J. RaptorRes. 34(2):102-107 ¸ 2000 The Raptor ResearchFoundation, Inc.

HABITAT SELECTION BY TAWNY - ( KETUPA FLAVIPES) IN

YUAN-HSUN SUN Departmentof WildlifeConservation, National PingtungUniversity of Scienceand Technology,Pingtung, Taiwan 912

YING WANG Departmentof Biology,National TaiwanNormal University,Taipei, Taiwan 117

CHING-FENG LEE Departmentof WildlifeConservation, National PingtungUniversity of Scienceand Technology,Pingtung, Taiwan 912

ABSTRACT.--Fourradio-ta. gged Tawny Fish-Owls (Ketupaflavipes) were studiedto determinetheir choice of roosting and foraging/resting habitats in Taiwan. A total of 51 roost siteswere located in two terri- toriesmeasuring 6.7 and 5.7 km in length. Owlsselected old-growth more than expectedac- cording to availabilityand tended to roost uphill during late springand fall. They foragedalong streams more frequently than expected (P < 0.05) and along creeksless than expected (P < 0.05) with use of small creeks and fish farms increasingafter streamsflooded. KEYWORDS: Tawl•yFish-; Ketupa flavipes;radio-telemetry; territory; habitat selection.

Seleccitn de habitat de Ketupaflavipes en Taiwan R•SUMEN.--CuatroKetupa flavipes dotados de radiotransmisoresfueron estudiadospara determinar la seleccitn de perchas,forrajeo y habitatsde descansoen Taiwan.Un total de 51 sitiosde descansofueron localizadosen dos territoriosde 6.7 y 5.7 kms de longitud.Los buhosseleccionaron bosques maduros mas de lo esperadode acuerdo a la disponibilidady tendian a posarsecuesta arriba a finales de la primaveray el otofio. Forrajearon a lo largo de los arroyosmas frecuentementede lo esperado(P < 0.05) y a lo largo de los riachuelosmenos de lo esperado(P < 0.05) utilizandolos riachuelosy granjas piscicolasen •poca de inundaciones. [Traducci6n de C•sar Mgrquez]

There are four speciesof fish-owlsin Asia in the Bureau 1995) on eastern and southern banks of streams Ketupaand three speciesin Africa in the ge- Plantationsof Makino bamboo (Phyllostachysmakinoi) and Cryptomeriajaponica, as well as farmland and human set- nus Scotopelia(Fogden 1973). Except for descrip- tlements,occur on westernbanks. Native riparian forests tions of the nesting biology and circadian rhythm bordering the streamssupported large epiphyticbird's of the Tawny Fish-Owl (Ketupaflavipes) (Sun and nest ferns (Pseudodrynariacoronans) on mature trees >1 m in diamter. Wang 1997, Sun et al. 1997), little is known about We captured tbur Tawny Fish-Owlswith a variety of this rare and secretivespecies (Voous 1988). Here- foot-snaretraps while they foraged or restedalong stream in, we present data on the breeding territories and bankson tree branches,or on the banksof pondsat fish habitat selection of this . farms. The owls were banded, measured and radio tagged prior to release. Radio transmitters (MD-205; Te- METHODS lonics Inc., Mesa, Arizona, U.S.A.) weighed 70-80 g (<3.5% bodymass) and had a lifespanof approximately We studied Tawny Fish-Owlsat Nanshih Stream in Fu- 2 yr. Radios were attached with a backpackharness of san village (24ø48'N, 121ø30'E;400 m elevation), 30 km wire (1.5 mm in diameter) placed inside tubular teflon south of Taipei, Taiwan. The main streamis fed by three ribbon. streams 10-30 m in width (Hawun, Chakung and Talo- Owls were tracked using a directional, hand-held H- lan) and by nine creeks2-5 m in width. At the village, antenna with a Telonics TR-2 receiver and locations were there are four fish farms where rainbow trout (Onc0rhyn- obtained by triangulation taking at least two bearings chusmykiss) and/or ayu (Pleooglossusaltivelis) are raised. each time. If an owl moved while being tracked, new Vegetationin the area consistsmostly of tropical rainfor- bearings were taken. Most owl locations were obtained estsdominated by Ficusand Lauraceae (Taiwan Forestry along a road parallel to and <150 m from the streams.

