Space Use and Pesticide Exposure Risk of Male Burrowing Owls in an Agricultural Landscape Author(S): Jennifer A

Home , Athene (bird)

Space Use and Pesticide Exposure Risk of Male Burrowing Owls in an Agricultural Landscape Author(s): Jennifer A. Gervais, Daniel K. Rosenberg and Robert G. Anthony Source: The Journal of Wildlife Management, Vol. 67, No. 1 (Jan., 2003), pp. 155-164 Published by: Wiley on behalf of the Wildlife Society Stable URL: http://www.jstor.org/stable/3803071 . Accessed: 09/06/2014 12:44

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Wiley and Wildlife Society are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Wildlife Management.

http://www.jstor.org

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions SPACE USE AND PESTICIDEEXPOSURE RISK OF MALE BURROWINGOWLS IN AN AGRICULTURALLANDSCAPE

JENNIFERA. GERVAIS,'1,2Oregon Cooperative Fish and WildlifeResearch Unit,Department of Fisheriesand Wildlife,Oregon State University,Corvallis, OR 97331, USA DANIELK. ROSENBERG,Oregon Cooperative Fish and WildlifeResearch Unit,Department of Fisheriesand Wildlife,Oregon State University,Corvallis, OR 97331, USA ROBERTG. ANTHONY,U.S. GeologicalSurvey, Oregon Cooperative Fish and WildlifeResearch Unit,Department of Fisheries and Wildlife,Oregon State University,Corvallis, OR 97331, USA

Abstract:We estimatedhome-range size and habitatselection in a populationof burrowingowls (Athenecunicular- ia) livingwithin an agriculturallandscape in the CentralValley of California,USA, in 1998and 1999.We modeled home-rangesize and habitatselection of breedingmale owls (n = 33) as a function of biologicaland physicalfactors. Biologicalfactors included number of young fledged and diet, and physicalfactors included cover-typecom- positionaround the nest. We also examinedpatterns of space use in conjunctionwith agriculturalpesticide appli- cationrecords for evidenceof secondarypoisoning risk to the owls.Owl home rangesvaried in size within (but not between) years,and not in conjunctionwith any of the biologicalfactors we measured.Foraging versus random locationswere differentiatedmost stronglyby distancefrom the nest, with 80%of nocturnalforaging observations fallingwithin 600 m of the nest burrow.No single cover typewas selected when distanceto nest wasalso included in the model. Owlsdid use agriculturalfields recentlytreated with pesticides,although we did not find evidence of owls selectivelyforaging in these fields. JOURNALOF WILDLIFEMANAGEMENT 67(1):155-164 Keywords: agricultural habitat, Athene cunicularia, burrowing owls, California, habitat selection, home range,pesticide exposure,radiotelemetry.

Burrowing owls were once widespread and configuration of cover types and distances to the common throughout western North America, nest burrow. These factors may impact patterns but some populations have declined in recent of habitat selection (Rosenberg and McKelvey years (Haug et al. 1993, Sheffield 1997). There 1999). Nearly all other studies that reported for- has been much speculation regardingpotential aging observations for burrowing owls were diur- causes of these declines (James and Fox 1987, nal, when the owls remained close to the nest and Haug et al. 1993,Jamesand Espie 1997,Desmond appeared to prey primarily on invertebrates and Savidge 1999), and habitat destructionand (Coulombe 1971, Thomsen 1971, Martin 1973, degradationare major concerns. Many burrow- Thompson and Anderson 1988, Green et al. 1993). ing owl populations persist in areas of urban Agricultural environments can support very developmentor agriculturalproduction (DeSante high densities of burrowing owls (Rosenberg and et al. 1997,Rosenberg and Haleyin press). Iden- Haley in press). These also may pose threats to tifying components of these altered environ- owl populations from pesticide exposure (James ments that are most important to the owls and and Fox 1987, Gervais et al. 2000), destruction of those that pose the greatestthreats will be useful nest burrows by farm equipment, seasonal food in conservationplanning. scarcity exacerbated by farming practices, or Despite the species'frequent proximity to areas extermination of the fossorial mammals that dig inhabitedby people and relativeresistance to dis- the burrows used by the owls (Desmond et al. turbance, burrowing owl habitat selection and 2000). Given that large expanses of the burrow- space-usepatterns remain little-studied.Work by ing owl's range are dominated by agriculture, Haug and Oliphant (1990) in Saskatchewanindi- understanding how the owls survive in these envi- catedthat male owlsselected grass-forb vegetation ronments is necessary for conservation strategies. cover for foraging during the breeding season. We explored space use and habitat selection by However,their analysesdid not consider spatial a resident population of burrowing owls living in an area of intensive row-crop agriculture. We postulated that space use would be linked to diet and I E-mail: [email protected] risk of Burrowing owls pri- 2 Presentaddress: Department of Forest,Range, and pesticide exposure. consume rodents and Wildlife Sciences, Utah State University,Logan, UT marily (Green Anthony 84332-5230,USA. 1989, Silva et al. 1995), and owls have shown both 155

