition as the proportions of chl-a in major taxonomic groups (f_chl-a taxa ). Diatoms were the predominant taxa in the bay, with significant contributions from dinoflagellates, cryptophytes, and cyanobacteria throughout the study period. The proportion of diatoms in the flora was affected by climatic conditions in all seasons with higher f_chl-a diatom during wet years than the LTA or dry years (Fig. 6). Diatoms were the most abundant taxa in spring with 4% higher f_chl-a diatom during wet years than the LTA and 12% higher than dry years. Cryptophytes were less abundant in spring during wet years with 3 to 5% lower f_chl-a crypto than the LTA or dry years, while dinoflagellate abundance was unchanged by climatic conditions, with nearly identical f_chl-a dino during wet, LTA, and dry years (Fig. 6a-c). Diatoms and dinoflagellates were more abundant in summer during wet years than the LTA or dry years. f_chl-a diatom was 4%    Table 1. Statistics for generalized additive models (GAM) of nutrient loadings and chl-a for Chesapeake Bay. Property = log 10 TN, NO 2 + NO 3 , or log 10 chl-a; predictor variables = year and log 10 annual SRF (10 9 m 3 yr −1 ). 1 Generalized cross-validation score. 2 *p < 0.05; **p < 0.01; ***p < 0.001; non-significant -NS.
higher during wet years than the LTA and 7% higher than dry years, while f_chl-a dino was 6% higher during wet years than the LTA and 7% higher than dry years. Cyanobacteria were less abundant in summer during wet years with 6% lower f_chl-a cyano than the LTA and 10% lower than dry years ( Fig. 6d-f). Diatoms and cryptophytes were the most abundant taxa in fall and reflected climatic conditions as in other seasons. f_chl-a diatom in fall during wet years was 6% higher than the LTA and 12% higher than dry years, while cryptophytes were less abundant in fall during wet years with 5% lower f_chl-a crypto than the LTA and 15% lower than dry years ( Fig. 6g-i). ANOVA showed significant contributions by season, salinity zone, and year to the variances of f_chl-a taxa for several major taxonomic groups, with highly significant (p ≪ 0.001) effects of climatic conditions on all but one group (Table 2). Variable climatic conditions also affected cell size and floral composition, evident for consecutive dry (1995) and wet (1996) years as the cumulative proportion of cell volume (CV) vs. equivalent spherical diameter (ESD) and corresponding f_chl-a taxa (Fig. 7). The cell-size distribution in spring during the wet year (1996) showed a shift to larger cells compared to the dry year (1995) (Fig. 7a,b). Differences of cell-size distributions were manifested as higher 10 th percentile, median, and 90 th percentile of ESD, and higher f_chl-a diatom during the wet year (1996) than during the dry year (1995). Similar effects of climatic conditions were observed in summer, with a shift of the cell-size distribution to larger cells, shown as higher 10 th percentile, median, and 90 th percentile of ESD and higher f_chl-a diatom during the wet year (1996) than during the dry year (1995) (Fig. 7c,d). A shift of cell-size distribution to larger cells during 1996 reflects the increased abundance of large, centric diatoms in MH and PH salinity zones where seasonal N-limitation is partly alleviated by higher SRF and increased N-loadings.
