Toward a Predictive Understanding of Primary Productivity in a Temperate, Partially Stratified Estuary

Abstract

This paper addresses the development of a predictive understanding of phytoplankton primary productivity (PP) in estuaries, drawing on an extensive set of observations in Chesapeake Bay and contemporary modeling approaches coupled to empirical data. PP was measured at 575 stations on 68 cruises, 1982–2000. Descriptions presented here are based on PP data from 455 stations occupied 1982–1998, and from 120 stations occupied 1999–2000 that were used for model validation. The nearly two-decade sampling period encompassed a broad range of freshwater flow and nutrient loading to the Bay. Mean net14 C-PP for 1982–1998 was 1055mgCm−2d−1 (±SE=46·0) in the main stem Bay for all seasons and regions. Phytoplankton dynamics included a spring maximum of biomass, expressed as euphotic-layer chlorophyll (chla), and a summer maximum of net 14C-PP displaced by approximately 4 months from the biomass maximum. The annual maximum of euphotic-layer chla exceeded 100mgm−2 in mesohaline and polyhaline regions of the Bay during April and May, whereas highest net14 C-PP of 1700–2500mgCm−2d−1occurred in the mesohaline Bay in July and August. Euphotic-layer chla and net 14C-PP were much lower in the light-limited oligohaline Bay and no time lag was observed. Mean gross 14C-PP was 1564mgCm−2d−1 (±SE=85·0) for 1995–1998. Concurrent measurements using 14C and O2 methods generated estimates of the photosynthetic quotient (PQ) to confirm our interpretation of net and gross PP from 14C uptake in full- and partial-day incubations, respectively, and allowed us to reconcile apparent differences in 14C- and O2-determined PP in recent studies. PQ values (=O2 produced/CO2 fixed) were estimated as 1·48 from regression of stoichiometrically converted net O2-PP on net 14C-PP, and as 1·38 from the regression of gross O2-PP on gross 14C-PP. PQ values in this range typically correspond to phytoplankton that are using oxidized nitrogen (NO−3-N) as the main N source, consistent with the view that N-limitation occurs on an annual scale in the Bay. We used empirical data for net and gross 14C-PP to estimate annual integrated production (AIP) of 282 to 538gCm−2yr−1 (net) and 347 to 662gCm−2yr−1 (gross). Simple, linear regression of net AIP on annual, mean euphotic-layer chla was significant and explained ∼62% of the variance. Inter-annual variability of net AIP was related to the volume of freshwater flow and to total nitrogen (TN) and total phosphorus (TP) loading during February and March. We tested the performance of published models to estimate PP in Chesapeake Bay. The Vertically Generalized Productivity Model (VGPM) overestimated net and gross PP, and adjusted forms of VGPM termed VGPM-A for net and gross PP gave significantly improved performance for Chesapeake Bay. We explored an alternative approach to VGPM that allowed us to obtain non-unity exponents for independent variables, using step-wise and multiple regressions in log-space first to identify independent variables that predicted net and gross PP, and subsequently to determine coefficients of the terms. This approach resulted in the Chesapeake Bay Productivity Model (CBPM-1) that estimated net and gross PP with root mean square error (RMSE) of 28·3% and 34·9%, respectively. We then developed models of the physiological input, PBopt, a variable that expresses optimal photosynthesis in the water column normalized to chla, using several independent variables and measured values of PBopt. Substitution of the models of PBopt in CBPM-1 resulted in CBPM-2 that required no explicit input of PBopt and estimated net and gross PP with RMSE of 120% and 49·8%, respectively. Lastly, we validated CBPM-1 and CBPM-2 using observations from 1999–2000 not included in development of the models. CBPM-1 estimated net and gross PP with RMSE of 17·8% and 35·8%, respectively, and CBPM-2 estimated net and gross PP with RMSE of 80·3% and 47·6%, respectively. We believe the use of contemporary PP models that require simple inputs amenable to remote sensing, such as the adjusted VGPM, CBPM-1, and CBPM-2, should permit us to resolve spatial and temporal differences of net and gross PP in estuaries and give improved estimates of AIP, an essential measure of ecosystem health and productivity.