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions 156 BURROWINGOWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al.J.Wildl. Manage. 67(1):2003

functional and numerical responses to increasing ends of runways. Owls nested in burrows excavat- vole (Microtusspp.) populations (Silva et al. 1995). ed by California ground squirrels (Spermophilus Grass is the dominant food of the California vole beecheyii),artificial burrows, cable housings, cul- (M. californicus;Gill 1977) and is the most stable verts, and holes excavated under concrete slabs. cover type for rodents in agricultural systems. The population was composed primarily of year- In addition, we frequently observed burrowing round residents (J. A. Gervais and D. K. Rosen- owls foraging along the edges of roads and berg, Oregon State University,unpublished data). drainage ditches. We hypothesized that owls should select grass and edge cover types while METHODS Also, home with amounts foraging. ranges greater Field Methods of grass and edge cover near the nest should be smaller than home ranges of owls nesting adja- We captured adult male burrowing owls during cent to cropland. the April-June breeding season in 1998 and 1999 Because burrowing owls are central-place for- (Gervais 2002). Nesting males were fitted with agers when nesting, distance from the nest should elastic radiocollars (mass = 3.6-4.5g; Model PD-2C, also influence habitat selection (Rosenberg and Holohil Systems Ltd., Ontario, Canada) represent- McKelvey 1999). We postulated that owls foraging ing 2.3-2.9% of adult breeding mass. Batteries with primarily in crop fields should have greater home- a 14-week life expectancy were used in 1998. range sizes and lower reproductive success due to Heavier 24-week batteries were used in 1999. We lower rodent densities and greater pesticide collected location data from 15 May-1 September exposure risk. Finally, based on earlier findings of 1998 and 1 May-15 September 1999. Females chlorpyrifos, a broad spectrum organophosphate were not radiomarked because they tend to pesticide, in footwash samples (Gervais et al. remain near the nest burrow through the early 2000), we predicted that these owls would select fledgling period. During this stage in the cycle, fields recently sprayed with pesticides in response males do most of the foraging (Haug et al. 1993). to the availability of dead and dying prey. We used a dual antenna receiving system with a We addressed these questions by radiomarking null combiner (Telonics, Inc., Mesa, Arizona, USA) and locating adult male owls in an agricultural mounted in the back of a pickup truck. In 1998, environment during the breeding season to cre- we used H antennas for the array,and obtained a ate an index of minimum habitat requirements maximum reception distance of 0.8 km. In 1999, and patterns of habitat selection within the home we used 4-element yagi antennas, which increased range. We also described diet, estimated numbers the reception range to 1.0 km. The antenna arrays of owlets surviving to fledging, and obtained doc- were approximately 3.5 m from the ground. umentation on pesticide spray applications with- Observers obtained sequential bearings at pre- in the study area. determined stations along a gridwork of farm roads that covered the area. All STUDYAREA study bearings used were taken <5 min apart. Frequent owl The population of owls we examined resided movements while foraging made obtaining more on an 80 km2 section in the center of Naval Air than 2 bearings on a single owl location difficult. Station (NAS) Lemoore, located 50 km southwest Because burrowing owls appear to move fre- of Fresno, California, USA, latitude 36018'N, quently while foraging, we recorded signal quali- 119056'W longitude. Naval Air Station Lemoore is ty as well as the time, station location and bearing in the center of the San Joaquin Valley, an area of angle. Signals were classified as either 1, strong intense agriculture (Griggs 1992). Major crops with obvious null; 2, strongest direction of a sig- include cotton, alfalfa, tomatoes, and corn (Cali- nal without a null (the bird was either moving, fornia Department of Pesticide Regulation 1998, underground, or vegetation and microtopogra- 1999). The Air Operations area at NAS Lemoore phy were interfering with signal transmission); or is surrounded by agricultural fields in active pro- 3, only a few good signal beats were detected. duction. Burrowing owls nested along runway This last scenario frequently occurred if owls easements, within the Air Operations taxiways were foraging in ditches or farm field furrows. and ramp systems, and in unmowed grassy areas Even limited topographic relief was enough to surrounded by agricultural fields. These patches cause substantial signal interference. Efforts were ranged from strips 20 m wide, extending the made to search areas >1 km from the nest site to length of runways, to fields of 45-179 ha at the avoid biasing observations near the nest. Loca-

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions J. Wildl.Manage. 67(1):2003 BURROWING OWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al. 157 tion attempts on the same owl were made 215 min 10 data sets based on the real data in which each apart. Each owl was tracked at least 2 nights/week, location estimate was drawn randomly from the 8 and several locations were obtained per night. alternative options for that location attempt. We tracked owls from dusk to 0300. These 10 data sets were then submitted to the In both years, we quantified radiotelemetry same analysis as the real estimated locations. error by placing radios in known locations. We Home-Range Size.-Fixed-kernel estimates of then estimated those locations using observers home range were calculated using least squares who were not aware of radio locations. Radios cross validation (Worton 1995, Seaman and Pow- were placed to mimic actual owl positions while ell 1996, Seaman et al. 1999). We did not calcu- perching or foraging, although the test radios late home ranges with kernel estimators for owls remained in fixed locations. with <26 locations due to instability of kernel esti- All nests were visited and pellets were collected mators with small sample sizes (Seaman et al. weekly or biweekly and prey remains noted. We 1999). Minimum convex polygons were estimated observed all radiomarked owl nests that were using all locations rather than 95% of observa- accessible using a standardized protocol to esti- tions because peripheral locations for central- mate productivity, which we defined as the maxi- place foragers are likely to be underestimated. mum number of owlets seen shortly before they There is more area to search at the periphery of were able to fly. Owlets from the same brood the range, and owls are therefore more likely to rarely scattered among several burrows after be detected when near the nest. This bias will be emergence from the natal burrow because of the particularly severe when a radiotagged animal paucity of available burrows. We recorded the cannot be detected over its entire range from a presence of invertebrate taxa in the pellets to single receiving location, as was true in this study. order or family, and we identified vertebrates to Habitat composition was estimated by determin- genus or species. We estimated individual ing the percentage of the fixed kernel home range rodents on the basis of dentary bone counts. composed of each of the major cover types. We used kernel estimates for this analysis because the DataAnalyses kernel estimator should not include large areas Telemetry locations were estimated using pro- of unused habitat relative to the MCP estimate. gram LOCATE II (version 1.5, Truro, Nova Sco- Factors affecting estimated 95% kernel home- tia, Canada). We removed locations from the range size were examined using multiple regres- data set that fell outside the estimated maximum sion and Akaike's Information Criterion (AIC) detection distance from stations m in 1998; for small sizes Burnham (800 adjusted sample (AICc; 1,000 m in 1999). We used program KERNELHR and Anderson 1998). We considered a suite of a (Seaman et al. 1998) to compute a 95% fixed-ker- priori models representing various potential nel home-range estimate, and program TELEM hypotheses that might explain patterns in the (version 1.0, U.S. Forest Service) to compute a data (Table 1; Franklin et al. 2001). Habitat fac- 100% minimum convex polygon (MCP) estimate tors included the amount of edge (road and for each owl (Jennrich and Turner 1969). ditch) and grass cover within 400 m of the nest, Location Estimation Error.-We estimated the which was the mean distance owls were detected maximum error angle from test radios whose loca- away from the nest over all owls and both years. tions led to some topographical signal interfer- Neighboring nests were defined as active nests ence, which prevented the null-peak signal recep- within 400 m of the focal nest. tion pattern. Displacement distance was calculated The relative importance of diet, numbers of as the tangent of the maximum error angle multi- fledglings raised, numbers of neighboring owl plied by the mean distance of the radio location pairs, and cover characteristics near the nest in to the receiving stations used in that location esti- explaining home-range size were evaluated using mate. This distance (Burnham and Anderson 1998). point-specific displacement AICcweights was then used to offset each estimated location sys- Habitat Selection.-We examined burrowing owl tematically on the cardinal directions and also habitat selection using estimated locations of NW, NE, SW, and SE. The square array of alter- radiomarked males. We defined the habitat avail- native points should encompass the extremes of able to each radiomarked owl as the area within habitat misclassification possibilities, particularly the circle, centered on the nest burrow, whose in our study landscape with its regular gridwork radius was the maximum distance the owl was of agricultural fields and runways. We generated detected from the nest (Rosenberg and McKelvey

