OUP user menu

Food Habits of an Endangered Carnivore,Cryptoprocta Ferox, in the Dry Deciduous Forests of Western Madagascar

Clare E. Hawkins , Paul A. Racey
DOI: http://dx.doi.org/10.1644/06-MAMM-A-366.1 64-74 First published online: 19 February 2008


We describe the diet of fossas (Cryptoprocta ferox) in a dry deciduous forest of western Madagascar from 376 scats collected between June 1994 and September 1996, from which 554 prey items were identified. More than 90% of these were vertebrates, and more than 50% were lemurs. No other nonprimate mammal includes such a high proportion of primate items in its diet. The principal prey comprised approximately 6 lemur species and 2 or 3 spiny tenrec species, along with snakes and small mammals. Significant differences were apparent in the composition of the scats in wet and dry seasons, with a higher proportion of Tenrec in the former, and fewer lemurs. Within the confines of a diet of vertebrates, fossas appeared to be opportunistic predators. For those prey types for which data were available, a significant relationship was found between the estimated relative number of individuals taken of any one type of prey and its abundance. Fossas were estimated to remove up to 19% of their prey populations per year. This high impact suggests that they were living close to the maximum population density possible on the available prey. Species of a wide range of body masses were included in the diet. Verreaux's sifaka (Propithecus verreauxi), weighing more than one-half of the body mass of the fossa, constituted approximately 11% of the prey biomass.

Key words
  • Cryptoprocta ferox
  • food habits
  • fossa
  • Kirindy
  • lemurs
  • Madagascar
  • tenrecs

To understand the ecological function of tropical forest Carnivora, many of which are threatened, it is important to know the extent to which their diets are opportunistic or specialized. Relatively little is known about the feeding ecology of this group (Schaller 1996) because they occupy relatively impenetrable habitat through which they may travel large distances. Emmons (1987), in her study of neotropical rain-forest felids, observed that this habitat appears to support only highly opportunistic, solitary predators. To our knowledge, there have been no similar published studies of carnivores (sensu Gittleman et al. 2001) in dry tropical forests.

The endangered fossa (Cryptoprocta ferox) reaches its highest densities in dry deciduous forests in Madagascar (Hawkins 2003), where it is the largest of 8 species of carnivores endemic to the island. As Ewer (1973) observed, its shortened braincase, together with its dentition, which includes enlarged carnassials and only 1 small upper molar, suggests that this species specializes in carnivory. Fossas were predicted on the basis of their body mass to take prey of an approximate mean body mass of 0.7 kg (Peters 1983).

Rasolonandrasana (1994) indicated that in terms of percentage biomass, lemurs were the predominant taxon in the diet of fossas. However, in terms of the percentage number of individuals consumed, introduced rodents were the most abundant diet in 1 of his 2 study areas, a montane humid forest. Rasoloarison et al. (1995) concluded that fossas were not entirely opportunistic, but preferred larger lemurs to smaller ones, and larger rodents to smaller ones, disproportionate to their relative abundances. Rasoloarison et al. (1995) calculated, as a very conservative estimate for their study site, that the number of Verreaux's sifakas (Propithecus verreauxi) taken annually by fossas approximately equated to 29% of the yearly population growth of sifakas.

Goodman et al. (1997b) provided evidence for flexibility in the choice of prey of fossas by examining scats collected above the treeline in a zone of ericoid meadows and exposed rock outcrops, at elevations ranging from 1,950 m to 2,600 m above sea level. These scats contained prey of a mean body mass less than one-half that of those identified by Rasolonandrasana (1994), with much higher proportions of the prey items identified (82 in total) being birds and insects. Only 3 prey items identified were lemurs. However, some of the prey items recorded were species not found above the treeline, indicating that the fossas did not subsist entirely in the area where the scats were collected. The flexibility of diet indicated by this study thus applies to fossas at the limits of their altitudinal range.

