Effects of Sample Size on Kernel Home Range Estimates

“%) are supported by the least amount of data and are therefore less reliable, whereas inner contours provide more robust representations of space use (Seaman et al., 1999).”

“Our description of fawn cover included only individuals with $30 locations (Seaman et al., 1999) and that survived $2-mo post- capture.”

“…following interpretation are transposed into individual territory boundaries (Bibby et al. 1997), (b) we seek to describe total defended area, not core activity areas, and (c) 96% of territories had fewer than 50 locations and 87% less than 30 locations (Seaman et al. 1999; Barg et al. 2005).”

“We used least-squares cross validation to select the amount of smoothing in fixed-kernel calculations (Seaman et al. 1999).”

“…study area, 2) we developed (Hooge et al. 1999) fixed-kernel estimates (Worton 1989, Seaman and Powell 1996) of the 85% utilization distribution (Seaman et al. 1999) with least squares crossvalidation (Bowman 1985) for the 3 composite or subpopulations, and 3) we drew boundaries (UTM northings)…”

“After applying the above filters, only individuals with 50 telemetry points were included, in order to reduce the effects of small sample sizes on home range estimates [15].”

“Thus, 95% fixed kernel methods (FK95) with least squares cross-validation (Seaman and Powell 1996; Seaman et al. 1999) and 95% local convex hull (LCH95—Getz and Wilmers 2004; Getz et al. 2007) were also used.”

“It is logical that a large number of recaptures will be needed to effectively estimate movement distances [15], an issue widely acknowledged for estimation of home range size from telemetry data [50, 51].”

“We did not use repeated values (coordinate matching) in a sample (monitoring day) because of limitations in the estimation of parameters using LSCV (Seaman et al. 1999), and we did not consider autocorrelations between locations because the removal of points limits the biological relevance of the analysis (De Solla et…”

“For all frogs, more than 30 observations were available as required by the method to exclude sample size effects (Seaman et al. 1999).”

“…of the home range for resident fishers with L30 radiolocations that we collected over L10 months using the 95% isopleth of the utilization distribution generated from the fixed kernel method with the smoothing parameter selected by least-squares crossvalidation (Worton 1989, Seaman et al. 1999).”

“We excluded panthers with ,50 locations to reduce bias in home-range estimation when using the fixed kernel method (Seaman et al. 1999).”

“Despite overestimating home range, MCP was chosen as it is appropriate for day range estimations when sample sizes are small (Seaman et al .”

“So as to reduce autocorrelation while maintaining a sufficient number of relocations ( 50; Seaman 1999), we cropped our dataset to one random relocation every third day before conducting KDEs.”

“…Worton 1989) for each cougar monitored at least 3 months with greater than 30 global positioning system or very-high-frequency relocations, or both (Seaman et al. 1999), using the Fixed Kernel Density Estimator in Hawth’s Tools (Beyer 2004) in ArcMap 9.3 (Environmental Systems Research Institute…”

“Because kernel distributions can be greatly influenced by sample size (Seaman et al. 1999), we maximized our sample size by using all years of location data to estimate the kernel distribution.”

“…for adult snakes with 30 locations and for each population using all individual locations combined as 95% and 50% fixed kernel home range utilization distributions (UD home ranges) and minimum convex polygon home ranges (MCP home ranges; Worton 1989, Seaman and Powell 1996, Seaman et al. 1999).”

“We used the adaptive kernel method (Seaman et al. 1999; Seaman and Powell 1996) to calculate areas, and used the least-squares cross-validation method recommended by Worton (1989) to obtain the smoothing parameter.”

“It is believed that at least 20 (Powell et al., 1997) or 30 (Seaman et al., 1999) locations or fixes are required for dependable fixed kernel estimates of home-range size.”

“We used a minimum of 30 locations (Seaman et al. 1999) and estimated boundaries for the 50, 75, and 95% probability kernels for each home range.”

“A least squares cross-validation was used to determine a smoothing parameter (Rodgers and Carr 1998) with minimum estimated error for fixed-kernel estimates (Seaman et al. 1999).”