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions 158 BURROWINGOWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al.J.Wildl. Manage. 67(1):2003

Table1. Modelsexploring the relationshipof 95%kernel home- able cover types for each owl (Table 2; Hosmer size and various range explanatoryvariables for 33 male bur- and Lemeshow Model factors rowingowls at Naval Air Station, Lemoore,California, USA, 1989). included 1998-1999. LowerAkaike's Information Criterion (AlCc) values distance to nest either as a log function or a third- indicaterelatively better models and weights are the propor- order function because owl use of an tionallikelihood of the models. polynomial area declines rapidly with distance from the nest. Model r2 AICc Weights This probably is due to the energy constraints of grassa + edgeb + rodentsc bringing back 1 prey item at a time to the nest. + chicksd+ nestse 0.104 14.635 0.0003 Log distance from the nest to edge cover account- + grass edge + rodents 0.054 7.622 0.0100 ed for the of owls on the chicks + nests 0.074 3.548 0.0764 difficulty locating directly roads or ditches to as these grass + edge 0.035 4.708 0.0428 adjacent them, are edge 0.023 2.015 0.1644 very narrow cover types. Most location estimates rodents+ chicks 0.031 4.801 0.0408 are likely to be near, but not on, edge features if numberof locations 0.041 1.476 0.2152 edge cover is utilized. Even small location error interceptonly (no effects model) 0.000 0.000 0.4502 can result in habitat misclassification for narrow, a Percentgrass within400 m of nest. linear cover types. However, owl use of these fea- b Amountof edge habitatwithin 400 m of nest. tures would also be indicated by the model para- c Mean numberof rodents/pellet. d Numberof chicks raisedto meter of distance to edge cover. Because owls e fledging. Numberof active nests within400 m. from different regions of the study area had somewhat different landscape compositions near their nests, we divided cover types into 3 general 1999). We then selected 1,000 random locations categories: GRASS, CROPLAND, and OTHER. within each circle. A cover type was assigned to GRASS included all runway easements, grass- each random location and owl location using land patches, and fallow fields. CROPLAND in- ARCVIEW (Environmental Systems Research cluded all fields in active production, including Institute, Redlands, California, USA, version 3.1). alfalfa hay. The OTHER category incorporated Distances to the nearest road and runway were ditches, industrial areas, ramps, taxiways, run- estimated for all locations using ARCINFO (Envi- ways, parking lots, and wetlands. Fallow fields ronmental Systems Research Institute, Redlands, were categorized as GRASS cover because they California, USA, version 7.2.1). typically were not disturbed by tilling or pesticide we evaluated a set of a the season. Using AICc, priori logis- applications during growing Despite tic regression models comparing used versus avail- its structure, permanence relative to other crops

Table2. Comparisonof mean Akaike'sInformation Criterion (AICc) values for habitatselection models among male burrowing owls at NavalAir Station, Lenmoore, California, USA, 1998-1999. LowerAAIC, values indicatebetter model fit to the data.Ten models were originallyevaluated. An additional2 models were evaluatedafter the initialanalysis and the AAICcvalues recalcu- lated over the 12 models.

Model SE Min Max r2a SE AAICc dnest + dnest2+ dnest3b+ ogdedge + cover typec 5.582 0.900 0 19.176 0.340 0.023 dnest + dnest2 + dnest3 + cover type 6.660 0.893 0 20.512 0.315 0.024 dnest + dnest2 + dnest3 7.980 1.805 0 41.448 0.308 0.025 logdnestd+ logdedgee + cover type 13.639 2.703 0 78.303 0.302 0.024 distance + cover type + distance*coverf 13.930 2.510 0 54.466 0.301 0.024 logdnest+ cover type 15.884 2.862 0 77.179 0.287 0.025 logdnest+ logdedge 16.751 3.216 0 76.566 0.279 0.025 logdnest 18.991 3.341 0 76.309 0.264 0.026 distanceg 19.309 3.251 0 63.620 0.265 0.025 cover type 88.988 10.863 5.010 272.669 0.093 0.011 logdedge 113.260 13.222 16.725 291.782 0.025 0.004 interceptonly (no effects model) 118.562 13.382 21.399 293.835

a Maximumrescaled generalizedr2. b Polynomialdistance function for distance to nest. C Definedas CROPLAND,GRASS, or OTHER. d Log distance to nest. e Log distanceto nearest habitatedge. f Interactionterm, a posteriorimodel. 9 Lineardistance to nest, a posteriorimodel.