Rasolonandrasana's (1994) data were derived from only 146 scats, all but 10 of which were taken from the dry season, over a period of 7 months, and from 2 very different sites. The data of Goodman et al. (1997b) were derived from only 20 scats and those of Wright et al. (1997) were from only 8 scats and 3 corpses.

The aims of the present study were to produce a detailed description of the year-round diet of fossas in the western dry forest of Madagascar, to identify factors affecting the composition of the diet, and to assess the impact of fossas on populations of their prey. Since our study was originally carried out, a similar study (Dollar et al. 2006) has been carried out at another western dry forest site. This 2nd study serves as a useful source of comparison for our work.

Materials and Methods

Study site.—The study took place in Kirindy Forest (44°41'E, 20°04'S). Also known as Kirindy/CFPF (Centre de Formation Professionelle Forestiè re) and described in more detail in Hawkins and Racey (2005), this forest lies 43 km north of the town of Morondava, and 23 km from the west coast of Madagascar at an elevation of 18–40 m above sea level (Sorg and Rohner 1996). It extends to 100 km2 and is composed of western dry deciduous forest with a very dense understory (5,000–19,000 stems < 10 cm diameter at breast height per hectare—Rakotonirina 1996).

Three species of carnivores are reported from Kirindy Forest: the fossa, the endemic narrow-striped mongoose (Mungotictis decemlineata), and the introduced small Indian civet (Viverricula indica), although the latter is rarely seen. Eight species of primates inhabit the forest, including the recently named pygmy mouse lemur (Microcebus berthaeRasoloarison et al. 2000). Ganzhorn and Kappeler (1996) report that the density of primates in Kirindy is one of the highest in the world. Other endemics occurring in the forest include the giant jumping rat (Hypogeomys antimena) and the flat-tailed tortoise (Pyxis planicauda).

The climate of the region is extremely seasonal. The year is divided into the wet season, occurring from October–November until April–May, and the dry season. The latter is characterized by a wide daily range of temperature, and little or no rain, whereas the wet season has a more constant, higher temperature. Mean monthly temperatures recorded throughout the study by resident researchers included monthly minima ranging from 11 °C to 22°C, and monthly maxima from 31°C to 37°C. The most extreme temperatures recorded during the study period were 4.0°C (in June 1996) and 40.7°C (in November 1995). The rain falls almost exclusively in the wet season, but is intermittent and increasingly variable from year to year; it is concentrated mostly in January (when the monthly peak of 296 mm was recorded in 1995) and February, but even during these months there may be periods of several days without rain.

Collection of scats.—Scats were collected opportunistically throughout the study period, from June 1994 to September 1996, except during October 1994, March–May 1995, September 1995, and most of April 1996. During the wet season, scats were collected primarily during the earlier drier months. Fossa scats were black or gray cylinders with twisted ends. They usually consisted of 1 or 2 such cylinders, typically 10–14 cm long and 1.5–2.5 cm wide. Most were largely composed of hair. Fresh scats had a strong, musky smell. They were easily distinguishable from scats of other carnivores in Kirindy.

All fossa scats encountered within the study area were collected, and the date of collection was recorded. Fossa scats were most commonly found on broad tracks, most likely because they were exposed and easy to spot there, and because most time was spent on tracks. A few were found around a lake, scattered throughout the sparse forest of this area, and some were collected on the dry bed of the Kirindy River, particularly near a fossa mating site. Many were also gathered on narrow paths cut by researchers studying lemurs.

Fecal analysis.—Most scats that were collected were already dry, but those collected in the height of the wet season were dried before storage. For analysis, the scats were broken up, and the few, small hard parts (bones and shell) picked out. In Kirindy, this was carried out under a transparent cover and wearing masks, to avoid risk of parasitic infection (Reynolds and Aebischer 1991). All scats were subsequently taken to Aberdeen, United Kingdom, where they were autoclaved before being further examined.

