—— Raptor Res. 36(4):287-293 J. © 2002 The Raptor Research Foundation, Inc. NESTING AND PERCHING HABITAT USE OE THE MADAGASCAR FISH-EAGLE James Berkelman^ and James D. Fraser Department ofFisheries and Wildlife Sciences, Virginia Polytechnical Institute and State University, Blacksburg, VA 24061-0321 U.S.A. Richard Watson T. The Peregrine Fund, 5668 West Flying Hawk Lane Boise, ID 83709 U.S.A. Abstract. We documented Madagascar Eish-Eagle (Haliaeetus vociferoides) nest and perch use on lakes and rivers and compared parameters ofused trees to unused reference trees. Nest and perch trees were broader and taller, had more unobstructed branches, and were less obstructed by adjacent trees com- pared to reference trees. Perch trees also were more often deciduous than reference trees. Nest sites had more shoreline perch trees than reference sites. Logistic regression models with tree height as the independentvariable distinguished nest and perch trees from randomly selected reference trees. Models m with number ofperch trees along a 1.25 ha (50 width) shoreline section distinguished nest sites from reference sites. These models suggest that the presence of trees >15 m tall within 50 m ofthe shoreline is a good predictor of Madagascar Eish-Eagle habitat use. KeyWords; MadagascarFish-Eagle, Haliaeetus vociferoides; habitat; Madagascar, nest tree, perch tree, shore- line. USO DE HABITAT DE ANIDACION Y PERCHA DEL AGUILA PESCADORA DE MADAGASCAR Resumen. -Documentamos el uso de nidos y perchas para el aguila pescadora de Madagascar {Haliaeetus vociferoides) en lagos y rios y comparamos parametros de arboles usados con arboles no usados de referenda. Los nidos y arboles percha fueron mas anchos y mas altos, tenian mas ramas despejadas, y estaban menos obstruidos por arboles adyacentes en comparacion con los arboles referenda. Los arboles percha fueron ademas algunas veces mas deciduos que los arboles control. Los sitios nido disponian de mas arboles percha costeros que los sitios de referencia. Los modelos de regresion logistica con la altura de los arboles como variable independiente distinguieron los nidos y arboles percha de arboles control m seleccionados aleatoreamente. Los modelos con numeros de arboles percha cerca a 1.25 ha (50 de ancho) de la seccion de costa distinguieron los sitios nido de los sitios referencia. Estos modelos sugieren que la presencia de arboles >15 m de alto dentro de 50 m de la linea costera es un buen pronosticador del uso de habitat del aguila pescadora de Madagascar. [Traduccion de Cesar Marquez] With a population estimate of 99 breeding pairs ies of Madagascar Fish-Eagle nesting or perching (Rabarisoa et al. 1997), the Madagascar Fish-Eagle habitat use. {Haliaeetus vociferoides) is one of the rarest birds of Nelson and Horning (1993) estimated from sat- prey in the world (Meyburg 1986). Until recently, ellite data that Madagascar’s forest cover had been little was known about the species’ ecology and sta- reduced to 10.4% of the island by 1990. Nest-site tus. Langrand and Meyburg (1989) noted that the availability is a key limiting factor for raptor pop- Madagascar Fish-Eagle used tall trees near water ulations (Newton 1979). Also perch-tree distribu- for nests and foraging perches, but prior to this tion is a reliable predictor ofBald Eagle {Haliaeetus study, there had been no detailed quantitative stud- leucocephalus) distribution on the Chesapeake Bay (Chandler et al. 1995). Thus, we focused our study ' Present address: Department of Wildlife Ecology, Uni- on both nest and perch trees, along with the sur- versity of Wisconsin, 218 Russell Labs, 1630 Linden Dr., rounding habitat conditions. The objectives of this Madison, W1 53706-1598; e-mail address: jberkelman® study were to determine characteristics of nest facstaff.wisc.edu trees, nest sites, and perch trees used by Madagas- 287 288 Berkelman et al. VoL. 36, No. 4 Table 1. Sites where Madagascar Fish-Eagle