(it is usually maintained for several years) and PesticideExposure Risk.-We were interested in potential prey populations, alfalfa was categorized location-specific exposure to pesticides as we had as CROPLAND due to regular cuttings, irrigation, some evidence that this occurred in the owls (Ger- pesticide applications, and other field operations. vais et al. 2000). Field-specific agriculturalchemical Overall, we modeled owl habitat selection as a use data for NAS Lemoore during the study peri- function of cover types, distance to edge, and dis- od were obtained from the California Department tance to nest both separately and in combination. of Pesticide Regulation (1998, 1999). The data After examining the original set of models, we con- were examined in conjunction with owl locations sidered 2 additional models that included linear to determine whether use of CROPLAND cover distance to nest and distance by habitat interaction. was exposing owls to recently applied pesticides. Eight owls did not have estimated locations in We noted dates and locations of all applications either the CROPLAND or OTHER cover types. of pesticides at NAS Lemoore, which had the To avoid quasiseparation of the random versus potential to create a pulse of dead or dying prey actual locations in the logistic regression analysis that might attract owls. Pesticides not necessarily (Allison 1999), we added a single fictional loca- toxic to birds (such as pyrethroids) were includ- tion to the unused cover type in each of those owl ed to better examine the general pattern of owl data sets. For these additional locations, distances response to pesticide application events that to features such as nest or roads were computed could provide a sudden food pulse. The risk zone as the mean distance over all locations and the was defined as a sprayed field 0-3 days post-spray. cover type was classified as the missing category. Three days post-spray were used as the risk peri- This approach allowed us to use all owl data on od because the pulse event of suddenly available the same set of models, rather than restricting prey is unlikely to last more than 3 days after our analyses to only a subset of the owls sampled. application of a pesticide. Most currently used To estimate the precision of our models, we cal- pesticides break down rapidly under environ- culated the maximum rescaled generalized r2 mental conditions (Kamrin 1997), and most prey value. This statistic is based on the likelihood ratio likely were exposed to and killed by the chemi- chi-square and is scaled to account for the dis- cals in that time. crete dependent variable being <1 (Nagelkerke Each owl location was classified as either in or 1991, Allison 1999). not in a risk zone. Available habitat was defined We examined the strength of selection for as fields used by the owls anywhere on the station GRASS by foraging owls as a function of its avail- during the days when at least 1 field was classified ability and as a function of avoidance of CROP- as a risk zone. This prevented the inclusion of LAND. To do this, we estimated the regression fields that may have had cover characteristics pre- coefficient of the parameter estimates for GRASS cluding foraging and were not used during the and CROPLAND, with the amount of grass cover risk period. We compared the use of risk-zone within 400 m of the nest. We used to fields to the use of all fields on the station at that AICcweights obtain model-averaged parameter estimates (Burn- time using odds ratios. ham and Anderson 1998) for the GRASS cover type from the logistic regression analyses. The RESULTS the selection of GRASS cover the greater type by Size owls, the larger the parameter estimate for Home-Range Relationships GRASS. The larger the parameter estimate, the We tracked 11 adult male owls in 1998 and 22 in better the distinguishing power of GRASSbetween 1999. Two individual owls were tracked in both random versus actual owl foraging locations. years. The total was 31 individual birds and 33 We examined habitat use ofjuvenile burrowing samples. Because fledging success, rodent con- owls before they dispersed from their natal nest. sumption, and habitat composition varied between We summed the number of juvenile locations by years, we retained each owl in both years of the cover type over all individuals to examine trends analyses. Twenty-eight of the samples had 26 loca- between years and among cover types. We did not tions, and 24 had 230 locations. estimate home ranges for juveniles nor did we Home ranges varied substantially among indi- define "available"habitat. These concepts are not viduals, but not between years (Table 3). Maxi- appropriate for young owls whose movements mum distance traveled from the nest was similar away from their natal area increase as they begin between years (? = 1,278 m, 95% CI = 855-1,697 m, = post-fledging dispersal (King and Belthoff 2001). n = 11 vs. ? 1,337 m, 95% CI = 1,033-1,641 m, n=

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions 160 BURROWINGOWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al.J.Wildl. Manage. 67(1):2003

Table3. Meanhome-range areas of burrowingowls at NavalAir Station, Lemoore, California, USA, 1998-1999, using95% fixed-ker- nal and 100%minimum convex polygon (MCP) estimates. Kernel home rangeswere notcalculated for owls with<26 observations. 1998 1999 95% kernel 100%MCP 95% kernel 100%MCP k 95%CI 95% CI 95% CI 95% CI , Size (ha) 139 1-277 177 52-302 98 64-132 189 122-256 Locations/owl 38.8 32.6-45.0 35.6 28.7-42.5 61.9 51.6-72.2 55.9 44.7-67.1 No. owls 9 11 19 22

