Further reviewer’s comments to revised manuscript by Y. Liu, H.E.M. Meier and K. Eilola “Nutrient transport in the Baltic Sea - results from a 30-year physical-biogeochemical reanalysis” submitted to “Biogeosciences”
The manuscript has been substantially improved in many aspects, from clarification of procedures and algorithms to its language. However, besides of possible further stylistic cosmetics, there is still a couple of important obscurities left, which need to be either clarified or entirely removed in order to increase the paper’s credibility because retaining clearly questionable issues reduces confidence in more plausible (reliable) results.
1. General comments and suggestions
1.1. As the major objectives of the data assimilation, the revised Introduction lists: a) the reconstruction of the water quality with high resolution, especially in under-sampled periods and areas, b) the estimation of nutrient transports from more realistic nutrient fields, c) the implementation of reconstructed fields as initial and boundary conditions. In addition to these, I would also explicitly stress the capability of dynamical estimating of the water nutrient pools, especially their long-term developments, as indicators of the trophic state (see 2.2 below).
1.2. However, I still miss the explicit indication of limitations and applicability of this approach already in Introduction and even, perhaps, in Abstract. In your response to me you indicated: “In the introduction section we further clarify the already listed limitations of data assimilation with respect to estimating nutrient budgets and we rewrite the objectives of this study.” However, the first mention: “As a reanalysis can never be dynamical consistent and does not preserve mass, momentum and energy…” (Lines 100 – 101) of non-conservativeness seems appeared here out of nowhere. Consider, please,
a) re-formulation as “As a reanalysis does not CONSERVE mass, momentum and energy and, (var. - therefore, consequently, thus) can never be dynamicalLY consistent and …” and
b) moving to- (or making similar statement) somewhere between lines 68-72.
1.3. Most important among such limitations is an inherent non-conservativeness of the approach. As you correctly admitted in several places of your responses: “The potential impact from artificial sources or sinks due to data assimilation is of course also included in the reanalysis results. Because of the unknown impact from this “process” it is better to avoid detailed discussions …especially about the changes in the nutrient pools… about the net sources and sinks.” First of all, why “potential” when it is a real artificial change of simulated masses of nutrients (see also lines 423-425)? Unfortunately, you have also neglected to estimate its possible magnitude as a difference between 3D pools estimated before and after assimilation, e.g. simultaneously with RMSD (lines 218-219). I understand that it might be too laborious to systematically calculate such differences over the entire integration period but then it could be made as a few cases, perhaps, at the very first “act of assimilation” and then a couple of times more, just to show a magnitude. Such estimates are important because I can hardly see a justification of your statement at lines 427-429 in Fig. 3, where RMSDs do not show really significant differences between the beginning and the rest of REAN experiment.
1.4. Furthermore, I now think that instead of “avoiding the discussion” about erroneous sources and sinks, their usage and analyses should be excluded entirely. Consider, please, for instance, your estimates of internal sources and thinks as a difference between integrated inflows and outflows of nutrients into every grid cell, where water transports is calculated as a product of the concentrations and the current velocity vectors in the considered and surrounding cells (Eq. 5). Without data assimilation these concentrations are determined by the explicitly parametrized pelagic and sediment biogeochemical processes and, thus, the resulted differences can justifiably be interpreted as sources and sinks of the model variables. In the case of data assimilation, these concentrations are artificially altered by unknown values, unknowingly variable both in time and space. These dynamical errors are then uncontrollably propagated over the simulation domain. As you correctly noted at lines 458-460: “…usually small artificial sources and sinks from the data assimilation are becoming as important as physically motivated sources and sinks when sums of fluxes are compared.” Therefore, your estimates are not strictly comparable to those in Eilola et al. (2012) and lead to errors if being implemented in budgeting. As you say: “In this study, we focus mainly on nutrient transports derived from the reanalysis” (Line 207) and I advise to stay strictly with that.
1.4. Nutrient budgets. 1.4.1. Definitions and algorithms. It is still not clear enough, how the “burial” terms in Figs. 9-10 were calculated. Terminologically, is it a proxy of real sediment burial (i.e. permanent removal from the model domain) or a sediment retention/accumulation, especially at a long-term (1970-1999) budgeting scale? Or even simpler – is it just integral sink or nutrient removal, accounting also for denitrification? Is it actually a difference between nitrogen fixation and denitrification that is accounted here for? Do you consider these “burial” values as “the sediment budgets” (line 99)? If it is the integration of sources/sinks described in Section 5.5 then the critique from 1.3 above apply. If “The sediment sinks (burial) are calculated from the difference between the net deposition of nutrients to the sediments and the release of nutrients from the sediments” (lines 227-228) then it should be better explained a) what does “net deposition” mean – was it explicitly calculated as the difference between total sedimentation of plankton+detritus+ resuspended sediments minus uplifted sediments and b) how the sediment release of nutrients was computed, why it is not presented and analysed, which could be more relevant and interesting than the entire Section 5.5? Please, consider carefully and explain more precisely.