Net primary productivity (NPP). Annual cycles of euphotic-layer chl-a and NPP based on data from 1982-2004 revealed displaced maxima of biomass in spring and NPP in summer (Fig. 8). Mean euphotic-layer chl-a ranged from 40 to 100 mg m −2 , with a spring maximum ~100 mg m −2 during wet years that was ~20% higher than the LTA or dry years. Consistent with the climatic effect on euphotic-layer chl-a in spring, we observed a summer mean >60 mg m −2 during wet years that was higher than the LTA or dry years. NPP showed a broad summer maximum during wet years, with a mean ~2250 mg C m −2 d −1 , higher than the LTA of 2000 mg C m −2 d −1 and 1500 mg C m −2 d −1 during dry years.   Seasonal and salinity-zone means for salinity, chl-a, euphotic-layer chl-a, and NPP from 1982-2004 aggregated by climatic conditions showed strong contrasts for dry, LTA, and wet years (Table 3). Significant contributions to the variances of salinity, chl-a, euphotic-layer chl-a, and NPP by salinity zone, season, and climatic condition were documented by ANOVA (p < 0.01). Chl-a and euphotic-layer chl-a were higher in MH and PH salinity zones during wet years than the LTA or dry years, and NPP was also higher in these salinity zones during wet years. Opposite responses to climatic conditions occurred in the OH salinity zone than in MH and PH salinity zones. Plots of euphotic-layer chl-a and NPP versus salinity for OH, MH, and PH salinity zones illustrate these effects (Fig. 9). The lower salinities corresponding to wet years were accompanied by lower euphotic-layer chl-a and NPP in the OH salinity zone, while both measures of phytoplankton dynamics were higher in MH and PH salinity zones (see arrows, Fig. 9). This pattern reflected a combination of light and nutrient limitation along the main axis of the bay, wherein high SRF and low salinity in the OH salinity zone in wet years were accompanied by high turbidity, leading to reduced euphotic-layer chl-a and NPP. Conversely, high SRF and lower salinities in MH and PH salinity zones were accompanied by increased N-throughput and fertilization of seaward regions of the bay, leading to higher euphotic-layer chl-a and NPP in wet years.

Discussion
Land-margin ecosystems are among the most productive on Earth, contributing disproportionately by area to secondary production and fisheries yields 34 . Biotic richness of these ecosystems hinges on high phytoplankton biomass and primary productivity, driven by nutrient loadings from anthropogenic and natural sources. Intense land-use in bordering watersheds often results in nutrient over-enrichment, producing symptoms of water-quality degradation such as high chl-a, reduced water clarity, and low dissolved oxygen (DO) in bottom waters 3,4 . Historical evidence of ecosystem responses to human activities in Chesapeake Bay is chronicled in stratigraphic signatures spanning several centuries 35,36 . During the 20 th century alone, TN and TP loadings to the bay increased 6-and 17-fold, illustrating the rate of cultural eutrophication 10 .
Water-quality trends in the bay have been described before: (1) Murphy et al. 15 reported improving DO based on hypoxic volume, density stratification, and TN loading from 1985-2009; (2) Zhou et al. 16 showed no significant DO trend, with climatic conditions explaining inter-annual variability of hypoxic volume; (3) Prasad et al. 37 estimated trends from gridded time-series data for chl-a, nutrients, and freshwater flow using monitoring data from 1985-2008; (4) Williams et al. 38 computed biotic metrics for the main stem bay and its tributaries from 1986-2008. These studies were restricted to monitoring data collected since the mid-1980s, however, and did not provide a historical perspective for the period prior to the most serious ecosystem impairments.
Here, we accessed long-term data for Chesapeake Bay covering earlier decades when cultural eutrophication accelerated, leading to multiple ecosystem impairments by the 1980s 3 . This approach allowed us to extend the analysis to a period when TN and NO 2 + NO 3 loadings were less than half the maximum values in the time series (Fig. 3). Our reasoning for focusing on TN and NO 2 + NO 3 loadings was that N limits phytoplankton growth and production on an annual scale in the bay 39 . We found that flow-adjusted TN loading reached a maximum of ~69 (×10 6 ) kg yr −1 in the mid-1980s and declined to ~53 (×10 6 ) kg yr −1 by 2012 (Fig. 3a). Flow-adjusted NO 2 + NO 3 loadings reached a maximum of 45 in the late-1980s to early-1990s and declined to 39 (×10 6 ) kg yr −1 by 2012 (Fig. 3b). Despite progress toward nutrient reductions in recent decades, annual TN and NO 2 + NO 3 loadings have yet to reach thresholds of 40 and 28 (×10 6 ) kg yr −1 , respectively, deemed necessary to achieve bottom-water DO >1 mg L −1 2,14 . Current loadings are also higher than unmet goals transcribed in the Chesapeake Bay Agreement for the year 2000 8 . Clearly, more aggressive action is required to counteract the steep increases of TN and NO 2 + NO 3 loadings after World War II.