Hairs were identified using a reference collection that was accumulated during fieldwork with the assistance of the other researchers. Given the endangered status of most of the prey species of the fossa, it was not permissible to set up a reference collection of bones, but the character and size of the fragments of bone, claw, teeth, and shell were used to aid identification of the prey species. In addition S. M. Goodman, Field Museum of Natural History, Chicago, Illinois, examined the bones from a subsample of 41 of the scats collected, identifying them with the aid of a reference collection in the Department of Palaeontology at the University of Antananarivo. To quantify the error in identification, results obtained through identification of hair were compared with Goodman's results. In the present study, prey type refers to a prey species or group of prey species. Prey species were grouped if their remains in the scats could not be distinguished from one another.

Data analysis.—Unless there was evidence of more than 1 individual in a single scat, each prey type identified in a single scat was scored as 1 item. Relative frequency of any 1 prey type in a sample of scats was expressed as: relative frequency

Data were examined and analyzed in contingency tables for seasonal and yearly changes in scat composition. Dry-season scats were defined as those collected from May to October, with the rest defined as wet-season scats, collected from November to April.

Fossas trapped in Tomahawk Bobcat traps (Tomahawk Live Trap Co., Tomahawk, Wisconsin) for population estimates and radiotracking studies often defecated, and these scats also were analyzed. The findings were used to test for any relationship between prey type and age (juvenile or adult) and sex of the trapped animal, using chi-square tests. Adult males were distinguished from juveniles by body mass (> 5.5 kg) and testis size (anteroposterior length > 25 mm); adult females were distinguished from juveniles by body mass (> 5 kg) and suckled mammae. Totals of each prey type found in scats of adult males, adult females, juvenile males, and juvenile females, respectively, were calculated. Where scats were collected from the same individual more than once, the mean numbers of each prey type found in each of that individual's scats were used. Totals were rounded up to integers to comply with the chisquare test, and prey types were grouped together to analyze the resulting small sample. To investigate whether prey size affected the composition of scats of fossas of different age or sex, prey types were grouped into those greater than or less than 300 g mean body mass. To investigate whether lemurs featured less in the diet of juvenile fossas, the prey types, in a further analysis, were grouped into lemur and nonlemur.

The relative frequency of any 1 prey type in the scats will relate not only to numbers of individuals of that type consumed, but also to the size of the prey involved. A correction for this effect was therefore attempted, to explore the importance of this factor. Because the main cues for identification of prey were fur, spines, or scales, the ideal correction factor would be in terms of surface area, rather than mass, because this allows for the greater surface area: volume ratio (i.e., fur: volume ratio) in smaller prey. Accordingly, the data were reanalyzed by correcting for prey surface area, with a view to obtaining an indication of numbers of prey individuals consumed.

If it is assumed that the density (i.e., mass per unit volume) of the bodies of all prey species was constant, then volume may be treated as proportional to body mass. A factor approximately proportional to surface area may therefore be obtained using body mass to the power of two-thirds. Thus, corrected incidence of any 1 prey type was expressed as: corrected incidence Embedded Image

Corrected relative frequency of any 1 prey type, relative to those of other prey types in a specified sample, was expressed as: corrected relative frequency Embedded Image

The surface area of the tails of lemurs was ignored because in the few lemur carcasses clearly killed by fossas found during the present study, the tail was consistently largely uneaten (C. E. Hawkins and P. A. Racey, in litt.; R. Warren, Natural England, Peterborough, United Kingdom, pers. comm.).

To calculate mean body mass of prey (for comparison with the prediction made on the basis of the body mass of fossas), the relative frequency (r) of each prey type (p) for which mass was known was multiplied by its mean body mass (w), summed and divided by the total percentage of prey items for which these prey types accounted: Embedded Image

For prey types comprising several species, a mean body mass was calculated for the type by multiplying the mean body mass (w) of each constituent prey species (s) by its abundance (a), summing these, and dividing them by the sum of their abundances: Embedded Image

Standard deviation was calculated on the same principle. The same calculation also was made replacing the relative frequency of each prey type with the corrected relative frequency, given that the latter frequency may be a better estimate of the proportions of individuals of each species consumed.