nests and perches were investigated in the region ofAntsalova, Mada- gascar, 1994. Site names are lakes unless otherwise indicated. Number oe Site Latitude, Longitude Eagle Pairs Masiadolo 18°41'S, 44°28'E H Besara 18°41'S, 44°16'E 1 Soahanina River 18°46'-48'S, 44°16'~19'E 3^ Antsahafa 18°48'S, 44°29'E P Masama 18°50'-5TS, 44“28'-29'E 2 Tsiandrora 18°58'S, 44°38'E 1 Andranolava 1'’ 19°0'S, 44°2TE 1 Befotaka 19°1'-2'S, 44°24'-25'E 3 Soamalipo 18°59'-19°2'S, 44°26'-27'E 3^ Ankerika 19°T-2'S, 44'^27'-44°28'E 4 Andranovorimirafy 19°3'S, 44°27'E 1 Andranolava 2'^ 19°4'S, 44°25'E 1 Antsakotsako 19°6'S, 44°33'E 1^ Ampozabe 19°9'S, 44°40'E 1 Bevoay 19°9'S, 44°25'E 1 Manambolo River 19°8'-9'S, 44°44'-49'S 2 Maromahia 19°11'-12'S, 44°37'-38'E 1 Bejijo 19°12'-14'S, 44°32 -33'E 1 No nest was found for one of the fish-eagle pairs at five of the sites. Two of the lakes in the study had the same name. car Fish-Eagles and to develop predictive models (>20 cm diameter at breast height [DBH]) near the to identify fish-eagle nesting and perching habitat. same bodies of water. To compare nest and perch trees to available large trees, we randomly selected a reference SiUDYArea and Methods tree for each nest or perch tree. We selected trees at the We conducted the study during the first half of the Tsoamdeoditshitsa,ncweefmreoamsuthreedwawtietrhaas hthiep ncehsatinorthpeerdcihsttarnecee Madagascar Fish-Eagle breeding season from 21 May-14 August 1994. We investigated fish-eagle nesting and from the nest tree to the nearest water (nest-water dis- peiching habitat in a 3000 km^ area in the Antsalova re- tance), from perch tree to nearest water (perch-water dis- tance), and the distance along the shore between the gion of western Madagascar (18°40'-19°15'S, 44^15'- nest and perch trees (nest-perch distance). We then ran- 44°50'E) that included the drainages of the Manambolo, Beboka, and Soahanina rivers west of the Bemaraha Pla- domly selected a shoreline reference point on the same teau. Topography consisted of coastal plains and low roll- body of water as the nest tree that we had measured ing hills with elevations ranging from sea level to 126 m. (within 1.5 km along the banks for nest trees on rivers) Soils were shallow and sandy, and the vegetation was a To select each nest reference tree, we went to the patchwork of dry deciduous forest, savanna, wetlands, shoreline reference point and moved inland a distance mangrove swamps, and rice paddies. The climate was sub- equal to the nest-water distance and selected the nearest humid and tropical with a dry season from April-Octo- tree >20 cm in DBH as our nest reference tree. We used ber and a wet season from November-March. Mean an- the same shoreline reference point to select a perch ref- nual rainfall in the region ranged from 1000-1500 mm erence tree by moving the nest-perch distance in the (Donque 1972). same direction (left or right) along the shoreline as that We defined a nest tree as any tree in which we ob- between the used nest and perch trees. We then moved served nest construction, incubation, or brood rearing. inland the perch-water distance and selected the nearest A nest site was the area within 300 m of the nest tree. A tree >20 cm in DBH as our perch reference tree. We peich tree was any tree in which we observed adult fish- used 20-cm DBH as a minimum size for reference trees eagles perching. We measured nest and perch trees of based on the minimum size of Bald Eagle perch trees on every known Madagascar Fish-Eagle pair in the study area the Chesapeake Bay (Buehler et al. 1992). (Table 1). Our perch-tree sample {N = 29) was larger We measured DBH of nest trees to the nearest cm and = than our nest-tree sample (N 24) because we did not used a clinometer to measure height to the nearest me- find a nest for five of the fish-eagle pairs. ter. We counted branches in the tree canopy that we es- We measured characteristics of fish-eagle nest trees