22), as was the mean distance traveled from the Table 4. Percent habitatcomposition of 95% kernel home- nest = 378 95% CI = 255-501 n = 11 vs. i range estimates of burrowingowls at Naval Air Station, (X m, m, Lemoore,California, USA, 1998-1999. = 409 m, 95% CI = 280-538 m, n = 22). Percentages of GRASS and CROPLAND in owl CROPLANDa GRASSb OTHERc w home ranges were similar among owls and n SE SE SE between years (Table 4). Individuals with high 1998 11 38.2 5.6 49.1 4.3 12.8 3.2 1999 22 32.4 5.1 56.6 3.9 11.0 2.9 percentages of the cover type OTHER nested within the Air area of the station. Operations a Includedall regularlytilled fields, includingalfalfa hay. b Home-range size was not well explained by cover- Includedfallow fields as well as runwayeasements and unmowed areas. type composition, number of owlets raised to fledg- grass C Includedrunways, roads, drainage ditches, wetlands, etc. ing, number of neighboring nests, nor was it related to the quantity of rodents in pellets (Table 1). The null models (intercept only and number of locations/owl) were among those with the lowest tal factors considered. However, owl foraging value. This further that the bio- locations were described AAICc suggested frequently adequately logical variables we measured are not related to by only distance to nest modeled as a third-order home-range size. Also, none of the variables we polynomial function, and polynomial distance considered had a overall relative likelihood and cover based on values. large, type together, AAICc as indicated summed (Table 5). This that distance to the nest was a by AICcweights suggests primary factor in male owl foraging-site selection. HabitatSelection Selection intensity for GRASS did not seem Habitat-selection patterns varied widely among related to its availability, nor to the proximity of individual owls with no clear "best" model evi- CROPLAND. Parameter estimates of each cover dent (Table 2). Habitat selection was equally well type were highly unstable between models that explained by distance to the nest and cover-type included habitat only and those that also includ- composition (Table 6). However, distance-only ed a distance function, with the SE of estimates models had greater explanatory power than the exceeding the parameter estimates themselves. habitat-only model as indicated by the generalized r2 value (Table 2). Distance to nest was of JuvenileOwl Habitat Use great importance in distinguishing foraging loca- Juveniles were most likely to be found in tions from random ones. Further, 80% of all for- GRASS cover, but this pattern differed between aging observations fell within 600 m of the nest years with greater numbers of locations occur- (Fig. 1). In contrast, distance to edges of roads and irrigation ditches did not explain owl forag- locations well to other factors ing compared Table5. SummedAkaike's Information Criterion (AICc) weights (Table 2). over all models foreach parameterin home-rangesize analy- Some individual owl were sis of male burrowingowls, NavalAir Station, Lemoore, Cali- foraging patterns fornia,USA, 1998-1999. best explained by the 2 a posteriori models with linear distance and distance by habitat interac- Parameters tion. For most of the owls, these models were not Grass Nest Chicks Edge Rodents competitive with the first 3 models we considered 95% kernela 0.0530 0.0767 0.1175 0.2174 0.0511 0.0832 0.2620 (Table 2). The foraging locations of most indi- MCPb 0.2863 0.3265 0.0486 vidual owls were best described the by global a 95%fixed kernelhome-range estimate. model containing all biological and environmen- b 100%minimum convex polygonhome-range estimate.

Table6. Meanweights for variables summed over all models in Table7. Mean percentage of observationsof juvenileburrow- a habitatselection analysis of male burrowingowls at NavalAir ing owls in differenthabitat types, NavalAir Station, Lemoore, Station,Lemoore, California, USA. Values are averaged over California,USA. 11 adultmale owls in 1998 and 22 adultmales in 1999. No.locations Cropa Grassb Otherc Variable x-b SE Min Max Year n W SE i SE k SE SE , dnest + dnest2 + dnest3a 0.6868 0.0683 0.0058 1.0000 1998 30 11.5 1.5 11.8 2.3 78.5 3.0 9.8 2.6 COVERtype 0.5548 0.0603 0.0775 1.0000 1999 31 14.2 1.9 35.8 8.1 57.5 5.3 13.3 4.1 log distance to edge 0.4271 0.0600 0.0221 0.9997 log distance to nest 0.3114 0.0680 0.0000 0.9942 a Cropsinclude all regularlytilled fields, including alfalfa hay. b Grass includedfallow fields as well as runwayeasements a Polynomialfunction, distance to nest. and unmowedgrass areas. b The greaterthe weight,the greaterthe contributionof that c Other included runways, roads, drainage ditches, wet- variableto the model'sfit to the data. lands, etc.

LocationEstimation Error ring in CROPLAND in 1999 (Table 7). We frequently observed recently fledged juveniles for- Just over half of all location estimates were aging along farm roads and edges of fields. made with at least 1 bearing whose signal quality was <1, that there was no Riskand Habitat meaning null-peak sig- PesticideExposure nal pattern. Based on test transmitter location Selection estimates with less than perfect signal reception, No adult owls were detected foraging in pesticide we estimated 15 degrees as the maximum error risk zones in 1998. In 1999, 4 different individual angle still leading to estimated locations within owls were detected in pesticide risk zones on at receiver detection range. All 10 data sets result- least 1 occasion. The odds ratio for adult male use ing from the randomly drawn maximum error of recently sprayed fields in 1999 was 0.467 (95% displacement supported the conclusions reached CI: 0.169-1.286, n = 52 observations). This suggest- regarding habitat selection with the actual esti- ed no tendency to use or avoid risk-zone fields. We mated locations (Table 8; Gervais 2002). There- detected 5 recently fledged juveniles in risk zones fore, we concluded that neither location error in 1998 and 3 individuals in risk zones in 1999. nor habitat misclassification had influenced Juveniles in 1998 had lower odds of using a recent- analysis results. ly sprayed field than in 1999 (1998: odds ratio = 0.218, 95% CI = 0.072-0.662, n = 29 observations; DISCUSSION 1999: odds ratio = 0.387, 95% CI = 0.106-1.409, n = Home-range sizes of breeding male burrowing 35 observations). Juvenile odds ratio values sug- owls during the nesting season were highly vari- gest that recently fledged owls are less likely to able among individuals and were not accounted forage in CROPLAND than in other cover types. for by reproductive output as defined by number of fledgling owlets, numbers of nearby nests, near the nor the 50 cover-type composition nest, numbers of rodents in the diet as indexed by pellet The of these 40 analysis. poor explanatory power o factors is surprising, given that intuitively they would seem to be in the i 30 important determining 0 amount of space required to obtain adequate 2 food. factors be more 20 Relationships among may complex than we assumed, and we may have failed to measure Home- 10 important components. range size estimates may be a poor indicator of the and of resources an organism 0 types quantities 0 500 1000 1500 2000 2500 3000 needs for a given life-history stage. Distanceto nest (m) Burrowing owls in this study area used the agricultural fields extensively. However, habitat selec- Fig. 1. Meanpercentage of observationsof foragingmale bur- tion patterns were not nearly as clear as those rowingowls at distance intervalsfrom the nest, NavalAir Sta- found in earlier work in other tion, Lemoore,California, USA, 1998-1999. Errorbars are 1 landscapes (Haug standarderror of the mean. and Oliphant 1990). Breeding burrowing owls

