Aquaculture Europe 2021

October 4 - 7, 2021

Funchal, Madeira

Add To Calendar 07/10/2021 11:20:0007/10/2021 11:40:00Europe/LisbonAquaculture Europe 2021MODELLING THE Sparicotyle chrysophrii OUTBREAKS IN GILTHEAD SEABREAM Sparus aurata MEDITERRANEAN AQUACULTURECongress AuditoriumThe European Aquaculture Societywebmaster@aquaeas.orgfalseDD/MM/YYYYaaVZHLXMfzTRLzDrHmAi181982

MODELLING THE Sparicotyle chrysophrii OUTBREAKS IN GILTHEAD SEABREAM Sparus aurata MEDITERRANEAN AQUACULTURE

 

M. Berrettini*a, R. Baricb, S. Colakb, M. Kolegab, D. Mejdandžicb, M.L. Fioravantic, A. Gustinellic, L. Parmac and C. Virolia

 

aDepartment of Statistical Sciences, University of Bologna (Italy)

bCromaris, Zadar (Croatia)

cDepartment of Veterinary Medical Sciences, University of Bologna (Italy)

E-mail: marco.berrettini2@unibo.it

 



Introduction                     

Sparicotyle chrysophrii represents one of the main parasitic threats for gilthead seabream (Sparus Aurata) cultures in the Mediterranean basin (see e.g. Alvarez-Pellitero, 2004), causing mortality and poor fish quality. A data set has been collected by the fish farm Cromaris, including ectoparasite counts and environmental hydrographic variables. In this study, a logistic regression model is fitted to identify factors that are connected to Sparicotyle outbreaks.

Data

The proposed analysis focuses on a single Cromaris farm, located in Bisage. In particular, seven cages have been monitored on eight examination days in the first year of life of the 2017 generation. For each of the resulting 56 observations, Sparicotyle presence has been checked on the left gill arches of 12 or 13 randomly drawn fishes. To understand the factors that may influence parasite presence, the dataset is integrated with some concomitant information about fishes (and the environment where they are bred) on the day of examination, such as actual water temperature (°C), average fish weight (g), actual fish density (kg/m3) and average haematocrit value (%).

Model specification

Let Yi be a random variable describing the observed number of Sparicotyle-infected fishes out of ni (12 or 13) sampled from a single cage on a specific date. For each observed sub-sample denote with the values taken by P concomitant variables on the ni sampled fishes (or the cage they are drawn from). Conditional on , each Yi is assumed to be independent and to follow a Binomial distribution with parameters ni and , where the latter represents the probability to observe the presence of Sparicotyle chrysophrii in one of the four left gill arches of a single fish. This probability is assumed to depend on the concomitant covariates through a logistic function, as

Results

All the covariates introduced in the Data Section have been included in the logistic regression model, together with the degree days (i.e. the sum of daily temperatures), which take into account the amount of time that passed between stocking fish on farm to the day of examination. Nevertheless, the latter has been removed from the model, as well as the average haematocrit value, during the variable selection phase, carried out via backward approach. The resulting model is adequate according to the goodness-of-fit test (p-value = 0.258).

According to Table 1, the odds of a fish being sparicotyle-infected increase by about a 10.4% with each additional degree Celsius of temperature, and about a 4% with each additional gram of weight, keeping constant the values of the two remaining covariates. On the contrary, fish density seems to be negatively related with the parasite presence. However, it is worth noting that the latter coefficient estimate presents a sensibly higher standard error with respect to the two other variables.

Conclusions and further developments

The logistic regression model, fitted on the Cromaris data, highlights some statistically significant potential determinants of the Sparicotyle chrysophrii outbreaks in the gilthead seabream Mediterranean aquaculture. Nevertheless, the results presented are referred to a preliminary analysis which will be extended by integrating additional variables and observations. Moreover, some further work can be done to better fit the logistic regression model to the Cromaris data.  For instance, random effects for each cage could be included in the analysis, as in the mixed paradigm (see e.g. McCulloch & Neuhaus, 2005). Furthermore, although a portion of the effect of time is already being caught by the variables considered (temperature and weight, in particular), the within-cage time dependence could be made explicit, e.g. by specifying an autocorrelation structure based on autoregressive-moving-average (ARMA) models (see e.g. Diggle et al., 2002).

Acknowledgement

This research was undertaken under the NewTechAqua (New technologies Tools and Strategies for a Sustainable, Resilient and Innovative European Aquaculture ) project, which has received funding from the European Union’s Horizon 2020 Programme under grant agreement No 862658  (https://www.newtechaqua.eu/).

References

Alvarez-Pellitero, P. (2004). Report about fish parasitic diseases. Etudes et Recherches, Options Mediterranennes. CIHEAM/FAO, Zaragoza, 103-130.

Diggle, P., Diggle, P. J., Heagerty, P., Liang, K. Y., Heagerty, P. J., & Zeger, S. (2002). Analysis of longitudinal data. Oxford University Press.

McCulloch, C. E., & Neuhaus, J. M. (2005). Generalized linear mixed models. Encyclopedia of biostatistics, 4.

/p>