timated to be >5 cm in diameter and unobstructed for and randomly selected reference trees to determine if 1 m above and below. We recorded arc ofaccessibility by trees used by fish-eagles differed from average large trees standing at the base of the tree and using a compass to December 2002 Mada(;asc]AR Fish-Eagle Nests 289 measure the total arc (0°-360°) that was unobstructed by probability of fish-eagle use of trees and sites based on m other trees for an estimated distance of 10 from the the measured habitat variables using stepwise analysis trunk and 3 m below the tree’s crown (Buehler et al. Our significance level for variables to both enter and exit 1992). We recorded nest-tree species and classified models was P = 0.05. We used dummy variables for growth form following Keister and Anthony (1983). Our growth form categories in the logistic regression (Hos- classification was based on the location ofthe lowest fork mer and Lemeshow 1989). We constructed classification in the trunk, and whether the tree was dead. We classi- tables for each logistic regression model by using the es- fied growth form as large if the lowest fork was in the timated logistic probabilities for each tree or site to pre- lower third of the trunk, medium if the lowest fork was dict fish-eagle use (Hosmer and Lemeshow 1989). We in the middle third of the trunk, and small if the lowest considered trees or sites as correctly classified as used hy fork was in the upper third of the trunk. We recorded fish-eagles if the predicted probabilities were ^0.5. growth form as dead top if the top third of the crown was dead and as snag if the entire tree was dead and Resui.ts leafless, regardless of the location of the lowest fork in the trunk. Nest-tree Characteristics. Nest construction, in- We measured minimum distance of each nest tree to cubation, or brood rearing was observed at 21 water with a hip chain and minimum distance to human (87.5%) of the 24 measured nest trees in 1994. disturbance, building, road, and fish-eagle nest from maps and aerial photos. Human disturbances included The remaining three nest trees were used in 1993, agricultural clearings, rice paddies, villages, tombs, and but not in 1994. Nest trees were taller, had more fishermen’s camps. Temporary, seasonal shelters that unobstructed branches, and a greater arc of acces- were not used during the fish-eagle breeding season were sibility than reference trees (Table 2). Mean nest- not considered buildings. There were no paved roads tree DBH was more than twice that of reference and few motor vehicles in the area, and the most traveled roads were traversed by less than one motor vehicle per trees. Twenty-two of 24 (91.7%) nest trees versus day, even in the dry season. Oxcarts frequentlywere used only 14 of 24 (58.3%) reference trees had a >270° to transport materials, so we recorded any oxcart track arc of accessibility. acessWsaiebriolcaiodty.nsfirdoemretdhetrseheosre^l6i.n1e mto bheigphotaenndtiawlitpher^c3h0°treaecs- = N7e)s,t-CtorredeylsapemcaideasgaisnccalriuednesdisTa{mNar=in4d)u,sAidnadniscoan{iNa based on the smallest recorded perch tree used by Bald sp. {N — 2), Colvillea racemosa {N — 2), Neobeguea Eagles on the Chesapeake Baym(Buehler et al. 1992). We mahafaliensis {N = 2), Acacia sp. {N = 1), Albizia mcounted perch trees within 50 ofthe water along a 250 greveana {N = 1), Alleanthus greveanus {N =1), Foe- encsehotrreelein(eChsaencdtlioenr cetenatl.er1e9d95o).nWteheclnaessstiftireedemoeranrefseurr-- tidia sp. {N = 1), Pandanus sp. {N = 1), and un- rounding canopy height to 5-m intervals ranging from 0- identified {N = 2). T indica was, the most frequent- m = 25 based on visual observation. ly recorded species of nest reference tree {N 6). We measured the perch tree that we saw fish-eagles use Its proportion among nest trees (29.2%) was not most frequently for foraging for each of the 29 fish-eagle pairs in the study area. Eleven (37.9%) of