102 JUNE 2000 TAWNYFISH-OWL HABITAT SELECTION 103

Radio locationswere plotted on habitat maps made from utaries (Talolan, Chakung and Hawun), a linear 1:10 000 aerial photos. distance of 6.7 km. We radio tracked each owl 1-2 times daily during day- light hours to determine roost sites.Due to the bearing- In October 1994, a subadultowl (W508) wascap- reduced inaccuracy of triangulation, only owl locations tured and radio tagged at the Hsiapen fish farm with error polygonswithin a specific riparian habitat (Fig. 1). It moved to the Loshanchun fish farm and patch were included in the analysisof roosting habitat remained there for about three months before selection. We defined a fragmented old-growth as leaving the area. It flew over the adjacent Fusan one that wasopened by loggingor was<100 m of a road. Other features of roost sites, such as the distance to water, territory and wandered along the upper stretch of edge and human developments (e.g., villages or roads) Talolan Stream and Maen, Wuchun and Tunlu were also measured. To determine foraging habitat use, Creeks in January-February 1995. In early Febru- we radio tracked each owl for 1-1.5 hr at night. This ary, it returned to the Loshanchun fish farm after sampling interval was assumed to give independent lo- an untagged owl wasshot and killed at the Hsiapen cationssince it waslong enough for owlsto move to any location within their territories (Lair 1987). We assumed fish farm. In late April, it left and returned to Tal- all owls were involved in foraging activity during our olan Stream. We found it once in the Fusan terri- tracking becauseof the difficulty in distinguishingresting tory and it sometimes roosted near the Hsiensen from foraging behaviorat night. Owl locationswith error fish farm while owls W503 and W513 nested on the polygonscentered -•100 m from water were categorized opposite side of Chakung Stream. We lost the sig- as either stream or creek habitat, depending on their stream order (streams3-4; creeks 1-2); otherwise, they nal of W508 in May 1996 and never found it again. were classifiedas riparian habitat. A second adult owl (W494) was radio tagged in We calculated the percent area of each water course this same area in earlyJune 1995 (Fig. 1). This owl (--•100 m fi•om streams or creeks) and riparian forest and its untagged mate moved into the vacant ter- (>100 m from the streams or creeks) within 500 m of ritory of the owl killed at Hsiapen fish farm in Feb- either side of a 14-km section of Nanshih Stream (ca. ruary. The pair did not feed at a fish farm until 1130 ha) by overlying an aerial photograph (1:10000) with a 1 X 1 cm grid matrix and counting the number November 1995 when W494 foraged at Loshan- of intersecting triangulation points within each habitat chun fishfarm. This territory was5.7 km in length. type. We used five riparian cover types:old-growth forest, The two breeding territories were mutually exclu- second-growth forest, forest plantation, grassland and sive with a boundary somewherenear the mouth farmland/village. Each point was checked to determine if there had been any recent changesin cover type. To of Awang Creek. We located a total of 51 different roost sites determine owl selection of roosting and foraging habi- tats, we used the percent area of each riparian and water along Nanshih Stream based on 106 radio loca- cover type for availablehabitat valuesand the percent of tions (Fig. 2). Each Tawny Fish-Owl used 14-17 roosting and foraging sitesin each cover type for habitat roost sites over the study period. Roost sites of use values. W494 and W508 were clustered in a riparian old- Chi-squareanalysis (Conover 1980) was used to deter- mine if actual habitat use differed from expected (Neu growth forest near the Loshanchun and Hsiapen et al. 1974). We used Wilcoxon signed-ranktests (Con- fish farms across Nanshih Stream. Two of 16 roost over 1980) to compare owl roost distancesto human de- sites were near Fusan fish farm ((500 m) and velopment and to edge types (e.g., second-growthforest, W503 and W513 used them most often with 40 of plantations and shrub/grass). Data were analyzed with 64 radio locations occurring there. the StatisticalAnalysis System (SAS Institute 1987). On 10 occasions, W503 and W513 roosted (100