“25 location estimates were recorded were used in these analyses (Seaman et al. 1999; Morris et al. 2011).”

“We estimated home ranges using a fixed kernel estimator (Worton 1995) with a 95% utilization distribution and smoothing determined by least-squares cross-validation (Seaman et al. 1999).”

“We adopted the fixed kernel method, as recommended by Seaman et al. (1999).”

“We used Least Squares Cross Validation to estimate the smoothing parameter, and set the probability contour at 80% for all species (Seaman and Powell 1996; Seaman et al. 1999).”

“It is well documented that the number of locations affects the size of the home-range estimated (White and Garrot 1990; Seaman and Powell 1996; Seaman et al. 1999).”

“…we chose the size of the grid such that it was sufficiently large to completely encompass all distributions in the mixture to 4 standard deviations and sufficiently fine that the calculated area of the true UD 5 1.00 (to 2 decimal places; adapted from Seaman and Powell 1996, Seaman et al. 1999).”

“The fixed-kernel method is preferred because it is a measurement of the intensity of use, excluding areas used minimally (White and Garrott 1990, Seaman et al. 1999).”

“We calculated Key deer annual ranges (95% probability area) and core areas (50% probability area) using a fixed-kernel home-range estimator (Worton 1989, Seaman et al. 1998, Seaman et al. 1999) with the animal movement extension in ArcView (Version 2.2; Hooge & Eichenlaub 1999).”

“We restricted our sample of marked juveniles to those individuals with 30 detections for subsequent analyses because simulation research has indicated 30 is the minimum sample size required for stable home range estimates using kernel estimators (Seaman et al. 1999).”

“Fixed-kernel estimators allow the identification of disjunct areas of activity (Seaman et al. 1999), a particularly important consideration in intensively managed landscapes.”

“We used fi xed-kernel estimation with least-squares cross-validation (see Worton 1995, Seaman et al. 1999) within ANIMAL MOVEMENT, version 2.0 (Hooge and Eichenlaub 1997) to calculate 95% probability contours of home-range boundaries.”

“(2015), for example, calculated the KDE for individuals seen at least 30 times (Seaman et al., 1999), although other studies indicated that 100–300 observations per individual were necessary to obtain highly precise representation of space use (Girard et al.”

“Fixed kernels Bond et al. Spotted Owls and Fire 3 are less biased than adaptive kernels at outer contour levels and have better surface overlap when compared to the true distribution (Seaman et al. 1999).”

“We used Least Squares Cross Validation (LSCV) to select the smoothing parameter (Seaman & Powell 1996).”

“…and 2 nonwinter) we used ArcView 3.3 Home Range Extension (HRE—Rogers and Carr 1998) with a least-squares cross-validation smoothing parameter (Seaman et al. 1999; Seaman and Powell 1996; Worton 1995) to estimate 50% fixed kernel seasonal core areas and 85% fixed kernel seasonal ranges…”

“A low number of observation days (Topping et al., 2005), fewer than 30-50 localizations (Seaman and Powell, 1996, Seaman et al., 1999) and spatial autocorrelation between observations (Swihart and Slade, 1985) tend to lead to the home range size being underestimated.”

“We quantified home ranges in ArcView 3.3 using the Animal Movements extension (Hooge et al. 2000) by implementing a 99% fixed-kernel estimator (Seaman et al. 1999).”

“%) using a fixed-kernel home range estimator (Worton 1989; Seaman et al. 1998, 1999) following methods identical to Lopez et al. (2005) to allow direct comparison.”

“Recommendations on the number of data points required to accurately estimate home ranges vary greatly, with a minimum of 10 (Börger et al., 2006), 30–50 (Seaman et al., 1999) and 100 (Girard et al., 2002) locations suggested.”

“Turtles were included in home range analysis if they had at least 30 locations during the active season (Sep - Apr) (Plummer and Mills 2000; Donaldson and Echternacht 2005), as home ranges estimates are sensitive to sample size (Seaman et al. 1999).”

“A home range built with fewer points tends to be larger and oversmoothed, and the number of points required to build a precise home range varies between systems (B€orger et al. 2006), though 30–50 are often necessary (Seaman et al. 1999, Millspaugh and Marzluff 2001).”