For each prey type for which there was sufficient information uncorrected numbers of prey items were tested against the effects of the following attributes of each prey type: mean body mass, average abundance, biomass (= abundance × body mass), daily activity rhythm (nocturnal, diurnal, or crepuscular), and seasonal activity (hibernating, regularly torpid, or never torpid). These data and their sources are given in Table 1. For attributes daily activity rhythm and seasonal activity, the Kruskal–Wallis test was applied; for the others, least-square regression analyses were used. The same analyses were repeated, using corrected incidences of prey items of each prey type in place of the uncorrected numbers.

View this table:
Table 1

Description of prey types identified in fossa scats during the present study. All data were collected from the Kirindy Forest. Biological data on lemurs are from Ganzhorn and Kappeler (1996—abundances calculated from standardized transect walks), except for Microcebus seasonal activity (Schmid and Kappeler 1998). All data on mammals other than lemurs are from Ganzhorn et al. (1996), except for Hypogeomys abundance (S. Sommer, pers. comm.). Data on birds are from Hawkins and Wilmé (1996) and all data on reptiles are from Bloxam et al. (1996). All species specifically mentioned are endemic to Madagascar, except for Rattus rattus. Daily activity pattern: N = nocturnal; D =diurnal; C = cathemeral (diurnal and nocturnal). Seasonal activity: H = extended periods of inactivity in the dry season; T = becomes torpid during cold nights; N = no periods of torpor.

Prey typeaPotential prey species includedFamilyCommon nameBody mass (g)Daily activity patternSeasonal activityAbundance (individuals/km2)
TenrecTenrec ecaudatusTenrecidaeTail-less tenrec1,500NH
Ech/SetEchinops telfairibTenrecidaeLesser hedgehog tenrec170NH
Setifer setosusbTenrecidaeGreater hedgehog tenrec220NH
PropithecusPropithecus verreauxiIndriidaeVerreaux's sifaka3,600DN38
EulemurEulemur fulvusLemuridaeBrown lemur2,300CN33
L/Ph/MzLepilemur ruficaudatusLepilemuridaeRed-tailed sportive lemur750NN195
Phaner furciferCheirogaleidaeFork-marked lemur450NN51
Mirza coquereliCheirogaleidaeCoquerel's dwarf lemur300NN12
CheirogaleusCheirogaleus mediusCheirogaleidaeFat-tailed dwarf lemur280NH264
MicrocebusMicrocebus murinusbCheirogaleidaeGray mouse lemur60NH/T171
Microcebus berthaebPygmy mouse lemur30NH/T171
Small mammalsSuncus madagascariensisbSoricidaePygmy shrew1.7N
Geogale auritabTenrecidaeLarge-eared tenrec4N
2 Microgale spp.bTenrecidae9N
Rattus rattusMuridaeRoof rat100NN
Eliurus myoxinusbNesomyidaeWestern tufted-tailed rat70N
Macrotarsomys bastardibNesomyidaeLesser big-footed mouse30N
HypogeomysHypogeomys antimenaCricetidaeGiant jumping rat1,100NN56
MungotictisMungotictis decemlineataEupleridaeNarrow-striped800DN
Birds68 speciesbAlmostN
all D
Snakes2 speciesbTyphlopidaeD/NH/N
2 speciesbBoidae
15 speciesbColubridae
LizardsTotal 23 speciesbGekkonidaeD/NH/N
Pyx/PelPyxis planicaudabTestudinaeFlat-tailed tortoiseDH
Pelomedusa subrufabPelomedusidaeFreshwater turtleDH
  • a Ech/Set = Echinops–Setifer; L/Ph/Mz = Lepilemur–Phaner–Mirza; Pyx/Pel = Pyxis–Pelomedusa.

  • b These species were not individually identified in the scats but they are the only ones of the relevant prey type that were present in the area.

