Marjorie A. M. Friedrichs

Skill assessment for coupled biological/physical models of marine systems

Coupled biological/physical models of marine systems serve many purposes including the synthesis of information, hypothesis generation, and as a tool for numerical experimentation. However, marine system models are increasingly used for prediction to support high-stakes decision-making. In such applications it is imperative that a rigorous model skill assessment is conducted so that the models capabilities are tested and understood. Herein, we review several metrics and approaches useful to evaluate model skill. The definition of skill and the determination of the skill level necessary for a given application is context specific and no single metric is likely to reveal all aspects of model skill. Thus, we recommend the use of several metrics, in concert, to provide a more thorough appraisal. The routine application and presentation of rigorous skill assessment metrics will also serve the broader interests of the modeling community, ultimately resulting in improved forecasting abilities as well as helping us recognize our limitations.

Assessment of skill and portability in regional marine biogeochemical models: Role of multiple planktonic groups

[1] Application of biogeochemical models to the study of marine ecosystems is pervasive, yet objective quantification of these models’ performance is rare. Here, 12 lower trophic level models of varying complexity are objectively assessed in two distinct regions (equatorial Pacific and Arabian Sea). Each model was run within an identical onedimensional physical framework. A consistent variational adjoint implementation assimilating chlorophyll-a, nitrate, export, and primary productivity was applied and the same metrics were used to assess model skill. Experiments were performed in which data were assimilated from each site individually and from both sites simultaneously. A cross-validation experiment was also conducted whereby data were assimilated from one site and the resulting optimal parameters were used to generate a simulation for the second site. When a single pelagic regime is considered, the simplest models fit the data as well as those with multiple phytoplankton functional groups. However, those with multiple phytoplankton functional groups produced lower misfits when the models are required to simulate both regimes using identical parameter values. The cross-validation experiments revealed that as long as only a few key biogeochemical parameters were optimized, the models with greater phytoplankton complexity were generally more portable. Furthermore, models with multiple zooplankton compartments did not necessarily outperform models with single zooplankton compartments, even when zooplankton biomass data are assimilated. Finally, even when different models produced similar least squares model-data misfits, they often did so via very different element flow pathways, highlighting the need for more comprehensive data sets that uniquely constrain these pathways.

Assimilation of JGOFS EqPac and SeaWiFS data into a marine ecosystem model of the central equatorial Pacific Ocean

Abstract A five-component (phytoplankton, zooplankton, ammonium, nitrate and detritus) ecosystem model developed for the central equatorial Pacific is reformulated in a data assimilative mode, using the variational adjoint technique. This method minimizes model/data misfits by adjusting six model parameters that were selected by assessing parameter co-dependencies and model sensitivity to parameter variations. Through the assimilation of cruise data from the US Joint Global Ocean Flux Study (JGOFS) Equatorial Pacific Process Study (EqPac), and ocean color data from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS), it is possible to reduce model/data misfit by estimating optimal parameters governing processes such as phytoplankton and zooplankton mortality, zooplankton grazing, phytoplankton growth, and the recycling of nutrients from detritus remineralization. The success of this approach is evident in that similar parameter sets are obtained even when independent data sets are assimilated. For example, the assimilation of in situ EqPac (depth-resolved) data from the 1991–1992 El Nino produces a parameter set that is nearly identical to that estimated via the assimilation of remotely sensed (surface) SeaWiFS data collected during the 1997–1998 El Nino. The assimilation of biological data also allows objective determination of whether or not a given model structure is consistent with a specific set of observations. For example, the assimilation process demonstrates that data collected during and after the 1991–1992 El Nino are consistent with the same single-species ecosystem model, thereby suggesting that El Nino conditions may not necessarily be associated with shifts in species composition. In contrast, the increased abundance of diatoms associated with the passage of a tropical instability wave in October 1992 as well as a brief period of macronutrient limitation observed from November 1997 through January 1998 violate key assumptions of the model. Assimilation of data that include these dynamics results in unrealistic simulations of the lower trophic levels. The successful simulation of these particular data sets will require that the model dynamics allow for species composition changes and alternation between macro- and micronutrient limitation. In this way, assimilation of biological data into marine ecosystem models cannot necessarily overcome inappropriate model dynamics and structure; rather, it can serve to guide model reformulation.