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions 162 BURROWINGOWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al.J.Wildl. Manage. 67(1):2003

Table8. Radiotelemetryerror test resultsfor the space use study of male burrowingowls at NavalAir Station (NAS) Lemoore, California,USA, 1998-1999. Tentests incorporatingmaximum estimated location error were performed,and each was summa- rizedover the 33 owls. The resultsof the 10 tests are averagedbelow and the AkaikeInformation Criterion weights (AICc)cal- culatedfor the mean result.

Model AICc 95% CI AAICc AICcweights dnest + dnest2+ dnest3 414.7032 414.1834-415.2230 0 0.7639 dnest + dnest2+ dnest3+ cover type 418.0771 417.5796-418.5747 3.3739 0.1414 dnest + dnest2+ dnest3 + log dist to edge + cover type 420.2999 419.6848-420.9150 5.5767 0.0465 lineardistance to nest + grass cover + dist*grasscover 422.4255 421.7597-423.0913 7.7223 0.0161 lineardistance to nest 422.7802 421.9656-423.5949 8.0770 0.0135 log distanceto nest 423.6412 422.8044-424.4780 8.9380 0.0087 log distanceto nest + log distanceto edge 424.7089 423.8376-425.5802 10.0057 0.0051 log distance to nest + cover type 425.4778 424.7060-426.2497 10.7746 0.0035 log distanceto nest + log distance to edge + cover type 427.5112 425.6286-429.3937 12.8080 0.0013 cover type 490.8776 489.9950-491.7602 76.1744 0.0000 log distance to edge 509.6501 509.0140-510.2862 94.9469 0.0000 interceptonly 511.1309 510.8591-511.4027 96.4277 0.0000

carry single prey items back to the nest burrow Burrowing owls are opportunistic foragers that and consequently fit the classic-central-place, sin- eat a wide variety of vertebrate and invertebrate gle-prey-loader foraging models (Stephens and prey (Green et al. 1993, Haug et al. 1993, Gervais Krebs 1986). It is therefore not surprising that et al. 2000, York et al. 2002). The birds consis- distance was consistently an important compo- tently choosing one cover type over another nent of models distinguishing random from actu- seems unlikely when faced with spatially and tem- al foraging locations during the breeding season. porally varying prey populations, provided that Although energetic demands, on male owls in cover characteristics did not preclude successful particular fluctuate greatly depending on nesting foraging altogether. Agricultural lands composed stage, we were unable to explore finer-grained of row crops can be adequate habitat in and of home-range sizes or habitat selection by individ- themselves, as demonstrated by the high densi- ual owls. Too few locations during each individ- ties of burrowing owls in the Imperial Valley ual owl's nesting cycle existed to make such a where few fallowed or uncropped fields exist comparison meaningful. (Rosenberg and Haley in press). Contrary to our initial predictions, male owls Owls in all parts of the study site were detected selected neither GRASS cover nor edges of roads foraging in CROPLAND recently treated with or irrigation ditches for foraging. Owls also did agricultural pesticides, although we found no evi- not appear to avoid CROPLAND, which suggest- dence of selection for the CROPLAND cover ed that the resources needed by the owls may have type. Ingestion of pesticide-contaminated prey is been distributed across owl home ranges indepen- a demonstrated threat to nontarget species dent of cover types. Therefore, no selection for (Henny et al. 1985, White and Kolbe 1985, Hunt any particular cover type was needed for success- et al. 1991). Whether animals may be attracted to ful foraging. In our study, it was difficult to deter- pesticide application events is less clear. Burrow- mine whether male owls selected particular cover ing owls certainly are capable of tracking shifting types once distance was accounted for. All nest resource availability. We witnessed a shift in burrows were either in GRASS cover or located their diets in 1999 in response to a major increase under structures immediately adjacent to GRASS. in rodent densities (Gervais 2002). In any case, Haug and Oliphant (1990) found that burrow- owls used agricultural fields in our study land- ing owls were more likely to use grass-forb areas scape. Depending on chemical persistence and than croplands or grazed pasture. The landscape toxicity, they may be at risk from either direct in their study had much greater interspersion of exposure of pesticides or ingestion of contami- cover types than that of NAS Lemoore. But it would nated prey. Owls foraged in fields recently treat- be interesting to reexamine those data including ed with compounds that are highly toxic to birds, distance to nest as a potential explanatory vari- including the organophosphate compound chlor- able. After distance has been accounted for, inter- pyrifos and the carbamate compound aldicarb pretation of apparent habitat selection may change (Gervais 2002). No radiomarked owls died after dramatically (Rosenberg and McKelvey 1999). foraging in these fields during the study, howev-