Estimates of prédation impact on prey populations could only be made for those prey types (p) for which abundance in Kirindy was known, and thus only the proportion of the annual biomass consumed which was made up of these prey types could be considered. According to Albignac (1970, 1975) and T. Hornsey (pers. comm.) of Suffolk Wildlife Park, a captive fossa consumes 0.5–1 kg/day of either pure meat or part plucked and boned meat. Recognizing the greater energy requirements of wild animals and allowing for some extra weight of fur or feathers and bones, we have taken the annual food requirement of a single fossa to be 365 kg/year. To calculate the proportion of the 365 kg comprising these prey types, the uncorrected relative frequencies (r) of these prey types (taken as an indication of percentage biomass consumed of each prey type), were summed (Σ(r)). Thus: Embedded Image

This mass is composed of a multiple of the sum of the percentage of individuals of each prey type consumed multiplied by the mean body mass of that prey type. Assuming that corrected frequencies of prey types in the scats were equivalent to the proportion of individuals of each prey type consumed, then the mass [365 Σ(rp)] kg was composed of a multiple (m) of the sum of, for each prey type, its corrected frequency in the scats (c) multiplied by its mean body mass (w): Embedded Image or Embedded Image

By this reasoning, the total number of individuals of all prey types (p) consumed in a year by a fossa will be m E(cp), and for any 1 prey type, a total of (m × c) individuals will be consumed. Taking the population density estimate of fossas of 0.26 individuals/km2 in Kirindy (Hawkins and Racey 2005), then: Embedded Image and Embedded Image


Three hundred seventy-six scats were collected, from which 554 prey items were identified into 18 different types. Details of potential prey, and the resolution to which they were identified, are given in Table 1. Of these, 111 scats were collected during the wet season, although few (13) were obtained during the height of this season (January#x2013;March).

Of the 41 scats whose bone fragments were examined by S. M. Goodman, and the other remains examined using the methods described for the present study, discrepancies between results occurred in only 8 scats. In 2 of these cases, a scat that was identified by one method as containing Phaner was identified by the other method as containing Lepilemur. In 2 other cases, there was similar disagreement between Mirza and Lepilemur. By grouping these 3 species together as a single prey type, the discrepancy between the results generated by the 2 methods was reduced to 10%.

The relative frequencies of the different prey types found in the 376 scats analyzed are shown in Fig. 1. Ninety-four percent of the items identified were of vertebrate origin. The only vegetable matter identified was of seeds of several different taxa. More than 50% of all items were identified as lemur. The medium-sized lemurs (Lepilemur,Phaner, and Mirza) were by far the most commonly identified items in the scats, accounting for 32% of items identified. Although it was difficult to distinguish between the remains of these 3 species, the majority of identifications of medium-sized lemurs appeared to be either Lepilemur or Mirza. Propithecus and Cheirogaleus also were commonly recorded. However, the relative frequency of Eulemur was very low compared with those of the other lemurs. Other prey types with high relative frequencies were Tenrec and snakes.

Fig. 1

Relative frequencies of different prey types found in 376 fossa (Cryptoprocta ferox) scats from Kirindy, Madagascar. Relative frequency of a prey type is defined as the number of items identified of the prey type, expressed as a percentage of the total prey items identified in all scats; prey item is defined as a single prey type identified in a single scat. Prey types are defined in Table 1. Pyx/Pel = Pyxis–Pelomedusa; L/Ph/Mz = Lepilemur#x2013;Phaner#x2013;Mirza; Ech/Set = Echinops#x2013;Setifer.