Anthropogenic and climatic influences on carbon fluxes from eastern North America to the Atlantic Ocean: A process‐based modeling study

The magnitude, spatiotemporal patterns, and controls of carbon flux from land to the ocean remain uncertain. Here we applied a process-based land model with explicit representation of carbon processes in streams and rivers to examine how changes in climate, land conversion, management practices, atmospheric CO2, and nitrogen deposition affected carbon fluxes from eastern North America to the Atlantic Ocean, specifically the Gulf of Maine (GOM), Middle Atlantic Bight (MAB), and South Atlantic Bight (SAB). Our simulation results indicate that the mean annual fluxes (±1 standard deviation) of dissolved organic carbon (DOC), particulate organic carbon (POC), and dissolved inorganic carbon (DIC) in the past three decades (1980–2008) were 2.37 ± 0.60, 1.06 ± 0.20, and 3.57 ± 0.72 Tg C yr−1, respectively. Carbon export demonstrated substantial spatial and temporal variability. For the region as a whole, the model simulates a significant decrease in riverine DIC fluxes from 1901 to 2008, whereas there were no significant trends in DOC or POC fluxes. In the SAB, however, there were significant declines in the fluxes of all three forms of carbon, and in the MAB subregion, DIC and POC fluxes declined significantly. The only significant trend in the GOM subregion was an increase in DIC flux. Climate variability was the primary cause of interannual variability in carbon export. Land conversion from cropland to forest was the primary factor contributing to decreases in all forms of C export, while nitrogen deposition and fertilizer use, as well as atmospheric CO2 increases, tended to increase DOC, POC, and DIC fluxes.

An assessment of phytoplankton primary productivity in the Arctic Ocean from satellite ocean color/in situ chlorophyll-a based models.

Abstract We investigated 32 net primary productivity (NPP) models by assessing skills to reproduce integrated NPP in the Arctic Ocean. The models were provided with two sources each of surface chlorophyll‐a concentration (chlorophyll), photosynthetically available radiation (PAR), sea surface temperature (SST), and mixed‐layer depth (MLD). The models were most sensitive to uncertainties in surface chlorophyll, generally performing better with in situ chlorophyll than with satellite‐derived values. They were much less sensitive to uncertainties in PAR, SST, and MLD, possibly due to relatively narrow ranges of input data and/or relatively little difference between input data sources. Regardless of type or complexity, most of the models were not able to fully reproduce the variability of in situ NPP, whereas some of them exhibited almost no bias (i.e., reproduced the mean of in situ NPP). The models performed relatively well in low‐productivity seasons as well as in sea ice‐covered/deep‐water regions. Depth‐resolved models correlated more with in situ NPP than other model types, but had a greater tendency to overestimate mean NPP whereas absorption‐based models exhibited the lowest bias associated with weaker correlation. The models performed better when a subsurface chlorophyll‐a maximum (SCM) was absent. As a group, the models overestimated mean NPP, however this was partly offset by some models underestimating NPP when a SCM was present. Our study suggests that NPP models need to be carefully tuned for the Arctic Ocean because most of the models performing relatively well were those that used Arctic‐relevant parameters.