RESULTS m from each other) but, on 6 d, they roosted as far as 2 km from each other, with the male always Four owls were captured between September near the territory boundary at Maen Creek. This 1993-August 1996. A pair of owls (W503-male, happened whenever the pair from the Hsiapen ter- W513-female) was captured at Fusan fish farm in ritory were found at the upper stretch of Talolan September1993 (Fig. 1). The pair wastracked for Stream. about a month until the transmitters fell off when Roost sites ranged from 20-550 m (• -- 138 ñ the harnesses broke. The male and female were 85 m, ñSD) from water and 29 roost sites (56.9%) recaptured in May 1995 and August 1995, respec- were located (200 m from water (Fig. 2). Most tively, near the Hsiensen fish farm. Based on the roost sites (82.4%) were situated on eastern and radio locations in 1995-96, this pair continued to southern banks of streamswhere disturbanceby use the same territory used in 1993. It included a people was minimal. Owl roostswere usuallylocat- stretch of Nanshih Stream and its three main trib- ed farther from human developments (• • 418.3 104 SUN ET A_L. VOL. 34, No. 2

Wuchun Creek

Nanshih Stream

o W503 ø W513 /• W508 •, W494 Hsiap•

# Fish farms Loshanchun

•..,/ MaenCreek Awang Creek •_•.• TalolanStream Fusan

Stream

Hsinshen

0 500m ChakungStream F•gure l. Nightly radio locationsof four radio-taggedTawny Fish-Owls in two territoriesat Nanshih Streamfrom September 1993-August 1996.

-- 323.5 m), such as villagesand roads,than from expectedwithout significance (P > 0.05). Highly- edgessuch as second-growth forest, plantation, and disturbed habitatssuch as second-growthforest, shrub/grass (i = 263.7 --- 167.7 m; Wilcoxon plantations,grassland, farmland and villageswere Signed-rank test, P = 0.001). significantlyavoided by the owls (P • 0.05). Use of roostsvaried with seasons(Fig. 2). In the We had a total of 303 radio locations of owls at colder months of winter and early spring,the owls night. They showedthat owls used water as op- tended to roost in the vicinity of streams,while posed to riparian habitatsmore often (X'• = 844.8, from late spring to fall, they shifted to roostslo- df = 1, P = 0.001). They also foraged along cated uphill from streams. streamsmore frequently than creeks (X'• = 12.5, df Roostinghabitat use did not vary amongowls (X 2 = 1, P = 0.001). Use of streamswas affectedby = 1.97, df = 6, P = 0.92; Table 1) and they all waterlevel (X'• = 86.9,df = 4, P = 0.001;Fig. 3). choseold-growth forests more often than excepted They foragedat fishfarms more during periodsof (Bonferroni Z test, P < 0.05). Fragmented old- medium water level and at creeks when water levels growth forest was used slightlymore or lessthan were high. JUNE 2000 T^WNYFIsH-OWL Iq_ABITAT SFLECTION 105

WuchunCere

,,• NansaShS• • ß•

-1000m2

..

oeo / ß o ß

Cree•' o

...'•c ß• .

ß

500 m Talolan Stream +

Original forest Fragmentedoriginal forest Second-growth Plantation, grass, farmland, and Fish farm village

ß late spring-fall o winter-early spring

Figure 2. Roost sitesof four radio-marked Tawny Fish-Owlsat Nanshih Stream from September 1993-August 1996. Black dots indicate summer to mid-fall roost sitesand white dots indicate late fall to spring roost sites. 106 SUN ET )•L. VOL. 34, NO. 2

Table 1. Occurrence of Tawny Fish-Owlroost sitesin three cover typesof a 1300-ha area in Nanshih Stream, Taiwan