Complicating analyses of long-term trends, climatic conditions produce irregular dry and wet years, leading to high variability of annual SRF and nutrient loadings in the past several decades (Figs. 2 and 3). These conditions generate ecosystem-scale responses of phytoplankton dynamics observed as contrasting chl-a, floral composition, and NPP during dry and wet years. Here, we argue that distinguishing long-term trends of phytoplankton dynamics from variability traceable to climatic conditions is essential to quantify secular changes. Specific results for nutrient loadings (Fig. 3), chl-a (Figs. 4 and 5), floral composition (Figs. 6 and 7), and NPP (Figs. 8 and 9) support this argument.
Climatic conditions strongly affect nutrient loadings, with important ramifications for phytoplankton dynamics. N inputs are higher during wet years, extending the areal extent of N-sufficiency seaward to MH and PH salinity zones. Conversely, N inputs are lower during dry years and N-limitation occurs further landward in the MH salinity zone. Long-term data on surface chl-a from shipboard and aircraft measurements spanning several decades show the effects of climatic conditions on spatio-temporal distributions of phytoplankton biomass (Figs. 4 and 5). Climatic conditions also affect inputs of bio-optically active constituents, with higher inputs of dissolved and suspended materials in wet years than dry years, influencing light penetration as K D (PAR) 2 . The spatial extent of light-and N-limitation is expressed in distributions of surface chl-a, floral composition, and NPP that reflect strong forcing by climatic conditions 13,17,20 .
Combined data from satellite ocean color (SeaWiFS), shipboard surveys, and aircraft remote sensing have been used to document effects of climatic conditions on inter-annual variability of chl-a in waters of the bay and adjacent W. Atlantic shelf 18,30,40 . Wet years with high annual SRF deliver more nutrients to the bay and support higher chl-a, while dry years with low annual SRF deliver less nutrients and support lower chl-a (Figs. 2-4). Wet years with higher chl-a also show changes of floral composition as the area of N-sufficiency is expanded seaward, expressed as a higher proportion of diatoms (Fig. 6), and accompanied by a rightward shift of the cell-size distribution (Fig. 7) 20 . Accordingly, management-driven derivations of numerical chl-a criteria for Chesapeake Bay encompass regulatory goals and thresholds for low-flow, mid-flow, and high-flow conditions, separating short-term effects on chl-a traceable to climatic conditions from long-term trends signaling responses to nutrient reductions 12 .
Wet years show a higher frequency of sea-level pressure patterns corresponding to warm, wet conditions, coinciding with high annual SRF, increased nutrient delivery, and higher spring-bloom chl-a in MH and PH salinity zones as nutrient-limitation is alleviated. Conversely, dry years show a higher frequency of sea-level pressure patterns corresponding to cool, dry conditions, coinciding with low annual SRF, decreased nutrient delivery, and lower spring-bloom chl-a in these same salinity zones (Fig. 4b,c) 18 . This fertilizing effect of high SRF along

Table 3. Mean (±SE) salinity, chl-a, euphotic-layer chl-a, and NPP from 1982 to 2004 using data aggregated by climatic conditions.