Hemispheric asymmetry of deep water transport modes in the western Atlantic

Subtropical studies of the Atlantic meridional cold water flow show a hemispheric contrast in the dominant southward transport mode below 2000 m; in the North Atlantic, lower deep water (LDW) (1.8° ≤ θ ≤ 2.4°C) dominates with small transport of middle deep water (MDW) (2.4° ≤ θ ≤ 3.2°C), while in the South Atlantic, the opposite is observed. We use numerous observations in the western basins of the tropics to show that the transition occurs rapidly near the equator in the western Atlantic. A meridional section in the central Brazil Basin suggests zonal flows are responsible for the transition. LDW transport from the Guiana Basin (north of the equator) flows eastward in the northern Brazil Basin and is inferred to continue on through the Romanche Fracture Zone into the eastern Atlantic. An opposing flow of MDW from the eastern tropical Atlantic flows toward the western boundary, where it bifurcates to supply MDW to the Deep Western Boundary Current (DWBC) of the Brazil Basin, as well as to feed the northward flow of MDW in the Guiana Basin offshore of the DWBC. The magnitude of each of these oppositely directed flows is roughly 7 Sv. We furthermore speculate that they are connected predominantly by upwelling from LDW to MDW within the low-latitude eastern basin. The overall deep water transport system below 2000 m in the western basins of the mid- and low-latitude Atlantic is thus found to comprise the following three distinct components. (1) A strong DWBC transport of LDW with associated recirculation dominates the Guiana Basin north of the equator. (2) In the northern Brazil Basin (just south of the equator) a narrow eastward flow absorbs the LDW and carries it eastward, while a somewhat broader westward flow imports MDW into the western basin. (3) This MDW flow then bifurcates, with the southward branch causing the MDW dominance in the Brazil Basin, where the MDW dominated DWBC and associated recirculations are the third component of the deepwater transport system.

Chesapeake Bay nitrogen fluxes derived from a land‐estuarine ocean biogeochemical modeling system: Model description, evaluation, and nitrogen budgets

Abstract The Chesapeake Bay plays an important role in transforming riverine nutrients before they are exported to the adjacent continental shelf. Although the mean nitrogen budget of the Chesapeake Bay has been previously estimated from observations, uncertainties associated with interannually varying hydrological conditions remain. In this study, a land‐estuarine‐ocean biogeochemical modeling system is developed to quantify Chesapeake riverine nitrogen inputs, within‐estuary nitrogen transformation processes and the ultimate export of nitrogen to the coastal ocean. Model skill was evaluated using extensive in situ and satellite‐derived data, and a simulation using environmental conditions for 2001–2005 was conducted to quantify the Chesapeake Bay nitrogen budget. The 5 year simulation was characterized by large riverine inputs of nitrogen (154 × 109 g N yr−1) split roughly 60:40 between inorganic:organic components. Much of this was denitrified (34 × 109 g N yr−1) and buried (46 × 109 g N yr−1) within the estuarine system. A positive net annual ecosystem production for the bay further contributed to a large advective export of organic nitrogen to the shelf (91 × 109 g N yr−1) and negligible inorganic nitrogen export. Interannual variability was strong, particularly for the riverine nitrogen fluxes. In years with higher than average riverine nitrogen inputs, most of this excess nitrogen (50–60%) was exported from the bay as organic nitrogen, with the remaining split between burial, denitrification, and inorganic export to the coastal ocean. In comparison to previous simulations using generic shelf biogeochemical model formulations inside the estuary, the estuarine biogeochemical model described here produced more realistic and significantly greater exports of organic nitrogen and lower exports of inorganic nitrogen to the shelf.

The assimilation of satellite‐derived data into a one‐dimensional lower trophic level marine ecosystem model