NUMBER PROPORTION BONFERRONI INTERVAL OWL COVERTYPE a OF ROOSTS OF ROOSTS (95% CONFIDENCEINTERVAL) W503 Old-growthforest 45 0.918 0.827 -< P1 -< 1.010 Fragmented old-growth 3 0.061 -0.030 -< P3 -< 0.068 Highly-disturbedhabitat b I 0.021 -0.020 -< P2-< 0.141 W513 Old-growthforest 14 0.875 0.682 -< P1 -< 1.068 Fragmented old-growth 2 0.125 0.000 -< P3 -< 0.000 Highly-disturbedhabitat 0 0.000 -0.070 -< P2-< 0.318 W508 Old-growth forest 29 0.935 0.832 -< P1 -< 1.039 Fragmented old-growth 2 0.065 0.000 -< P3 -< 0.000 Highly-disturbed habitat 0 0.000 -0.040 -< P2 -< 0.168 W494 Old-growth forest 10 0.909 0.706 -< P1 -< 1.112 Fragmented old-growth I 0.091 0.000 -< P3 -< 0.000 Highly-disturbed habitat 0 0.000 -0.020 -< P2 -< 0.144 Proportions(expected use) of eachcover type are 0.575 (old-growthforest), 0.116 (fragmentedold-growth forest) and 0.309 (highly- d•sturbed habitat). Includessecond-growth forest, forest plantation, farmlands, grassland and village.

