Improved consistency between the modelling of ocean optics, biogeochemistry and physics, and its impact on the North-West European Shelf seas

We use a recently developed spectrally resolved bio-optical module to better represent the interaction between the incoming irradiance and the heat fluxes in the upper ocean within the (pre-)operational physical-biogeochemical model on the North-West European (NWE) Shelf. The module attenuates light based on the simulated biogeochemical tracer concentrations, and thus introduces a two-way coupling between the biogeochemistry and physics. We demonstrate that in the late spring-summer the two-way coupled model heats up the upper oceanic layer, shallows the mixed layer depth and influences the mixing in the upper ocean. The increased heating in the upper oceanic layer reduces the convective mixing and improves by ~5 days the timing of the late phytoplankton bloom of the ecosystem model. This improvement is relatively small compared with the existing model bias in bloom timing, but sufficient to have a visible impact on model skill. We show that the changes to the model temperature and salinity introduced by the module have mixed impact on the physical model skill, but the skill can be improved by assimilating the observations of temperature, salinity and chlorophyll into the model. However, in the situations where we improved the simulation of temperature, either via the bio-optical module, or via assimilation of temperature and salinity, we have shown that we also improved the simulated oxygen concentration as a result of the changes in the simulated air-sea gas flux. Overall, comparing different 1-year experiments showed that the best model skill is achieved with joint physical-biogeochemical assimilation into the two-way coupled model.

We use a recently developed spectrally resolved bio-optical module to better represent the interaction between the incoming irradiance and the heat fluxes in the upper ocean within the (pre-)operational physical-biogeochemical model on the North-West European (NWE) Shelf. The module attenuates light based on the simulated biogeochemical tracer concentrations, and thus introduces a two-way coupling between the biogeochemistry and physics. We demonstrate that in the late spring-summer the two-way coupled model heats up the upper oceanic layer, shallows the mixed layer depth and influences the mixing in the upper ocean. The increased heating in the upper oceanic layer reduces the convective mixing and improves by ∼5 days the timing of the late phytoplankton bloom of the ecosystem model. This improvement is relatively small compared with the existing model bias in bloom timing, but sufficient to have a visible impact on model skill. We show that the changes to the model temperature and salinity introduced by the module have mixed

25
Physical-biogeochemical ocean models are an essential element in mon- 26 itoring and forecasting of global and shelf-sea ecosystem indicators ( [1,2]). 27 However, coupled physical-biogeochemical marine modelling is a complex un-28 dertaking and a common way to simplify coupled models is to neglect the 29 impact of the biogeochemical model state on physics ( [3,2]). Although ma- (e.g. [5]) having an overall impact on the radiative terms and Earth energy 37 budget, (iii) some biogeochemical tracers influence light attenuation, mod-38 ifying the short-wave heat fluxes in the water column and therefore ocean 39 stratification ([6, 7, 8, 9, 10, 11, 12, 13, 14]), and (iv) marine ecosystems have 40 an impact on cloud condensation nuclei through the production of dimethyl 41 sulfide (DMS, [15,16,17,18]), or more directly via bubble formation ([19]). 42 The size of life's impact on Earth's physics has been subject to much de-       The irradiance at the ocean surface was calculated using the bio-optical  and also log-chlorophyll derived from the fluorescence measurements by the 228 same AlterEco glider Cabot, that was used in the physical data assimilation.

229
The assimilation is performed for log-chlorophyll, rather than chlorophyll, as 230 chlorophyll is widely known to be log-normally distributed ( [63]).

231
The assimilated in situ (EN4, glider) observations were thinned to a res-232 olution of 0.08 • (EN4), or up-scaled to the AMM7 grid (glider), with addi-233 tional temporal averaging applied to the same-day glider observations. The     Table 1: The AlterEco gliders and the variables measured by the gliders used for assimilation (6-th column), or validation (7-th column). The table uses the following abbreviations: deployment:"dpl", data assimilation:"DA", temperature:"T", salinity:"S", oxygen concentrations:"O 2 ", chlorophyll a concentrations:"Chl a" and sum of nitrate and nitrite concentrations:"NO x -". where NEMOVAR calculates directly the set of 3D increments using flow- The performance of the different simulations will be evaluated using two 328 skill metrics. The first metric is the model bias (∆Q mo ): where Q o are the observations mapped into the model grid and the Q m are 330 the corresponding model outputs. The second metric is the bias-corrected (2) The different experiments compared in this study. The first column shows the abbreviated experiment name, the second column indicates whether the two-way coupling is used and the following columns list the assimilated data. The table uses the following abbreviations: satellite:"sat", Cabot glider:"Cabot", EN4 dataset:"EN4", temperature:"T", sea surface temperature:"SST", salinity:"S", chlorophyll a:"Chl a".  The reference one-way coupled model simulates well the seasonal increase 336 of temperature in the surface ocean in late-spring / summer ( Fig.2:A, Fig.3).