the pairs were different from its proportion among reference - = P = observed for at least 6 hr, at least once per week during trees (20.8%) (x^ 0.44, df I, 0.51). Pro- the breeding season (May-October) in 1992, 1993, and portions of nest trees and reference trees in each 1994 as part of a related study (Watson et al. 1999). The growth form class were similar (x^ — 4.58, df — 4, remaining 18 (62.1%) pairs were observed for at least 6 P — 0.33). Eight of the nest trees (33.3%) and hr, at least three times per breeding season from 1992- 94. We measured the same tree characteristics for perch three (12.5%) of the reference trees were decidu- = = P= trees that we measured for nest trees. ous (x^ 2.95, df I, 0.09). We tested the null hypothesis ofno difference between Eish-eagle nest-tree use was positively associated trees or sites used by breeding Madagascar Fish-Eagles with tree height, producing a logistic regression and reference trees or sites for each of the numerical variables using the Wilcoxon signed-ranks test. We paired model of each fish-eagle nest or perch tree with the randomly se- lteescttefdorredfiefrfeenrcenecterseeinondistthaensceamteo wwaatteerrbboedcya.uWseetdhiids wnaost 0 = 1/1+ exp 5.52 - X 0-38x, a criterion for selecting reference trees. We used the chi- square test ofequal proportions to determine iffish-eagle habitat use was different from expected use for the fol- where 0 is the probability of fish-eagle use and x, lowing categorical variables: tree species, deciduous ver- is the height of tree i. This model correctly classi- sus evergreen trees, growth form, and surrounding can- fied 83.3% of 48 trees measured. opy height. If >20% ofexpected valueswere <5, we used Nest-site Characteristics. Number of shoreline the likelihood ratio chi-square test statistic (Agresti perch trees was greater at nest sites than at random 1990). We developed logistic regression models to predict the sites (Table 3). There was a positive relationship 290 Berkelman et al. VoL. 36, No. 4 Table 2. Characteristics of Madagascar Fish-Eagle nest trees, perch trees, and paired reference trees in the region ofAntsalova, Madagascar in 1994. Paired Paired Reference Reference Nest Trees Trees Perch Trees Trees (V = 24) (N = 24) {N= 29) (N = 29) X ± SE X ± SE X ± SE X ± SE Variable (Range) (Range) pa (Range) (Range) pa DBH (cm) 87.8 ± 11.8 38.4 ± 4.2 <0.001 65.3 ± 7.2 36.9 ± 3.3 <0.001 (29-245) (22-114) (27-270) (21-120) Height (m) 18.7 ± 0.8 10.5 ± 0.9 <0.001 16.7 ± 0.8 9.8 ± 0.4 <0.001 (10.7-25.9) (5.0-23.3) (9.4-30.3) (4.9-15.8) No. of branches’^ 5.5 ± 0.7 3.2 ± 0.8 0.021 7.9 ± 1.2 1.8 ± 0.4 <0.001 (1-14) (0-19) (2-39) (0-15) Arc of accessibility 346.7 ± 5.4 260.2 ± 25.0 <0.001 336.7 ± 7.1 231.4 ± 21.4 <0.001 (265-360) (0-360) (190-360) (0-360) *Wilcoxon signed-ranks test significance level. Number of branches in the tree canopy >5 cm in diameter and unobstructed for 1 m above and below. Arc (0°-360°) that was unobstructed by other trees ^10 m of the trunk and ^3 m below the crown (Buehler et al. 1992). between fish-eagle nest-site use and the number of 1), Adansonia sp. (N = 1), Cedrelopsis grevei {N ~ — = shoreline perch trees. The model was 1), Pandanus sp. (N 1), Raphia sp. {N 1), and unidentified {N — A). T. indica was the most fre- 9 = 1 1 + exp 3.49 - 2 quently recorded perch reference tree species {N = 10). Its proportion among perch trees (10.3%) was smaller than among reference trees (48.3%) where 0 is the probability of fish-eagle use and = = P = is the number of perch trees within a 1.25 ha (50 (x2 5.96, df 1, 0.02). m Perch trees and reference trees had similar wide) shoreline section centered on the point on the shoreline nearest nest tree Correct clas- growth forms (x^ = 8.04, df = 4, P = 0.09). Pro- i. portion of deciduous trees among perch trees sification offish-eagle use for this model was 72.9% (34.5%) was greater than among reference trees of 48 sites. Minimum distance