DISCUSSION tories measuring only 1-2 km in length along riv- Territory size decreasesas the number of com- ers in Borneo (Fogden 1973). We found that Taw- petitors increasesin areas of abundant food (Da- ny Fish-Owlsfavored riparian old-growthforests for vies and Houston 1984). At Nanshih Stream where roostingwhich may partially accountfor the distri- fish farms provided a abundant food resourcefor bution of the speciesin old-growthforest areason owls, territories were smaller than at Sakatang Taiwan Island (Sun 1996). Old-growthforests, with Stream where there were no fish farms. In general, their tall canopy (20-30 m), may provide greater larger usually maintain larger territories safety from ground predators and their open na- (Schoener 1968). In Asian fish-owls,there is a pos- ture may allow these large owls more maneuvera- itive correlation between body size and territory bly. size. The large Blakiston Fish-Owl (K. blakistoni) Hayward et al. (1987) found that Boreal Owls which is 20% larger than the Tawny Fish-Owl (Aegoliusfunereus) roosted near their last foraging (Voous1988), occupiedstretches of river 10-19 km site before dawn. This may explain why TawnyFish- in length in Russia(Pukinskiy 1973), whereasthe Owls roosted more frequently near fish farms. smallestMalay Fish-Owl (K. ketupa)which is about They also roosted at higher elevationsduring the 20% smaller than the Tawny Fish-Owlhave terri- hot summer and early fall seasons,probably in re- sponseto heat stress.Barrows (1981) alsoreported 90 that summer roostsof Spotted Owls (Strix occiden- 8O talis) tended to be in cooler habitats, such as north- • 70 ern slopesand under dense canopy.A male ob- •60 I-ILow served nest guarding showed signsof heat stress • 40 ßHigh when air temperatureswere between 30-33øC in early May (Sun 1996). This might explain why two • •0 of the owlswe studied moved uphill and downhill within their territories and why one owl temporar- •o•502o 0 • • / glMedium Stream Fish faro Creek ily left its territory moving upstreamin summer. Habitat •pe Tawny Fish-Owlsrely heavilyon aquatic habitats Figure •. Proportions of nightly radio locationsof four becausethey eat mostlyfish, amphibiansand crus- •adlo4agg•d Tawny Fish-Owlsat th• diff•nt habitats taceans (Sun 1996). Fogden (1973) noted that its dunng periods of low, medium •d high water levels at uniquely-scaledfeet make it well-suited to take N•shih stream, Taiw•, September 199•-August 1996. such prey. We found that Tawny Fish-Owls pre- JUNE2000 TAWNYFISH-OWL HABITAT SELECTION 107 ferred to use streamsover creeks during low water FOGDEN,M. 1973. Fish-owls,eagle-owls and the Snowy levels.Most creeksin our studyarea were <5 m in Owl. Pages 53-85 in J.A. Burton [ED.], Owls of the width and some were even narrower, with dense world--their evolution, structure and ecology.A. and overhanging vegetation blocking some channels W. Visual Library, New York, NY U.S.A. HAYWARD, G.D., P.H. HAYWARDAND E.O. GARTON. 1987. along the creeks. This may have discouragedTaw- Movements and home range use by Boreal Owl in ny Fish-Owls,with a 1.5-m wingspan,from foraging central Idaho. Pages175-184 in R.W. Nero, RJ Glark, in these areas despite their abundant food. RJ. ICmaptonand R.H. Hamre [EDS.], Biology and Tawny Fish-Owls increased their use of creek conservationof northern Ibrest owls.Symposium Pro- habitat when streamswere flooded. This may have ceedings,Gem Tech. Rept. RM-142, USDA Forest Ser- been due to the fact that creeks were much cleaner vice, Denver, GO U.S.A. than streamsand prey specieslike fish and crabs LAIR,H. 1987. Estimatingthe locationof the Ibcal center were much less difficult to find. This may explain in red squirrel home ranges.Ecology 68:1092-1101. why Tawny Fish-Owls also turned to fish farms as NEU, C.W., C.R. BYERSAND J.M. PEEK.1974. A technique stream conditionsdeteriorated. Like Ospreys(Pan- Ibr analysis of utilization-availabilitydata. J. Wildl. Manage.38:541-545. dion haliaetus),fish-eating species like the Tawny POOLE,A.E 1989. : a natural and unnatural his- Fish-Owl would be expected to leave turbid estu- tory. Cambridge Univ. Press,Cambridge, U.K. aries and fly to inland poolswith clear water (Poole PUKINSKIY,Y.B. 1973. Ecologyof BlakistonFish-Owl in the 1989). Bikin River basin. Byull. Mosk. Obshch.Prir. Otd. Bzol. 78:40-47. ACKNOWLEDGMENTS SASINST•TUTV,. 1987. SASuser guide. SASInstitute, Inc., The fbllowing people assistedin the collectionof field Cary, NC U.S.A. data: P. Chiang, C. Fang, T. Fu, L. Hsiao, T. Hsu, H. Lee, SCHOENER,T.W. 1968. Size of Ibeding territories among L. Liao, Y. Liao, H. Mai, P. Mark, Y. Sun, T. Tin, T. Wang, H. Wu, S. Wu and Y. Wu. Without their great assistance, birds. Ecology49:123-141. the studywould not have been completed. This manu- SUN, Y. 1996. Ecology and conservationof Tawny Fish- script wassubstantially improved by CJ. Ralph, NJ. Silvy, Owls in Taiwan. Ph.D. dissertation, Texas A&M Univ., R.D. Slackand T.L. Thurow. The researchwas supported College Station, TX U.S.A. by the Council of Agriculture, Republic of . --AND Y. WANG.1997. Activity pattern of TawnyFish- Owl. Wilson Bull. 109:377-381. L•TV, RATURr C•TV, D

, -- ^ND K.A. ARNOLD. 1997. Notes on a nest BARROWS,C. 1981. Roost selection by Spotted Owls: an of Tawny Fish-Owl at SakatangStream, Taiwan.J. Rap- adaptation to heat stress.Condor 83:302-309. tor Res. 31:387-389. CONOVrR,WJ. 1980. Practical nonparametric statistics, TAIWAN FORESTRY BUREAU. 1995. The third forest resourc- 2nd ed. John Wiley and Sons,New York, NY U.S.A. es and land use inventory in Taiwan. Taiwan Forestry DAVIES,N.B. ANDA.I. HOUSTON.1984. Territory econom- Bureau, Taipei, Taiwan. (in Chinese) ics. Pages 148-169 in J.R. Koreas and N.B. Davies Voous, K.H. 1988. Owls of the northern hemisphere lEDS.I, Behavioralecology: an evolutionaryapproach, MIT Press,Cambridge, MA U.S.A 2nd ed. Blackball Scientific Publications, Oxfbrd, U.K. Received17 May 1998; accepted19 December 1999