Ernesto Jardim (doutorando do Paulo Ribeiro)

Ernesto Jardim ernesto@ipimar.pt, is an Assistant Scientist at the Department of Marine Resources in IPIMAR, with a Degree in Marine Biology and Fisheries and a MSc on Probability and Statistics. At the moment he is doing his PhD in Fisheries Science on the University of Algarve with Prof. Manuel Afonso Dias (U.Algarve), Prof. Paulo Ribeiro (Universidade Federal do Paraná) and Dr. Manuela Azevedo (IPIMAR). Stock coordinator for Southern Hake since 2001 on the ICES Working Groups on the Assessment of Hake, Monk and Megrim. Also working on data collection programmes, he has been involved in several EU funded projects on these subject: Study Contract 94/013, FIEFA and SAMFISH. Since 2006 he is the chairman of the ICES Planning Group on Commercial Catches, Discards and Biological Sampling. He has been regularly invited to participate in STECF on (i) the Sub-Group on Research Needs dealing with several aspects of the Data Collection Regulation (DCR), evaluate MS proposals, evaluate MS activities and revision of the DCR (REG.CE. 1639/2001); (ii) the Sub-Group on Management Options regarding the development of a proposal for a Recovery Plan for Hake and Nephrops in Iberian waters; and (iii) the Sub-Group on Resource Status dealing with effort management. He has skills on database and software development in particular in data analysis systems.

1. Geostatistical Assessment of Sampling Designs for Portuguese Bottom Trawl Surveys (submetido)
2. Sampling Designs for Bottom Trawl Surveys: The Portuguese Autumn Survey Field Experience (submetido)

Submetido à Fisheries Research em 29/Junho/2006 (versão submetida).

Ref.: Ms. No. FISH896 Geostatistical Assessment of Sampling Designs for Portuguese Bottom Trawl Surveys Fisheries Research

Dear Dr. Jardim,

I can now inform you that the Editorial Board has evaluated the manuscript FISH896: Geostatistical Assessment of Sampling Designs for Portuguese Bottom Trawl Surveys.

The Editor has advised that the manuscript will be reconsidered for publication after major revision.

The comments below should be taken into account when revising the manuscript. Along with your revised manuscript, you will need to supply Revision notes in which you list all the changes you have made to the manuscript, and in which you detail your responses to all the comments passed by the reviewer(s) and the Editor. Should you disagree with any comment(s), please explain why.

To submit a revision, please visit http://ees.elsevier.com/fish/ and log in as an Author. You will see a menu item called Submission Needing Revision. The revised manuscript and covering letter can be submitted there.

You are kindly requested to submit your revised manuscript within 90 days. If your revision is received after that deadline, it may be treated as a new submission.

Kind regards,

Antoinette van den Brakel Journal Manager Fisheries Research fish@elsevier.com

Important note: If a reviewer has provided a review or other materials as attachments, those items will not be in this letter. Please ensure therefore that you log on to the journal site and check if any attachments have been provided.

Reviewer #1: Report on manuscript "Geostatistical Assessment of Sampling Designs for Portuguese Bottom Trawl Surveys" by Ernesto Jardim and Paulo J. Ribeiro

General evaluation: Acceptable after minor revision

This is an interesting, straightforward manuscript assessing the effect of sample size and spatial configuration of Portuguese bottom trawl surveys in fish abundance estimates through geostatistical methods. The writing is clear and the figures and tables appropriate. The simulations are carefully designed, including the simulated data and the set of correlation parameters with their respective maximum likelihood estimates.

General Comments

1. Line 100. "The spatial model assumed here is a Log-Gaussian geostatistical model".

In the discussion section, the authors justify the use of isotropic models (lines 338 to 341) but no explanation and/or justification about the log-Gaussian geostatistical model selected are given. Further explanation about the reasons of the model selection will clarify the results.

2. Line 241. "Table 2 summarizes the checks of the results of the parameter estimates which were considered satisfactory and coherent".

It is not thoroughly clear in the text what the authors mean with satisfactory and coherent. More detail will be relevant to better understand the sampling design and survey processes

3. Line 347. "Furthermore, the results can be retained for all species with a spatial behaviour covered by these parameters".

It seems like the authors assume all the species surveyed have similar spatial behavior. This is not necessarily true, especially if the survey is targeting species with different life history traits and aggregation behaviors under different spatial scales (i.e. demersal fishes vs. sedentary invertebrates). Furthermore, the autocorrelation structure in the data is not explicitly mentioned or described. Additional information and discussion on the effect of spatial correlation for the different stocks targeted by the trawl survey on the model selection will improve the robustness of this study.

Minor comments

Line 123: repeated word: the the

Line 125: Unnecessary word: are

Lines 183 and 206: different notation for sampling designs <LAMBDA>d and ?d

Line 234: confusing sentence/notation: "…and also included in the Table ®…"

Table 3: Summary statistics units are not specified.