1.4.2. There are also internally contradicting statements and unfulfilled intentions:
Lines 100-102 – “As a reanalysis … does not preserve mass… the calculated budgets are compared to … other studies … as consistency check”. OS: – if only within an order-of-magnitude, because of different approaches and compared periods.
Lines 459-464 – “…one cannot expect that budgets calculated … with data assimilation are more accurate… Hence, we calculated budgets with the aim … to estimate the magnitude of artificial sources and sinks by comparing our results with other studies…”
OS: Such estimate was not made.
1.4.3. Most likely, the above inconsistencies are the reason why the presented budgets as a whole are highly unrealistic. The budget’s terms are presented as average for 30 years including “tendencies” from Table 1. Then the starting and final total stocks (TS) can formally be calculated as follows: TSinit = TSaverage – dTS*15 and TSfinal = TSaverage + dTS*15. For the Gulf of Finland such estimates give in kT: Pinit = 29.9 – 3.7*15 = -25.6, Pfinal = 29.9 + 3.7*15 = 85.4, Ninit = 127 – 16*15 = -113, Nfinal = 127 + 16*15+ 367; for the Kattegat: Pinit = -29.3, Pfinal = 51.7, Ninit = -378, Nfinal = 522, etc.
At the same time, both the total pools and estimated nutrient exports and imports across the basin boundaries looks realistic enough to serve as another solid source of information for other studies and conclusions, if not being compromised by the doubts in “burial” terms estimated with artificial effects of assimilation.
1.5. Algorithm of the data assimilation. Now it looks more comprehensive and save the reader from looking into similar papers. Unfortunately, it still too formal and, to my mind, is not narratively explanative enough for fellow marine biologists and biogeochemists. Perhaps, that is why it leaves (or arises) some additional questions, such as, for instance, about construction of BEC from simulation for 1962-1968, when erroneous effects of the poorly prescribed sediment (especially, P) nutrients should be even larger.
2. Specific comments and suggestions.
2.1. Consider, please, re-estimation of nutrient budgets using only external nutrient inputs and nutrient flows across basin’s boundaries. The sinks/sources can then be found as the rest terms assuming either steady state or accounting also for the differences between initial and final pools (see 2.2) below. Then, the analysis of budgets should be revised accordingly both in Results and Discussion.
2.2. The potential of the integrated nutrient pools estimation from more realistic (?) 3D nutrient fields reconstructed by data assimilation into 3D coupled physical-biogeochemical could be exploited much more for description of eutrophication. For that as well as for the budgeting, it is not even necessary to calculate annual dynamics but just, say, 5 (10?) year averages in the beginning and end of simulation.
2.3. In addition to running RMSDs, a comparison of such average pools between REAN and FREE simulations (perhaps, also with BED) as the measure of improvement could be interesting too,
2.4. End of Section 3. A few sentences are needed here to explain what does indication “in BED” in Sections 5.2-3 mean and how the long-term average March fields in 5.3 are reconstructed.
2.5. Lines 220-224. As follows from Eq. 5, the product of velocity vector “Vk” and concentration “Ck”, both taken at the location “k”, represents just transports (or outflowing nutrient mass) at this location and not “net transports”, which word better to keep for real differences between integral inflows and outflows, like you use for derivation of (semi-artificial) internal sources and sinks in Section 5.5. as well as for more plausible really net transports between basins and along cross-sections in Section 5.6. Please, check carefully the usage of “net transports” and correct accordingly, for instance, at line 295.
2.6. Line 273-274 – “…an improved model description of vertical transports of nutrients in the layers above the halocline.” Since Fig. 4 says nothing about water movements, explain, please, the reason(s) for a statement about improvement of transports – is it due to realistic vertical gradients or something else? Where is this claimed analysis of vertical transports?
2.7. Consider, please, making a separate Section starting from line 372
3. Minor things, technical corrections and language cosmetics
Title – I feel, “Nutrient transportS…” would sound better
Line 97 - “This information can’t be obtained from neither observations alone or…” The grammar construction should be either “…CAN NOT … EITHER… OR…” or “ CAN… NEITHER… NOR…”, I would recommend the former
Line 100 - “…dynamicalLY consistent..” - it is adverb
Line 101 – replace, please, “preserve” by “conserve” because we are dealing here with the laws of conservation of mass and energy…
Section 5.1 – Please, restore reference to Fig. 3
Line 408 – “…nutrients are not detailed enough known.” Please, rephrase less clumsy: “are poorly known”, “are insufficiently known”, “are known not detailed enough”, are known not in enough detail”
Line 752 – Ref. to Terruzi et al (2014) should be to pp. 200-217
Line 817 –“…Northwestern Gotland Basin, Bothnian Sea…” – Gulf of Finland is missing in this listing. |
- Printer-friendly version
- Supplement