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions J. Wildl.Manage. 67(1):2003 BURROWING OWL SPACE USE AND PESTICIDE EXPOSURE RISK * Gervaiset al. 163 er. Use of agricultural fields alone cannot be a ACKNOWLEDGMENTS basis for formal risk assessment, it indi- although Our work was supported by U.S. Navy EFA West cates that some estimation of risk exposure may and California Department of Fish and Game. be necessary. Our thanks especially to J. Crane and G. Buma of We estimated a much greater radiotelemetry NAS Lemoore for their support and assistance. error than is of typical null-peak receiving sys- M. Abright, M. Anderman, M. Bond, C. Bailey, V. tems and 1990). (e.g., Haug Oliphant Burrowing Franke, T. Lanman, and J. Podulka assisted in owls are candidates for fine-scale generally poor radiotracking, and additional field assistance was work because move fre- radiotelemetry they provided by C. Dalton and S. Solomon. We thank and while quently rapidly foraging. Sampling B. Glenn, M. Ricca, and L. Whitney for assistance error biases location estimates toward cover types with GIS analyses and helpful discussions and H. with the detection These greatest probability. Packard andJ. Rosier who examined owl pellets. J. have not been well in radio- issues explored Barth, E. Forsman, B. Glenn, M. Schulz, R. Steidl, data collection and a telemetry analysis, although and an anonymous reviewer commented on drafts tremendous exists for results potential spurious of this manuscript. This project was conducted as and Noon 2001). However, our (McKelvey study part of the Burrowing Owl Research Program, a was of landscape composed large, contiguous collaborative research program including The blocks of cover as individual farm fields are type Institute for Bird Populations, Oregon State Uni- 65 ha, and GRASS areas typically approximately versity, the Oregon Cooperative Fish and Wildlife between 45 and 179 ha, the ranged decreasing Research Unit, and San Jose State University. of impact relatively poor system performance. Cooperators of the Oregon Cooperative Fish and MANAGEMENTIMPLICATIONS Wildlife Research Unit included the U.S. Fish and Wildlife Service, Oregon State University, Ore- In view of the tremendous in the variability gon Department of Fish and Wildlife, the Wildlife estimates individual birds, we home-range among Management Institute, and the Biological Re- caution number as an against using any single sources Division of the U.S. Geological Survey. indication of how much space a breeding owl pair may need. However, habitat selection was LITERATURECITED heavily influenced by distance to the nest. Habi- ALLISON, P. D. 1999. Logisticregression using the SAS for owls should focus tat improvements breeding system. SAS Institute, Cary, North Carolina, USA. efforts within 600 m of nest burrows to maximize BURNHAM,K. P., ANDD. R. ANDERSON.1998. Model selec- USA. foraging efficiency. In addition, because owls are tion and inference. Springer, New York, risks be mit- CALIFORNIA DEPARTMENT OF PESTICIDE REGULATION. 1998. central-place foragers, pesticide may Annual use and Fresno coun- near pesticide reports, Kings igated by avoiding pesticide applications ties. California Department of Pesticide Regulation, nest burrows (James and Fox 1987). Our data Sacramento, USA. suggested that adults concentrate foraging efforts -. 1999. Annual pesticide use reports, Kings and within 600 m of the nest burrow, as was observed Fresno Counties. California Department of Pesticide in Canada and 1990) and south- Regulation, Sacramento, USA. (Haug Oliphant COULOMBE,H. N. 1971. Behavior and population ecolo- ern California and in (Rosenberg Haley press). gy of the burrowing owl, Speotytocunicularia, in the Maintaining a buffer zone of 500-600 m would ImperialValley of California.Condor 73:162-176. most and secondary poisonings, DESANTE,D. S., E. D. RUHLEN,S. L. ADAMANY,K. M. BUR- prevent primary in this a substantial TON,AND S. AMIN.1997. A censusof burrowingowls although might pose negative central California in 1991. 38-48 L. Lincer on the farmer in the case of Pages inJ. impact cropland and K. Steenhof, editors. The burrowing owl: its biol- under active production. ogy and management. Raptor Research Report No. 9. Dispersingjuvenile burrowing owls used a num- Allen Press, Lawrence, Kansas, USA. A. SAVIDGE.1999. Satellite burrow ber of burrows as they moved away from their DESMOND,M.J., ANDJ. use owl chicks its influence on nest natal territories (Desmond and 1999, by burrowing and Savidge fate. Studies in Avian Biology 19:128-130. King and Belthoff 2001). We recommend that ----- , AND K. M. ESKRIDGE.2000. Correlations disturbance or pesticide exposure of non-nest between burrowing owl and black-tailed prairie dog burrows in areas of nesting populations be mini- declines: a seven-year analysis. Journal of Wildlife mized in late summer, when young owls are most Management 64:1067-1075. FRANKLIN,A. B., T. M. SHENK,D. R. ANDERSON,AND K. P. likely to be using them. Nonbreeding owls can be BURNHAM.2001. Statistical model selection: an alter- cryptic and determining their presence should native to null hypothesis testing. Pages 75-90 in T. M. not be based only on casual observations. Shenk and A. B. Franklin, editors. Modeling in nat-

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions 164 BURROWINGOWL SPACE USE AND PESTICIDEEXPOSURE RISK * Gervaiset al.J. Wildl. Manage. 67(1):2003