to human distur- = = P = bance, minimum distance to nearest road, mini- (10.3%) (x^ 4.86, df 1, 0.03). mum distance to nearest building, and minimum There was a positive association between fish-ea- gle perch-tree use and tree height, producing a lo- distance to nearest fish-eagle nest did not differ gistic regression model of between nest sites and random sites (Table 3) The . m proportion of nest sites in each 5 canopy height interval did not differ between nest sites and ran- 0 = 1 1 + exp 8.68 2 dom sites (x^ = 4.93, df = 4, P = 0.30). Mean m = distance to water of nest trees was 70.8 (SE where 0 is the probability of fish-eagle use and = 12 6, range 6.8-199.2 m). is the height of tree i. This model correctly classi- Perch-tree Characteristics. Perch trees were larg- fied 84.5% of 58 trees measured. er (DBH and height), had more unobstructed branches, and had a greater arc of accessibility Disclissicjn than reference trees (Table 2). Twenty-six of 29 Nest-tree Use. Madagascar Fish-Eagles used nest (89.7%) nest trees versus only 16 of 29 (55.2%) trees that were taller and had a greater DBH, more reference trees had a >270° arc of accessibility. unobstructed branches, and a greater arc of acces- Perch-tree species included Colvillea racemosa (N sibility than reference trees. The suhstantial differ- — = 5), Ficus cocculifolia (N 4), Neobeguea mahafal- ence between nest trees and reference trees in lensis (N = 3), Tamarindus indica (N = 3), Albizia mean height and DBH suggests that the fish-eagle lebbeck (N — 2), Borassus madagascariensis {N —2), selects nest trees from among the largest trees = ~ Cordyla madagascariensis (N 2), Acacia sp. {N available near water. By placing its nests in the tops December 2002 Madagascar Fish-Eagi.e Nests 291 Table 3. Characteristics of Madagascar Fish-Eagle nest sites {N = 24) and paired reference sites (N — 24) in the region ofAntsalova, Madagascar in 1994. Nest Sites Paired Random Sites x± SE X ± SE pa Variable (Range) (Range) Minimum distance to human disturbance'^ (km) 0.8 ± 0.2 0.9 ± 0.1 0.742 (0-2.8) (0-2.8) Minimum distance to building (km) 1.8 ± 0.4 1.8 ± 0.3 0.814 (0.1-7.7) (0-5.6) Minimum distance to road (km) 1.7 ± 0.4 1.3 ± 0.3 0.055 (0-8.4) (0-5.4) Minimum distance to fish-eagle nest (km) 4.8 ± 0.9 4.3 ± 0.9 0.104 (1.3-20.3) (0.4-20.1) Number of perch trees'^ 30.8 ± 2.3 16.6 ± 1.9 <0.001 (10-53) (0-33) Wilcoxon signed-ranks test significance level. '’Human disturbances included agricultural clearings, rice paddies, villages, tombs, and fishermen’s camps. Number of perch trees within a 1.25 ha (50 m wide) shoreline section centered on the point on the shoreline nearest the nest tree We considered trees that we estimated to have a height ^6.1 m and >30° accessibility from the shoreline to be perch trees. of these trees, it maximizes accessibility and visibil- of the species’ range (Rabarisoa et al. 1997). Al- ity for foraging and territorial defense. These re- though our sample size was limited, the 29 breed- sults were consistent with those reported for other ing sites sampled represent 29.3%, ofthe 99 known nesting Haliaeetus species (McEwan and Hirth remaining Madagascar Fish-Eagle breeding sites 1979, Andrew and Mosher 1982, Anthony and (Rabarisoa et al. 1997). Bald Eagle management Isaacs 1989, Shiraki 1994). guidelines recommend conserving mature forest Nest-site Use. Number of shoreline perch trees around existing and potential nest sites (Anthony was the only variable that differed between nest et al. 1982, Wood et al. 1989). We offer guidelines sites and random sites. This suggests that the Mad- that are more specific to the range oftree sizes and agascar Fish-Eagle, like the Bald Eagle (Chandler densities found in the tropical dry forest and sa- et al. 1995), may avoid areas without a sufficient vanna habitats that surround the lakes where Mad- number of foraging perches. agascar Fish-Eagles occur. Perch-tree Use. Perch trees