337
The novel two-way coupling further increased the temperature in the upper 338 20m by around 1 • C ( Fig.2:B, Fig.3). This is a relatively major change with 339 respect to the reference run, when compared to the changes introduced to

384
We evaluated (Fig.6 and Fig.7) the skill of both the two-way coupled    the opposite is true beneath 40m depth (Fig.6:A). This means the observed 393 thermocline represents a larger gradient in temperature than the simulated 394 thermocline. The bio-optical module substantially (by > 1 • C) heats up the 395 upper 20-30m layer, increasing the vertical temperature gradient (Fig.6:C), 396 however the near-surface temperature of the two-way coupled run rises well 397 above the levels observed by the glider (Fig.6:D). The thermocline of the 398 two-way coupled model free run appears to be located above the glider ther-399 mocline (e.g. Fig.6:D) and the impact of the two-way coupling on the model  The main advantage of those left-hand (A,C,E) panels is that they allow relatively easy interpretation of the dynamical changes introduced to the reference run by the biogeochemical feedback to physics and/or data assimilation. Figure 7: Skill of the different model simulations to represent temperature ( • C, panels A-B) and practical salinity (panels C-D). The skill is measured by bias (x-axis, Eq.1) and BC RMSD (y-axis, Eq.2). The skill is evaluated for two half-year periods of 2018, the "summer" (panels A,C) defined as May-October and the "winter" (panels B,D) defined as November-April (data averaged through January-April 2018 and November-December 2018). The different simulations are represented by different colors: free run of the oneway coupled model (red), free run of the two-way coupled model (blue), assimilation of chlorophyll into the two-way coupled model (cyan), physical data assimilation into the one-way coupled model (lime), physical data assimilation into the two-way coupled model (grey) and joint physical data-chlorophyll assimilation into the two-way coupled model (orange). The different markers show comparison with different data-sets: the star stands for the VIIRS/in situ SST, the circle for the Cabot glider observations, the diamond for the remaining available glider observations (the 2018 AlterEco mission without Cabot) and the cross for the EN4 data-set. The data (SST, Cabot, EN4) which were assimilated in some of the simulations were used to validate only the simulations that avoided their assimilation. reanalyses look very similar to the assimilated data ( Fig.12:B-C, Fig.S5 and 481 Fig.S10 of SI) and also validate much better than the free runs relative to 482 the non-assimilated AlterEco glider data (Fig.11:A). senting Cabot oxygen (Fig.11:B), which is likely triggered by the fact that the 489 same simulation improves both Cabot chlorophyll (Fig.11:A) and the temper-490 ature bias (Fig.7:A). Equivalently, model skill in representing Cabot glider 491 oxygen can be improved by assimilating physical data into the model (phys 492 DA 1-way), and it is to some degree also improved by assimilating chloro-493 phyll (chl DA 1-way, chl DA 2-way), with the best performance achieved 494 when both the physical data and chlorophyll are assimilated into the model 495 ( Fig.11:B). However, the Cabot glider study is specific, since the glider mis-496 sion took place in the period of the largest discrepancy in the simulated 497 and observed productivity (Fig.8) and the oxygen concentrations were mea-498 sured by the same glider that provided temperature, salinity and chlorophyll 499 data for assimilation. For the remaining non-assimilated AlterEco gliders 500 the impact of two-way coupling and assimilation on simulated oxygen is less 501 clear ( Fig.11:B), i.e. even though AlterEco chlorophyll is improved by the 502 chlorophyll-only assimilative runs (Fig.11:A) they mostly degrade simulated 503 oxygen ( Fig.11:B). This is likely due to the complex relationship between phy- bias and a negative impact on the phosphate bias ( Fig.11:D-E). Silicate is 525 impacted more by the physical data assimilation than nitrate and phosphate, 526 but it is mostly degraded by all the assimilative runs ( Fig.11:F).  The purpose of the left-hand panels is to show the desired changes to the one-way coupled model (panel A) and how these changes are realized by the biogeochemical feedback in the free run (panel C) and in the physical data-assimilative run (panel E). The main advantage of those left-hand panels is that they allow relatively easy interpretation of the dynamical changes introduced to the reference run by the biogeochemical feedback to physics and/or data assimilation. The skill is measured by bias (x-axis, Eq.1) and BC RMSD (y-axis, Eq.2). The skill is evaluated for the full year 2018. The different simulations are represented by different colors: free run of the one-way coupled model (red), free run of the two-way coupled model (blue), assimilation of chlorophyll into the one-way coupled model (purple), assimilation of chlorophyll into the two-way coupled model (cyan), physical data assimilation into the one-way coupled model (lime), physical data assimilation into the two-way coupled model (grey), joint physical data-chlorophyll assimilation into the one-way coupled model (green) and joint physical data-chlorophyll assimilation into the two-way coupled model (orange). The different markers show comparison with different data-sets: the star stands for the satellite ocean color data, the circle for the Cabot glider observations, the diamond for the remaining available glider observations (the 2018 AlterEco mission without Cabot), the cross for the SOCAT data and the square for the NSBC climatological data-set.  Future work will use the two-way coupled model and expand the data assim-569 ilation scheme to include such strong coupling into our operational system. Supporting Information for "Improved consistency between the modelling of ocean optics, biogeochemistry and physics, and its impact on the North-West European Shelf seas" X -2 : into the two-way coupled model, "phys+chl DA 1-way": joint physical data -chlorophyll assimilation into the one-way coupled model, "phys+chl DA 2-way": joint physical data -chlorophyll assimilation into the two-way coupled model.