Figure 1: X and Y axis legends should be specified (i.e. Longitude West and Latitude North respectively).

Figure 2: Variables in the X axis are specified in the legend but not in the figure

1. There's a paragraph (lines 325-332) justifying the use of a log transform, in particular in lines 330-332 is mentioned that the log was found on previous analysis of the historical data.

2. Acho que temos que clarificar este parágrafo (lines 241-253).

3.1 We generalized our results for all species that fit in the range of the covariance parameters used. This may not apply to invertebrates but certainly apply for most demersal species, which are the target of our survey. This sentence was revised to clarify it's aim.

3.2 The autocorrelation structure in the data is presented in Table 1 where all the correlation parameters estimated are shown and in lines 231-240 we describe them and the most important particularities found.

3.3 Paulo, não percebo o que o revisor quer na terceira frase …

Minor comments ok !

Reviewer #3: I propose rejecting this submission because it is overly detailed on the simulation results (1), gives little insight how the simulations relate to the original Portuguese survey data (2), of which little is spoken, and because it is not clear why this is to be considered more than an exercise confirming what already has been stated in Diggle and Lophaven (3). The authors do show an understanding of the issues involved in simulation and did not, in my mind, make any errors. Some of the results are technical and issues of isotropy, parameter estimation and the like are discussed at a more technical level than would be understood by a general reader. The one significant result is that when there is autocorrelation in the underlying data it is better to use a combindation of regular survey with paired random additions (to provide points close to each other and better estimate autocorrelation I presume) than a pure random design for fisheries surveys. If this is indeed a new result (I'm really not sure whether it is) then this could be acceptable as a greatly reduced in size 'note' that gives the results and refers to a web document or report for details of the simulations (4). Certainly the geostatistical equations are not needed and are better found elsewhere (5). They are not new to the fisheries literature. Finally, in simulation work like this I am left unsure how general the results are to other areas (6). This the authors discussed some and think the results are general (maybe they are). There is little need in that case to focus on the real system (7). Otherwise, some evaluation using actual data would be useful (if there were a year when higher sampling intensity was used – it could be subsampled to see how much the estimates changed) (8). In fairness to the authors I did not study the results in detail. Maybe someone who does will find gold in it. I did not think it was worth looking.

Existem algumas inconsistências que podemos explorar na resposta a este revisor. No essencial podemos a valorizar outros resultados que obtivémos como o facto da variância da média amostral ser enviesada para a variância do estimador quando há correlação espacial, ou o procedimento para comparar desenhos com tamanhos diferentes.

(1) The detailed simulation results were included to allow readers to understand the scope of our work and have enough information to judge if their own situation is inside the range of our work.

(2) The historical data was used to condition the simulation work using the covariance parameters obtained with it to define the range of the parameters used for simulation.

(3) The results obtained by Diggle and Lophaven were theoretical and not applied to a real situation, like we did. On the other hand their work compares two specific ways of building sampling designs, "lattice plus close pairs" and "lattice plus infill", and never include a pure random or regular design, which we did. Also they use only geostatistical methods and we also included a comparison of the designs performance using sampling theory estimators. We included anysotropy and log transformation on our analysis. More important of all, we describe an easy way of building a sampling design that has the characteristics of "lattice plus close pairs", by overlaping the random design with a regular design that can be applicable to most European Bottom Trawl Surveys.

(4) This results are new at least in Fisheries Science once that there is no reporting of surveys using such sampling strategy. The authors can not guarantee that the theoretical results of Diggle and Lophaven were not implemented already in other scientific areas, but the bibliographic search did not show any relevant papers about its implementation. Also there are secondary results that are new in this work (i) the approach to build the sampling designs, (ii) the approach to compare sampling designs with different sample sizes, (iii) the result about the underestimation of abundance variance by the variance of the sampling mean. The authors agree that these results should be more visible and made the necessary changes.

(5) Section 2.1 was included to make the paper self contained, providing information so that readers clearly understand the scope of the work. Also it helps readers to get familiarized with notation. However, if the Editor finds it should be decreased, we can remove some parts of it and include bibliographic references.

(6) The results are generalized by the spatial behaviour of the resource. If in another area someone exploring the spatial correlation of a resource finds parameters that fit inside the range of parameters used for our simulations, there is a good chance that the sampling design of the survey collecting its data will gain by adopting a mixed random/regular design.

(7) As said in point (2) the focus on the real system is just enough to provide information for conditioning the simulation work so that the results are applicable to the real world. There was not the intention of explore deeply the data or completely ignore it.

(8) This would be a valid approach if the spatial correlation is ignored, once that the removal of a location would not only reduce the sample size but also the configuration of the sampling design with and impact extremely difficult to assess.

From the Editor-in-Chief : One reviewer asks for relatively small changes, while the other feels the paper is not acceptable unless it is substantially shortened and focused on what is new. If the authors will react to the reviewers comments , I will reconsider the assessment.

EJ: 03/Jan/2007