Scat composition was compared between seasons (Fig. 2). Significant differences in composition were found between scats collected in the wet season and those collected in the following dry season (ϰ2 = 39.079, d.f. = 8, P < 0.0001 for wet season 1994–1995, dry season 1995; ϰ2 = 32.652, d.f = 8, P < 0.0001 for wet season 1995–1996, dry season 1996). Comparisons between wet seasons and the preceding dry seasons (i.e., incorporating dry season 1994, which was omitted in the previous analysis) also gave significant chi-square values. The individual chi-square values for each prey type for the 2 years is given in Table 2. For the contingency tables, to avoid excessive proportions of cells with low expected values, prey types with low incidence in the scats were grouped together: “other large mammals” includes Eulemur, Hypogeomys, and Mungotictis; “other small mammals“ includes the types EchinopsSetifer and small mammals; “other nonmammals” includes birds, lizards, PyxisPelomedusa, and insects. One prey type with an overall relative frequency of 5%, seeds, was ignored for the contingency tables; these were assumed to have arrived in the scats via the stomachs of the lemurs consumed as prey, rather than being a food specifically selected by fossas.

Fig. 2

Relative frequencies of different prey types found in 111 fossa (Cryptoprocta ferox) scats from Kirindy, Madagascar, collected during wet seasons (1994#x2013;1995 and 1995#x2013;1996) compared with those found 265 scats collected during dry seasons (1994, 1995, and 1996). Definition of relative frequencies as in Fig. 1. Abbreviations as for Fig. 1.

View this table:
Table 2

Individual chi-square values for each prey type arising from 2 contingency tables to compare the composition of scats collected in the dry and wet seasons over 2 years.

Prey type1994–199519951995–19961996
Other large mammals0.0340.0270.0500.021
Other small mammals0.0060.0050.1080.046
Other nonmammals0.1650.1330.0500.021
  • a L/Ph/Mz = LepilemurPhanerMirza.

Almost all of the differences in scat composition between wet and dry season, for both years, were attributable to the much higher relative frequencies of remains of Tenrec identified in the scats found during the wet season. Remains of Propithecus and LepilemurPhanerMirza also featured more commonly in the dry-season scats. Seasonal differences in relative frequencies of other prey types were relatively trivial. Cheirogaleus featured marginally more commonly in the scats from the wet season. Microcebus and snakes both exhibited small differences in relative frequencies between seasons that were inconsistent between years. Other prey types contributed still less to the overall between-season difference in scat composition. Within each season, wet or dry, there was no significant difference in scat composition across years (Σ2=15.886, d.f. = 16, P = 0.46, for the 3 dry seasons; Σ2 = 6.374, d.f. = 6, P = 0.38, for the 2 wet seasons; degrees of freedom were fewer for the wet-season contingency table because, there being fewer data available, these were compiled into fewer groups). Mean ± SD of prey body mass was calculated using uncorrected relative frequencies as 1.10 ± 1.10 kg, and using corrected relative frequencies as 0.51 ±0.81 kg.

Thirty-five scats were collected from 12 adult males, 8 adult females, 3 juvenile males, and 5 juvenile females caught in traps. Their composition was compared to investigate whether age or sex of a fossa was related to the frequency of prey types of > 300 g body mass, or to the frequency of types of lemur prey, in its scats (Table 3). Chi-square tests revealed no significant difference (P > 0.5) between the proportion of large prey found in scats of all trapped males and that found in those of all females, nor in scats of all adults versus those of all juveniles. Furthermore, no significant difference was found between the proportion of lemur and nonlemur prey found in scats of all males and those of all females. However, this was not clear in the case of the proportions of lemur and nonlemur prey found in scats of all adults versus those found in the scats of all juveniles. In this case (Table 3), scats of adults contained 5 times as many lemur prey items as nonlemur ones, whereas scats of juveniles contained equal proportions of lemur and nonlemur prey items (Fisher exact P = 0.0863). The small sample size precludes a rigorous statistical treatment, but it appears that the scats of juvenile fossas may contain a significantly lower proportion of lemur prey items than do thescats of adults.

View this table:
Table 3

Summary of prey items identified in 35 scats collected from trapped fossas, including 12 adult males, 8 adult females, 3 juvenile males, and 5 juvenile females. Mean body masses of prey are shown in Table 1.