were larger in height We recommend that areas with a >32/ha density m and DBH, and had more unobstructed branches, of trees >15 tall should receive high priority for and had a greater arc of accessibility than refer- Madagascar Fish-Eagle conservation. Probability ence trees. Such trees probably have greater access that a shoreline tree would be used by Madagascar and provide better visibility over water than other Eish-Eagles for nesting or perching can be calcu- trees. This is consistent with Bald Eagle perch-tree lated by inserting tree height into the correspond- use (Stalmaster and Newman 1979, Steenhof et al. ing logistic equation (Fig. 1). Similarly, number of 1980, Buehler et al. 1992). Madagascar Fish-Eagle perch trees along a 1.25 ha (250 X 50 m) shoreline perch trees were more often deciduous than ref- section can be used to estimate the probability that erence trees. In contrast with the nest-tree results, Madagascar Fish-Eagles will use the shoreline sec- the fish-eagles in this study appeared to avoid T. tion for nesting (Eig. 1). These models are best indica for perching. T. indica is evergreen and often used under the conditions that were present dur- has a dense crown; therefore fish-eagles may use ing this study (e.g., same eagle population density, this species less often for perching than leafless same time of year) and apply to eagles nesting on trees or snags. lakes, rivers, and estuaries. Model Applications. The models we developed Presence of tall trees close to shoreline is the may be used to identify Madagascar Fish-Eagle best predictor of Madagascar Eish-Eagle nest-site nesting and perching habitat along lakes, rivers, use. The eagles often used the tallest trees near and estuaries in western Madagascar. They do not water both for nesting and for foraging perches. apply to a sub-population of at least 16 fish-eagle Rabarisoa et al. (1997) conducted Madagascar pairs that nest on offshore islands at the north end Fish-Eagle surveys from 1991-95, and found areas 292 Berkelman et al. VOE. 36, No. 4 Use Nest-tree Height (m) of Probability Perch-tree Height (m) 0 S 10 15 20 25 30 35 40 45 50 No. of Perch Trees Figure 1. Probability of Madagascar Fi.sb-Eagle use of iiesl trees, perch trees, and nest sites as a I’unclion of nest- tree height (A), perch-tree height (B), and number of shoreline perch trees (C), in the region ol Antsalova Mada- gascar, 1994. Probabilities were calculated by inserting dilferent values of the explanatory variable (tree height or number of perch trees) into the equation resulting from stepwise logistic regression analysis. Dec.ember 2002 Madagascar Fish-Eagle Nests 293 with dense forest adjacent to water that were un- perch habitat on the northern Chesapeake Bay. Wil- occupied by fish-eagles. Watson et al. (1996) are son Bull. 104:540-545. developing means to augment the hsh-eagle pop- Chandler, S.K., J.D. Fraser, D.A. Buehler, andJ.KD. Si:egar. 1995. Perch trees and shoreline development ulation and seek areas of unoccupied fish-eagle as predictors of Bald Eagle distribution on Chesa- habitat where young eagles may be released. Our peake Bay. y. Wildl. Manage. 59:325-332. models may be used both to identify areas of suit- Donque, G. 1972. The climatology of Madagascar. Pages able, but unoccupied, fish-eagle habitat and high 87-144 in R. Battistini and G. Richard-Vindard [Eds ], conservation priority areas of occupied habitat. Biogeography and ecology ofMadagascar. Dr. W.Junk The Tsimembo Forest surrounding Lakes Befo- B.V., The Hague, Netherlands. taka, Soamalipo, and Ankerika, where the highest Hosmer, D.W., Jr. and S. Lemeshow. 1989. Applied lo- gistic regression. John Wiley and Sons, New York, NY density of fish-eagles is found (Rabarisoa et al. U.S.A. 1997), should receive highest conservation priority. Keister, G.P., Jr. and R.G. Anthony. 