A Multi-Model Paradigm in Application to the Analysis of the Factors Defining the Epizootic Situation in the Communities of Non-Synanthropic Small Mammals in Ekaterinburg

Abstract

The aim of the work was to rank statistical models for assessing the contribution of a number of factors that determine the epizootic situation on natural-focal infectious diseases in the communities of non-synanthropic small mammals (SM) in the forests of Ekaterinburg.Materials and methods. The SM survey was carried out for three years in the summer-autumn period. The animals were caught with snap tpaps on a standard bread bait, 9705 trap-nights were worked out. To identify infections (hemorrhagic fever with renal syndrome, tularemia, leptospirosis, yersiniosis, pseudotuberculosis), 333 SM specimens (rodents and shrews of six species) were used, randomly taken from the general sample. To determine the pathogens, the methods of enzyme immunoassay and polymerase chain reaction were used. The cumulative infection rate of small mammals with all studied natural-focal infections was evaluated. Statistical data processing was carried out from the standpoint of a multi-model approach. The selection of models was performed using the Akaike information criterion with the calculation of the sum of the weights of the models based on their full spectrum (SW).Results and discussion. Carriers of all of the mentioned above infections, except for pseudotuberculosis, have been identified in the forest parks of Ekaterinburg. According to the ranking, the model with three predictors received the highest weight: species, year, and the abundance of SM of the previous year in a particular habitat. The most significant predictors for the full spectrum of models are the year (SW=1), species (SW=0.6), abundance of animals in the current and previous year (SW=0.48). The distribution of positive samples by species of small mammals corresponded to their ranking by abundance in the community. The influence of the factors “season” (summer or autumn) and “area” (place of capture of animals) turned out to be insignificant on the scale of the studies (SW = 0.3 and 0.16, respectively). The percentage of infected samples in total for all infections varied significantly by location and year (0–60%). The possible significance of unaccounted factors is discussed: landscape features and the mode of using forest parks, dispersal of small mammals. The conclusion is made about the usefulness of the multi-model approach in the analysis of the data from epizootiological studies.