the landward-seaward axis of the bay expresses a mix of hydrological effects on nutrient loadings and bio-optical properties that strongly influences phytoplankton dynamics. A notable example of this interplay of light and nutrients is evident as higher chl-a observed in MH and PH salinity zones following the phosphate detergent ban in the 1970s, resulting from increased N-throughput and a seaward extension of N-sufficiency 2 . Such effects on the chl-a distribution reflect slow, non-linear responses to gradual decreases of N loadings since the early 1980s: increased P-limitation now supports lower chl-a in the OH salinity zone, allowing more N to reach MH and PH salinity zones where ambient light conditions are conducive to phytoplankton growth. Non-linear trajectories for recovery of Chesapeake Bay to previous conditions are likely to resemble complex responses of phytoplankton dynamics to nutrient reductions in a number of estuarine and coastal ecosystems 41,42 . Based on results for other ecosystems where management actions have successfully reduced nutrients, we cannot anticipate significant decreases of chl-a, diatoms, and NPP with the modest decreases of TN and NO 2 + NO 3 loadings to the bay that have been achieved thus far. Nor do we anticipate decreases of chl-a, diatoms, and NPP will follow linear trajectories and return directly to past conditions. Changes in the efficiency of biomass production expressed as increased chl-a / TN in OH and MH salinity zones since 1995 is consistent with observations for recovering ecosystems 2 . This is likely because other biotic resources in the bay have changed over the past 60-70 years, among them precipitous declines of sea grasses, eastern oysters, and Atlantic menhaden 3 . The loss of these components of an intricate food web, combined with recent water-quality improvements, lend stability to mean annual surface chl-a, manifested as relatively constant flow-adjusted values in the PH salinity zone and a continuing, gradual increase in the MH salinity zone (Fig. 4b,c). This stability, once climatic conditions are taken into account, reflects multiple influences on phytoplankton dynamics in the bay, rather than resilience.
Comparable effects of climatic conditions on phytoplankton dynamics to those reported here occur in estuarine and coastal ecosystems around the world [22][23][24][25][26][27][28][29][30] . For Chesapeake Bay, water-quality properties were related to indices of the North Atlantic Oscillation (NAO) and El Niño -Southern Oscillation (ENSO), revealing significant relationships of chl-a and freshwater flow to NAO and ENSO that explained only 0.25 to 23% of the variances 37 . We found a synoptic climatology 31 proved more effective to explain inter-annual variability of phytoplankton, zooplankton, and fish than these basin-scale indices 18,32,33,43,44 , although neither approach produced a quantitative solution separating long-term trends from variability imposed by climatic conditions. The approach presented here accounts for variable climatic conditions by generating flow-adjusted outputs using GAM, thereby accomplishing that separation. Variable climatic conditions significantly affect euphotic-layer chl-a and NPP, reflecting effects of annual SRF on distributions of light and nutrients in the bay (Figs. 8 and 9; Table 3). Wet years with high annual SRF experience high N-loading and an increased spatial extent of high K D (PAR), while dry years with low annual SRF experience opposite effects on N loading and K D (PAR) 2 . These patterns were discussed in an earlier study based on a subset of data (1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998), defining annual cycles of euphotic-layer chl-a and NPP that were characterized by a spring chl-a maximum and a subsequent summer NPP maximum 13 . NPP data from that study supported estimates of AIP ranging from 282 to 538 g C m −2 yr −1 13 , values that were consistent with outputs of a depth-integrated model (DIM) applied to remotely sensed observations that estimated AIP ranging from 400 to 500 g C m −2 yr −1 33 . Coincident forcing of euphotic-layer chl-a and AIP by climatic conditions was evident in the significant, linear relationship of AIP to annual mean euphotic-layer chl-a that explained ~62% of the variance 33 . AIP values of this magnitude place the bay among 'eutrophic' ecosystems based on the classification scheme of Nixon 45 .
Sensitivity of chl-a, floral composition, and NPP to climatic conditions has consequences for secondary production in Chesapeake Bay 46 . Wet years with higher chl-a, diatoms, and NPP correspond to enhanced recruitment of Atlantic menhaden, an important Clupeid fish in the eastern United States. Since the late 1980s, age-0 menhaden have been more abundant in years of high phytoplankton biomass, particularly from April-June when juvenile menhaden acquire the ability to filter feed and consume phytoplankton cells. Although frequently hypothesized, such relationships between chl-a, NPP, and recruitment levels for marine fishes have rarely been confirmed 47,48 . These findings demonstrate that variable climatic conditions dominating phytoplankton dynamics may have a pervasive effect on the food web, consistent with Ryther's seminal paper linking photosynthesis to fish production in major ocean provinces 34 .

