Factors affecting habitat occupancy and densities of the Red-legged partridge in introduction areas of north-western Italy

Introduction. The Italian range of the red-legged partridge changed markedly in the time together with population abundance because of habitat changes, overhunting, and releases of hand-reared birds. Currently self-sustaining populations are recorded in plain areas outside the historical range that possibly originated by natural colonization and/or introductions carried out with partridges from game farms. . Historic range of red legged partridge in Italy extended mainly along the northern Apennine chain from the Southern Alps to the west, to the province of Modena in Emilia-Romagna region to the east, both on the south and north slopes. In central Italy the presence of the species was recorded in Tuscany, Umbria and Marche regions, approximately down to the Mount Vettore (29 provinces and 8 regions). In the years between 1980 and 1990 the red legged partridge was present in 18 provinces and 5 regions. The populations in central Italy disappeared with the exception of the Elba Island, and a strong decline and local extinctions were recorded for the population in northern Apennines. On the contrary massive releases of hand-reared partridges from the game farm of the region Emilia-Romagna caused an increase of the eastern part of the range that reached the provinces of Bologna, Ravenna, and Forlì-Cesena [1]. At the end of the past century, the hunting districts established by the National Law 157/92 and private estates carried out releases of red legged partridges from game farms outside the historical range causing an important range expansion, in particular in the Po plain and in central Italy (39 provinces and 9 regions). Methods. To collect information on the factors affecting the current distribution and abundance of red-legged partridge in new-colonized areas we carried out call spring counts in two (2015) and three (2016) protected areas located in the dry crop Po plain in the Piacenza province. The three study areas covered 9.2 (Borgonovo study area), 41.1 (Trebbia s. a.), and 35.6 km2 (Nure s. a.). They were mainly cultivated; the main crops were winter cereals (18.9%), maize (11.9%), vegetables (7.6%), alfalfa (7.0%) and hay fields (5.0%). Settlements occupied 10.7% of the study areas. Two main streams cross Trebbia and Nure study areas with pebbly shores occupying in total 20.0% of the surface. We carried out call counts with tape recorded call from random listening points (34 points in 2015 and 92 in 2016) in April and May from the sunrise until 9.30 a.m., recording the calling males of partridges and pheasants [2]. The densities were calculated using a 300 m fixed radius from the points. Moreover in each buffer of 300 m we calculated the percentages of the main land use classes and the main landscape metrics. We analyzed habitat selection by red-legged partridges by means of Resource Selection Probability Functions (RSPF) formulated by Binary Logistic Regression Analyses (BLRA) of presence points vs. an equal number of random points in which we calculated the same habitat variables (use vs. availability approach, [3]). Moreover we examined the effect of habitat variables on the partridge density by Multiple Regression Analyses (MRA) carried out considering only the presence points. For both analyses we selected the best models by the correct Akaike Information Criterion (AICc) [4]; the performance of logistic model was tested by ROC analysis and that of multiple regression model by the correlation between the predicted and observed values. We verified the multicollinearity of predictor variables by the Variance Inflation Factor (VIF), the normality of residual distribution by the Kolmogorov-Smirnov test, and the residual autocorrelation by Durbin-Watson test [5]. Results. The average density varied from 2.1 (SE=0.42) to 3.2 (SE=0.39) in 2015 and from 2.9 (SE=0.63) to 3.7 (SE=0.88) in 2016 without significant differences between years and study areas. For 15 habitat variables we found significant differences between presence and control points (Mann-Whitney U test, P<0.05); in particular winter cereals, pebbly shores, hedgerow density, Shannon diversity index, the patch number, edge density, and mean perimeter area ratio had greater values in presence points. Three habitat variables (winter cereals, pebbly shores and habitat diversity) entered the best logistic model with a positive effect on the presence probability of red-legged partridge. The model explained 37.0% of the variance and correctly classified 71.2% of the original cases (presence: 77.8%; controls: 64.9%); ROC analysis showed a good performance of the model (AUC=0.82, SE=0.048, P<0.0001) (Table 1). We found significant positive correlations between the partridge density and percentage of unpaved roads (study areas pooled r= 0.442, n=39, P=0.005), hedgerows (Trebbia study area r=0.661, n=16, P=0.005), hay fields (Trebbia s. a. r=0.601, 268 n=16, P=0.013), and patch size SD (Nure s. a. r=0.648, n=13, P=0.017).Considering the study areas pooled the best regression model explained 31.8% of the density variance by the inclusion of four habitat variables of which the percentage of unpaved roads, and the edge density with significant positive effects, and the patch number with negative ones (Table 2); predicted and observed density values were highly correlated (r=0,624; n=39; P