Density estimates comparison of a radio-tracked population of yellow-necked mouse Apodemus flavicollis

Density estimation of small mammal populations from capture-mark-recapture (CMR) data has played an important role in many studies of theoretical and applied ecology. The definition of the “effective area”, i.e. the actual area the trapping data should be referred to, is one of the main issues as to accuracy of density estimates. The aim of the present study is therefore the comparison between CMR density estimates based on trapping distance parameters and on ranging movements assessed by radio-tracking. From May to November 2005, a CMR experiment was carried out monthly in a beech forest of the province of Trento, northern Italy (National grid reference: 1651959E 5093546N), in a 18x18 trapping grid with 15m trap spacing (total area= 6,7 ha). Live trapping took place for 5 consecutive nights/month. For each individual rodent, species, sex, breeding condition and body mass were recorded. Ectoparasites (fleas, mites and ticks) were counted and samples of blood and tissue for DNA analysis were collected. Some of the yellow-necked mouse Apodemus flavicollis individuals captured from July to October were marked with radio-collars and then localised by radio-telemetry. Monthly home ranges were calculated as Minimum Convex Polygon (p=95%) and Kernel probability distribution (p=95%) with the program R, package Adehabitat. Mean maximum distance between fixes was computed by means of ArcGIS 9.0. Population size N was estimated from capture histories of each trapping session using closed population models. The program CAPTURE was run to compute all available population estimators: Mo, Mt, Mh, Mh-Chao, Mb, Mbh, Mtb. Density was estimated using several sampling area (A) values: (1) trapping grid area (TGA); (2) TGA plus a boundary strip (W) equal to trap-spacing and half trap-spacing; (3) TGA plus W equal to mean maximum distance moved (MMDM) and half mean maximum distance moved calculated by trapping data; (4) TGA plus W equal to MMDM and half MMDM calculated by radio-tracking data (i.e. mean maximum distance between fixes); (5) TGA plus W equal to the mean “radius” of monthly home ranges (defined as: r= (((P2/4) - 2A)0.5)/2), where A and P are area and perimeter of home ranges). The mean proportion of time spent in the grid by radio-collared individuals was also calculated and used as a correcting factor of the density estimates. For each trapping session, the density estimates referred to the TGA were similar when calculated by different estimators (e.g. in September: DM(o)= 11.99+0.15 in/ha, DM(h)= 13.49+0.72 in/ha, DM(b)= 12.44+0.42 in/ha, DM(h-chao)= 13.19+0.77 in/ha, DM(bh)= 12.44+0.42 in/ha, DM(t)= 11.99+0.07 in/ha). The software CAPTURE selected Mh (Jackknife) and Mbh (generalized removal) estimators, which were used for comparison between density estimates referred to different values of A. Density estimates referred to A=TGA+W with W=MMDM or 1/2 MMDM obtained by trapping data were lower than density estimates referred to A=TGA and generally higher than density estimates referred to A=TGA+W with W=MMDM or 1/2 MMDM obtained by radio-tracking data (RT) or W= mean “radius” of home ranges (e.g in September: with W=MMDM, DM(h)= 7,75+0,41 in/ha, DM(bh)= 7,15+0,24 in/ha; with W=1/2 MMDM, DM(h)= 10,06+0,54 in/ha, DM(bh)= 9,28+0,31 in/ha; with W=MMDM_RT, DM(h)= 4,08+1,13 in/ha, DM(bh)= 3,76+1,08 in/ha; with W=1/2 MMDM_RT, DM(h)= 6,75+1,27 in/ha, DM(bh)= 6,22+1,18 in/ha; with W= r_Kernel, DM(h)= 4,29+0,23 in/ha, DM(bh)= 3,95+0,13 in/ha; with W= r _MCP, DM(h)=7,67+0,41 in/ha, DM(bh)= 7,08+0,24 in/ha). The density values increased when corrected for the proportion of time spent in the grid (e.g. in September: with A=TGA, DM(h)= 13+0,70 in/ha, DM(bh)= 11,99+0,41 in/ha). The movement parameters provided by the radio-tracking study allowed to assess the effective area on the basis of the actual ranging behaviour of the population. The resultant density estimates were generally lower than the ones based only on the trapping data, suggesting the risk to overestimate the population as a consequence of edge effect and the importance of behavioural studies for analysing and interpreting population data.