Conclusions
We demonstrate here that climatic conditions affect phytoplankton dynamics in Chesapeake Bay in predictable ways, captured in a conceptual model that synthesizes our findings (Fig. 10). Summarizing: • Flow-adjusted N-loadings provide a measure of the trajectory of eutrophication and changes since management actions were put in play; • Time series of chl-a exhibit features reflecting both management actions (P-ban, OH salinity zone), and ecosystem-scale effects (N-throughput, MH and PH salinity zones) suggesting a non-linear recovery path; • Climatic conditions affect floral composition, with higher diatom abundance and a shift toward a larger cellsize distribution during wet years; • Primary productivity as NPP is higher during wet years, with ramifications for secondary production and fish recruitment; • Together, climatic conditions act on the ecosystem in predictable ways through effects on phytoplankton dynamics.  and summed over time (10 9 m 3 yr −1 ). Climatic conditions were categorized as dry, long-term average (LTA), or wet using annual SRF and a synoptic climatology for the mid-Atlantic region 31,32 . Flow-based categorizations were based on long-term mean annual SRF from 1945-2015 of 34.1 (×10 9 ) m 3 yr −1 , with 25 th and 75 th percentiles used to delineate dry and wet years, respectively (Fig. 2). We developed a synoptic climatology in previous studies using frequencies of 10 sea-level pressure patterns based on data from the National Climate Data Center (NCDC) 18,32,33,43,44 . Designations of individual years as dry, LTA, or wet using annual SRF and the synoptic climatology agreed and water-quality data were aggregated using these categories.
Water-quality properties. Data  Remotely sensed chl-a. Ocean color measurements from aircraft as part of the Chesapeake Bay Remote Sensing Program (http://www.cbrsp.org) were used to obtain remotely sensed chl-a data presented in Fig. 5 Floral composition. Samples to determine concentrations of algal photopigments (n = 540) were collected on spring, summer and fall cruises from 1995-2004. Surface water was vacuum-filtered (< 150 mm Hg) onto glass-fiber filters (GF/F or equivalent) to collect phytoplankton. Filters were flash frozen on dry ice, held at − 20 °C until return to the laboratory, and stored at −80 °C prior to analysis using high-performance liquid chromatography (HPLC). Van Heukelem et al. 51,52 provided additional details for extraction and chromatographic methods. Reconstructions of floral composition from pigment concentrations generated the proportions of chl-a in seven major taxonomic groups using CHEMTAX software 53,54 . The method relies on concentration matrices of diagnostic carotenoids and chlorophylls to compute abundances. Outputs consisted of proportions of diatoms, cryptophytes, dinoflagellates, cyanobacteria, prasinophytes, two groups of haptophytes (types 7 and 8), and chlorophytes for each group (f_chl-a taxa ). Data were aggregated by salinity zone, season, and climatic conditions. Harding et al. 20 provided additional details of floral composition determined using algal photopigments. Cell counts were obtained from semi-monthly to monthly cruises of the CBP monitoring program from 1985-2007 for nine stations in the bay. Samples were collected in duplicate 20-L carboys using a submersible pump at several equally spaced depths above the pycnocline. Aliquots (500 ml) were withdrawn from each carboy, fixed in buffered Lugol's solution, and preserved in formalin on return to the laboratory. Cell dimensions for each species were used to compute cell volume (CV) from species-specific shape codes. Corrections were applied to taxonomic groups with vacuoles (e.g., diatoms) to compute plasma volume (PV). Phytoplankton records were obtained from the CBP data hub (http://www.chesapeakebay.net/data), including identifications to species, National Oceanographic Data Center (NODC) database designations, two-digit codes for major taxonomic groups, cell counts, dimensions, CV, PV, and cell carbon (C) 55 . Data were aggregated by salinity zone, season, and climatic conditions. Harding et al. 20 provided additional details on methods for estimating floral composition from cell counts and dimensions.
Bottles were placed in simulated in-situ sunlight incubators on the ship's upper deck and cooled with flowing surface water. Layers of neutral density screens allowing 58, 34, 21, 11, 4 and 1% transmission were used