Lower trophic level marine ecosystem models are highly dependent on the parameter values given to key rate processes, however many of these are either unknown or difficult to measure. One solution to this problem is to apply data assimilation techniques that optimize key parameter values, however in many cases in situ ecosystem data are unavailable on the temporal and spatial scales of interest. Although multiple types of satellite-derived data are now available with high temporal and spatial resolution, the relative advantages of assimilating different satellite data types are not well known. Here these issues are examined by implementing a lower trophic level model in a one-dimensional data assimilative (variational adjoint) model testbed. A combination of experiments assimilating synthetic and actual satellite-derived data, including total chlorophyll, size-fractionated chlorophyll and particulate organic carbon (POC), reveal that this is an effective tool for improving simulated surface and subsurface distributions both for assimilated and unassimilated variables. Model-data misfits were lowest when parameters were optimized individually at specific sites; however, this resulted in unrealistic overtuned parameter values that deteriorated model skill at times and depths when data were not available for assimilation, highlighting the importance of assimilating data from multiple sites simultaneously. Finally, when chlorophyll data were assimilated without POC, POC simulations still improved, however the reverse was not true. For this two-phytoplankton size class model, optimal results were obtained when satellite-derived size-differentiated chlorophyll and POC were both assimilated simultaneously.

Increased nitrogen export from eastern North America to the Atlantic Ocean due to climatic and anthropogenic changes during 1901–2008

We used a process-based land model, Dynamic Land Ecosystem Model 2.0, to examine how climatic and anthropogenic changes affected riverine fluxes of ammonium (NH4+), nitrate (NO3−), dissolved organic nitrogen (DON), and particulate organic nitrogen (PON) from eastern North America, especially the drainage areas of the Gulf of Maine (GOM), Mid-Atlantic Bight (MAB), and South Atlantic Bight (SAB) during 1901–2008. Model simulations indicated that annual fluxes of NH4+, NO3−, DON, and PON from the study area during 1980–2008 were 0.019 ± 0.003 (mean ± 1 standard deviation) Tg N yr−1, 0.18 ± 0.035 Tg N yr−1, 0.10 ± 0.016 Tg N yr−1, and 0.043 ± 0.008 Tg N yr−1, respectively. NH4+, NO3−, and DON exports increased while PON export decreased from 1901 to 2008. Nitrogen export demonstrated substantial spatial variability across the study area. Increased NH4+ export mainly occurred around major cities in the MAB. NO3− export increased in most parts of the MAB but decreased in parts of the GOM. Enhanced DON export was mainly distributed in the GOM and the SAB. PON export increased in coastal areas of the SAB and northern parts of the GOM but decreased in the Piedmont areas and the eastern parts of the MAB. Climate was the primary reason for interannual variability in nitrogen export; fertilizer use and nitrogen deposition tended to enhance the export of all nitrogen species; livestock farming and sewage discharge were also responsible for the increases in NH4+ and NO3− fluxes; and land cover change (especially reforestation of former agricultural land) reduced the export of the four nitrogen species.

Dissolved organic carbon fluxes in the Middle Atlantic Bight: An integrated approach based on satellite data and ocean model products

Continental margins play an important role in global carbon cycle, accounting for 15-21% of the global marine primary production. Since carbon fluxes across continental margins from land to the open ocean are not well constrained, we undertook a study to develop satellite algorithms to retrieve dissolved organic carbon (DOC) and combined these satellite data with physical circulation model products to quantify the shelf boundary fluxes of DOC for the U.S. Middle Atlantic Bight (MAB). Satellite DOC was computed through seasonal relationships of DOC with colored dissolved organic matter absorption coefficients, which were derived from an extensive set of in situ measurements. The multiyear time series of satellite-derived DOC stocks (4.9 Teragrams C; Tg) shows that freshwater discharge influences the magnitude and seasonal variability of DOC on the continental shelf. For the 2010-2012 period studied, the average total estuarine export of DOC into the MAB shelf is 0.77 Tg C yr-1 (year). The integrated DOC tracer fluxes across the shelf boundaries are 12.1 Tg C yr-1 entering the MAB from the southwest alongshore boundary, 18.5 Tg C yr-1 entering the MAB from the northeast alongshore boundary, and 29.0 Tg C yr-1 flowing out of the MAB across the entire length of the 100 m isobath. The magnitude of the cross-shelf DOC flux is quite variable in time (monthly) and space (north to south). The highly dynamic exchange of water along the shelf boundaries regulates the DOC budget of the MAB at subseasonal time scales.