ural resource management. Island Press, Washington ecology and behaviour in central New Mexico. Con- D.C., USA. dor 75:446-456. GERVAIS,J. A. 2002. Evaluating space use and pesticide MCKELVEY,K. S., ANDB. R. NOON.2001. Incorporating exposure risk for burrowing owls in an agricultural uncertainties in animal location and map classifica- environment. Dissertation, Oregon State University, tion into habitat relationships modeling. Pages 72-90 Corvallis, USA. in C. T. Hunsaker, M. F. Goodchild, M. A. Friedl, and - , D. K. ROSENBERG,D. M. FRY, L. TRULIO, AND T. J. Case, editors. Spatial uncertainty in ecology: K. K. STURM.2000. Burrowing owls and agricultural implications for remote sensing and GIS applications. pesticides: evaluation of residues and risks for three Springer-Verlag, New York, USA. populations in California, USA. Environmental Toxi- NAGELKERKE,N. J. D. 1991. A note on a general defini- cology and Chemistry 19:337-343. tion of the coefficient of determination. Biometrika GILL,A. E. 1977. Food preferences of the California vole, 78:691-692. Microtuscalifornicus.Journal of Mammalogy 58:229-233. ROSENBERG,D. K., ANDK. L. HALEY.In press. The ecolo- GREEN,G. A., ANDR. G. ANTHONY.1989. Nesting success gy of burrowing owls in the agroecosystem of the and habitat relationships of burrowing owls in the Imperial Valley, California. Studies in Avian Biology. Columbia Basin, Oregon. Condor 91:347-354. ----, ANDK. S. MCKELVEY.1999. Estimation of habitat - , R. E. FITZNER,R. G. ANTHONY,AND L. E. ROGERS. selection for central-place foraging animals. Journal 1993. Comparative diets of burrowing owls in Oregon of Wildlife Management 63:1028-1038. and Washington. Northwest Science 67:88-93. SEAMAN,D. E., J. B. GRIFFITH,AND R. A. POWELL.1998. GRIGGS,F. T. 1992. The remaining biological diversity of KERNELHR:a program for estimating animal home the San Joaquin Valley, California. Pages 11-16 in D. F. ranges. Wildlife Society Bulletin 26:95-100. Williams, S. Byrne, and T. A. Rado, editors. Endan- , J. J. MILLSPAUGH,B. J. KERNOHAN,G. C. BRUNDI- gered and sensitive species of the San Joaquin Valley, GE, K. J. RAEDEKE,AND R. A. GITZEN. 1999. Effects of California: their biology, management, and conserva- sample size on kernel home range estimates. Journal tion. CaliforniaEnergy Commission, Sacramento, USA. of Wildlife Management 63:739-747. HAUG, E. A., B. A. MILLSAP,AND M. S. MARTELL.1993. , ANDR. A. POWELL.1996. An evaluation of the Burrowing owl (Speotytocunicularia). Pages 1-18 in A. accuracy of kernel density estimators for home range Poole and G. Gill, editors. The birds of North America, analysis. Ecology 77:2075-2085. vol. 61. Philadelphia Academy of Natural Sciences, SHEFFIELD,S. R. 1997. Current status, distribution, and Philadelphia, Pennsylvania, and American Ornitholo- conservation of the burrowing owl (Speotytocunicularia) gists' Union, Washington D.C., USA. in midwestern and western North America. Pages - , ANDL. W. OLIPHANT.1990. Movements, activity 399-407 in J. R. Duncan, D. H. Johnson, and T. H. patterns, and habitat use of burrowing owls in Saskatch- Nicholls, editors. Biology and conservation of owls of ewan. Journal of Wildlife Management 54:27-35. the northern hemisphere. U.S. Forest Service Tech- HENNY,C. J., L. J. BLUS,E. J. KOLBE,AND R. E. FITZNER. nical Report NC-190. 1985. Organophosphate insecticide (famphur) topi- SILVA, S. I., I. LAZO, E. SILVA-ARANGUIZ,F. M. JAKSIC,P. L cally applied to cattle kills magpies and hawks. Jour- MESERVE,AND J. R. GUTIERREZ.1995. Numerical and nal of Wildlife Management 49:648-658. functional response of burrowing owls to long-term HOSMER,D. W., ANDS. LEMESHOW.1989. Applied logistic mammal fluctuations in Chile. Journal of Raptor regression. John Wiley & Sons, New York, USA. Research 29:250-255. HUNT,K. A., D. M. BIRD,P. MINEAU,AND L. SHUTT.1991. STEPHENS,D. W., ANDJ. R. KREBS.1986. Foraging theory. Secondary poisoning hazard of fenthion to American Princeton UniversityPress, Princeton, NewJersey, USA. kestrels. Archives of Environmental Contamination THOMPSON,C. D., AND S. H. ANDERSON. 1988. Foraging and Toxicology 21:84-90. behavior and food habits of burrowing owls in JAMES,P. C., ANDR. H. M. ESPIE.1997. Current status of the Wyoming. Prairie Naturalist 20:23-28. burrowing owl in North America: an agency survey. THOMSEN,L. 1971. Behavior and ecology of burrowing Pages 3-5 inJ. L. Lincer and K. Steenhoff, editors. The owls on the Oakland Municipal Airport. Condor burrowing owl: its biology and management. Raptor 73:177-192. Research Report 9. Allen Press, Lawrence, Kansas,USA. WHITE, D. H., ANDE. J. KOLBE.1985. Secondary poison- , ANDG. A. Fox. 1987. Effects of some insecticides ing of Franklin's gulls in Texas by monocrotophos. on productivity of burrowing owls. Blue Jay 45:65-71. Journal of Wildlife Diseases 21:76-78. JENNRICH,R. I., ANDF. B. TURNER.1969. Measurement of WORTON,B. J. 1995. Using Monte Carlo simulations to non-circular home range.Journal of Theoretical Biol- evaluate kernel-based home range estimators. Jour- ogy 22:227-237. nal of Wildlife Management 59:794-800. KAMRIN,M. A., editor. 1997. Pesticide profiles: toxicity, YORK,M., D. K. ROSENBERG, AND K. K. STURM.2002. Diet environmental impact, and fate. Lewis Publishers, and food-niche breadth of burrowing owls (Athene Boca Raton, Florida, USA. cunicularia) in the Imperial Valley, California. West- KING, R. A., ANDJ. R. BELTHOFF.2001. Post-fledging dis- ern North American Naturalist 62:280-287. persal of burrowing owls in southwestern Idaho: char- acterization of movements and use of satellite bur- Received9 November2001. rows. Condor 103:118-126. Accepted22 August 2002. MARTIN,D. J. 1973. Selected aspects of burrowing owl AssociateEditor: Kelly.

This content downloaded from 128.114.163.7 on Mon, 9 Jun 2014 12:44:08 PM All use subject to JSTOR Terms and Conditions