Prey < 300 gPrey > 300 gLemurNonlemur
All males613145
All females513135
All adults820235
All juveniles3644

Three of the 4 lemur items identified in the scats of juvenile fossas were medium-sized nocturnal lemurs (LepilemurPhanerMirza), the 4th being a Microcebus. Fifteen of the 28 lemur items identified in the scats of adults were medium-sized nocturnal lemurs, 2 others were Propithecus, and the rest were prey types with smaller body mass (Cheirogaleus and Microcebus).

For all 376 scats collected, the relative frequencies of the different prey types for which mean body mass was known were corrected for surface area (Fig. 3) to give an indication of relative numbers of individuals of each prey type taken by fossas. In the corrected data, as for the uncorrected data, lemurs as a whole were identified at a corrected relative frequency of more than 50% (although in this analysis, fewer prey types were involved). In addition, the spiny tenrecs (EchinopsSetifer and Tenrec) constituted a substantial proportion of the estimated diet, similar to that in the uncorrected data. However, in other respects, the corrected data differed markedly from the uncorrected data. Small mammals were identified as the most common prey type, whereas in the uncorrected data they constituted a relatively rare prey type. Smaller lemurs comprised a much larger proportion of the estimated diet, with Microcebus being the 3rd most common prey type, whereas the largest lemur species in Kirindy, Propithecus, constituted a much smaller proportion than in the uncorrected data.

Fig. 3

Relative frequencies, corrected for prey surface area, of the prey types identified in 376 fossa (Cryptoprocta ferox) scats for which body mass was known. This correction was made to give an indication of relative numbers of individuals of each prey type taken by the fossa. Abbreviations as for Fig. 1.

Scat composition in the dry and wet seasons was compared in terms of the corrected frequencies of prey types for which mass was known (Fig. 4). The correction did not markedly affect the differences in relative frequencies between seasons, but the choice of prey appeared to be more evenly spread across the types than in the uncorrected analysis. Furthermore, in the corrected data for the wet season, neither Tenrec nor the medium-sized nocturnal lemurs were the most frequent prey type; small mammals were identified at the highest corrected relative frequency, with Microcebus and Cheirogaleus also at similar or higher corrected relative frequencies than Tenrec and LepilemurPhanerMirza.

Fig. 4

Relative frequencies of different prey types for which body mass was known, corrected for prey surface area, found in the same fossa (Cryptoprocta ferox) scats collected in the wet and dry seasons and analyzed in Fig. 2. The correction was made to give an indication of relative numbers of individuals of each prey type taken by fossas. Abbreviations as for Fig. 1.

A significant positive relationship was found between corrected incidence in the scats and abundance of prey types in Kirindy (R2 = 0.698, slope = 0.007, P = 0.039; Fig. 5). No relationship was found between uncorrected numbers of prey items and abundance of prey types, nor between either corrected incidence or uncorrected numbers of prey items and body mass, daily rhythm, or seasonal activity of prey types (P > 0.1 in all cases).

Fig. 5

Least-squares linear regression of the corrected incidence of 6 prey types identified in 376 fossa (Cryptoprocta ferox) scats from Kirindy, Madagascar, for which abundance (individuals/km2) was known, against abundance of each of those prey types. Corrected incidence of a prey type is defined as the number of items identified of the specified prey type, corrected for surface area as explained in the text. The correction was made to give an indication of the numbers of individuals of each prey type taken by fossas.

The highest of the estimated impacts of predation were exerted on populations of Microcebus,LepilemurPhanerMirza, and Propithecus (Table 4). It appears that more individuals of LepilemurPhanerMirza were eaten per square kilometer per year than any other prey type, but that the population from which the highest percentage was removed by the fossa each year was Microcebus. The prey types included in this calculation accounted for 61% of the diet (in terms of the sum of their uncorrected frequencies in the scats), and 62% of the diet in terms of corrected frequencies, despite the fact that the latter estimate involved a smaller number of prey types overall. Thus, the error through equating these 2 estimates to calculate predation impact was negligible.