1983. Characteris- The human population density around the lakes tics of Bald Eagle communal roosts in the Klamath was low until recent years when large numbers of Basin, Oregon and California. Wildl. Manasce. 47 J. fishermen began to migrate to the region (Watson 1072-1079. and Rabarisoa 2000). Increased harvesting of tall Langrand, O. and B.-U. Meyburg. 1989. Range, status, shoreline trees by migrant fishermen will have a and biology of the Madagascar Sea-Eagle Haliaeetus negative impact on the fish-eagles. People use the vociferoides. Pages 269-277 in B.-U. Meyburg and R.D. Chancellor [Eds.], Raptors in the modern world. tallest trees available for dugout canoes and build- WWGBP, Berlin, Germany. ing materials (Watson and Rabarisoa 2000) and McEwan, L.C. and D.H. Hirth. 1979. Southern Bald Ea- may prevent regeneration of tall trees by harvest- gle productivity and nest site selection. Wildl. Man- J. ing large amounts of fuel wood to preserve fish by age. 43:585-594. smoke drying. Deforestation probably has already Meyburg, B.-U. 1986. Threatened and near-threatened substantially reduced the amount offish-eagle hab- birds of prey of the world. Birds Prey Bull. 3:1-12. itat available, and as the human population contin- Nelson, R. and N. Horning. 1993. AVHRR-LAC esti- mates offorest area in Madagascar, 1990. Int.J. Remote ues to increase, available habitat will continue to Sens. 14:1463—1475. decrease unless steps are taken to conserve fish- Newton, I. 1979. Population ecology of raptors. Buteo eagle habitat. Books, Vermillion, SD U.S.A. Rabarisoa, R., R.T. Watson, R. Thorstrom, and Ber- J. At^KNOWITDGMENTS kelman. 1997. Status of the Madagascar Fish-Eagle The Peregrine Fund provided financial and logistical Haliaeetus vociferoides in 1995. OVncA 68:8-12. support for this research. We thank C. Razafimahatratra, Shiraki, S. 1994. Characteristics of White-tailed Sea-Eagle nest sites in Hokkaido,Japan. Condor96:1003-1008 G. Raoelison, Mampiandra, and L. Kalavah for help with data colleJc.tion. Thanks to Rajesy, R. Rabarisoa, R. Stalmaster, M.V. andJ.R. Newman. 1979. Perch-site pref- J. erences of wintering Bald Eagles in northwest Wash- I.ewis, P. Ravonjiarisoa, and M. Razafindrakoto for ad- ington./. Wildl. Manage. 43:221-224. ministrative and logistical support in Antananarivo and Steenhof, K., S.S. Berunger, and L.H. Fredrickson. in the field. We thank C.A. Haas, Ney, R.G. Oderwald, J.J. 1980. Habitat use by wintering Bald Eagles in South D.F. Stauffer, R. Thorstrom, R. Tingay, and A.R. Harmata Dakota. Wildl. Manage. 44:798-805. for comments on the manuscript. Watson, R.JT.. and R. Rabarisoa. 2000. Sakalava fisher- men and Madagascar Fish-Eagles: enhancing tradi- farERAit]RE Cited tional conservation rules to control resource abuse AciRESTi, A. 1990. Categorical data analysis. John Wiley that threatens a key breeding area for an endangered and Sons, New York, NYU.S.A. eagle. Ostrich 71:2-10. Andrew,J.M. andJ.A. Mosher. 1982. Bald Eagle nest site , S. Razaeindramanana, R. Thorstrom, and S. Ra- FANOMEZANTSOA. 1999. Breeding biology, extra-pair selection and nesting habitat in Maryland. Wildl. J. birds, productivity, siblicide, and conservation of the Manage. 46:383-390. Madagascar Fish-Eagle. Ostrich 70:105-111. Anthony, R.G. and F.B. Isaacs. 1989. Characteristics of S. Tiiomsett, D. O’Daniel, and R. Lewis. 1996 Bald Eagle nest sites in Oregon. J. Wildl. Manage. 53: Bree,ding, growth, development, and management of 148-159. the Madagascar Fish-Eagle {Haliaeetus vociferoides) J R.L. Knight, G.T. Ai.ien, B.R. McClelland, and , Raptor Rei 30:21-27. J.I. Hodges. 1982. Habitat use by nesting and roosting Wood, P.B., T.C. Edwards,Jr., and M.W. Collopy. 1989. Bald Eagles in the Pacific Northwest. Trans. N. Am. Characteristics of Bald Eagle nesting habitat in Flori- Wildl. Nat. Resour. Conf. 47:382-390. da./. Wildl. Manage. 53:441-449. Buehler, D.A., S.K. Chandler, T.J. Meicsmann,J.D. Fras- er, ANDJ.K.D. Seegar. 1992. Nonbreeding Bald Eagle Received 30 November 2001; accepted 9 July 2002