“Only locations recorded from individuals with at least 30 locations were included in home range analyses, the minimum number recommended for unbiased estimates of HR size (Kenward, 2001; Seaman et al., 1999).”

“Months were chosen as the time unit as they were the smallest standard temporal unit for which we could generate home range estimates, and only those elephant/month combinations in which at least 50 location points were obtained were included in the analysis (Seaman et al. 1999).”

“We attempted to obtain $30 locations per individual male (Laundre and Keller 1984, Bibby and Burgess 1992, Seaman et al. 1999).”

“We calculated the smoothing parameter (Kenward 2001) following the least squares cross validation method (LSCV) to minimise the mean integrated square error (Seaman et al. 1999).”

“We used a minimum of 30 locations after removing outliers (≥5 km from territory; Fuller, 1989) and defined the annual pack territory as the 95% volume isopleth boundary (Seaman et al., 1999; Uboni, Vucetich, Stahler, & Smith, 2015).”

“All KHR estimates were estimated using a sample size of at least 30 locations as recommended by Seaman et al. (1999).”

“The least squares cross-validation bandwidth estimator and P30 locations per individual have been recommended for home range studies (Seaman et al., 1999), but the likelihood cross-validation bandwidth estimator produces better fit and less variability with moderate to small sample sizes (i.e., 650…”

“We estimated home-range (HR), corresponding to the 95% utilization distribution (UD) [48], using data from radiocollared wild boar and the fixed-kernel function in the ADEHABITAT package [50] in R version 3.”

“The minimum number of relocations per individual employed to estimate seasonal home ranges was 25 according to previous studies [61,63].”

“…the kernel overlap function in the AdehabitatHR package (Calenge 2006) in program R (R Version 2.14.1; R Development Core Team 2011) using 500 randomly selected GPS locations and a smoothing parameter determined by reference bandwidths for each animal (Seaman and Powell 1996; Seaman et al. 1999).”

“There is no established minimum number of points necessary for calculating a MCP (Seaman et al., 1999), but the estimated area may stabilize with sample sizes greater than 40 (Arthur & Schwartz, 1999).”

“The 50% kernel home range was chosen as the study area for sampling because it is considered as a core center of animal activity; furthermore, the biological signifi cance of the 95% kernel home range can be unreliable (Dickson and Beier 2002, Seaman et al. 1999).”

“We used least-squares cross validation (LSCV) to determine smoothing factors in the fixed-kernel estimation procedure because LSCV is less biased and performs better than other methods, especially with sample sizes .50 (Seaman and Powell 1996, Seaman et al. 1999).”

“Estimates were based on 75 percent kernel contours to define central portions of a home range and exclude ‘‘occasional sallies’’ (Burt, 1943; Seaman et al., 1999).”

“We determined the value of using the least-squares cross-validation method, which has been demonstrated to be effective for ecological applications and is commonly used for the delineation of home range (Seaman et al. 1999).”

“The smoothing parameter, h, is usually calculated with the least-square crossvalidation technique (Seaman et al., 1999), but Hemson et al.”

“This number of locations is adequate for fixed kernel home ranges (Seaman et al. 1999).”

“The fixed-kernel method generally appears to have lower bias and better surface fit than the adaptive kernel (Seaman et al. 1999) and is more reliable when estimating the outer contours and centers of activity of the home range (Kernohan et al. 2001).”

“We calculated 100% minimum convex polygon, 50% kernel density, and 95% kernel density home ranges for turtles with >40 locations (Seaman et al. 1999).”

“However, as an overlap measure, Bhattacharyya’s affinity needs to be computed across two reasonably large distributions (30–50 observations, Seaman et al. 1999), which means that it would not be appropriate to calculate it separately for each of our speakers, since each has only eight tokens of…”

“We used only data from individuals with more than 25 locations due to minimum sample size limitations of KDE (Seaman et al. 1999).”

“The number of points used to generate annual and seasonal utilization distributions ranged from 381 to 3,033, providing robust estimates of kernel density [36].”