View this table:
Table 4

Estimates of the impact of predation by fossas on some of their prey species, “ c” is the relative frequency of each prey type found in 376 scats, corrected for surface area to estimate relative numbers of individuals eaten. Mean body mass and mean abundance of each prey species were obtained from Ganzhorn and Kappeler (1996), other than that for Hypogeomys, which was provided by Ganzhorn et al. (1996) and S. Sommer (pers. comm). Calculations for the other columns are explained in the text.

BodyAbundanceNo. individualsNo. individuals% population
Prey typec(%)mass (kg)(individuals/km2)eaten/fossa/yeareaten/km2/yeareaten/year
  • a L/Ph/Mz = LepilemurPhanerMirza.


The findings of our study are based on the largest collection of fossa scats collected over the longest period of time yet analyzed from a single area. The composition of the scats remained consistent when compared from year to year, reinforcing the conclusion that the findings represented the typical diet of fossas in Kirindy, and confirming that, regardless of the method of analysis, the diet of fossas is almost exclusively vertebrates, and primarily mammalian. In this respect at least, fossas converge with felids.

The level of carnivory of a species has implications for various aspects of its biology (Harvey and Gittleman 1992). For example, McNab (1963) showed that a carnivorous animal typically has a larger home range than an herbivore of the same body mass. The seeds in the scats were the only form of vegetable matter identified; a likely source of these was the stomachs of lemurs. An alternative explanation is that fossas occasionally eat fruit, perhaps as a water source; seeds appeared to be more prevalent in the diet in the dry season (Fig. 2).

Our study shows clearly that lemurs form a very high proportion of the diet, regardless of method of analysis, confirming the conclusion drawn by Rasolonandrasana (1994) on the diet of the fossa in Kirindy. No other nonprimate mammal is reported to include such a high proportion of primates in its diet. Although forest leopards (Panthera pardus) regularly take primates (Hoppe-Dominik 1984), they are not regarded as primate specialists. Although the intelligence, sociality, and low reproductive rate of primates may render them inaccessible to most predators, they are vulnerable to arboreal predators (e.g., Sunquist and Sunquist 1989). The only other mammals specializing in predation on primates are primates themselves; for example, Stanford et al. (1994) found more than 80% of the prey items of chimpanzees (Pan troglodytes) to be red colobus monkeys (Procolobus verus). Ganzhorn and Kappeler (1996) demonstrated that the density of primates in Kirindy is among the highest in the world. A lower proportion of primates was found in the diet of fossas in the montane forests of Montagne d'Ambre (Rasolonandrasana 1994), which has a lower density of primates than Kirindy (Ganzhorn et al. 1997). The principal prey types identified by our study were 2 or 3 spiny tenrec species and approximately 6 lemur species, along with, less specifically, snakes and small mammals. The medium-sized nocturnal lemurs (Lepilemur, Phaner, and Mirza) appeared to be the fossa's staple food, and were even the most common item in the limited sample of scats from juveniles. The findings of Dollar et al. (2006) in the western dry forests of Ankarafantsika National Park support these conclusions: they found a similarly high proportion of lemur remains in fossa scats, whereas another species of Lepilemur,L. edwardsi, was the most frequently identified prey species.

Prey availability in the study area varied with season: Cheirogaleus,Tenrec,Echinops,Setifer, and many snakes and lizards are inactive during the dry season, as are many individual Microcebus, and a significant difference was found between dry- and wet-season diets. However, no correlation was identified between seasonal activity and scat composition. Microcebus,EchinopsSetifer, and Cheirogaleus were not substantially more frequent in scats collected in the wet season than in those from the dry season, yet Tenrec was more frequent. Cheirogaleus,Echinops,Setifer, and Microcebus hibernate in tree holes or under bark, whereas Tenrec hibernates underground, indicating that arboreal, but not underground, shelters may be accessible to fossas. Although Dollar et al. (2006) did not carry out statistical analyses similar to ours, their analyses of fossa scats from the dry forests of Ankarafantsika National Park suggested, like ours, a shift away from Lepilemur predation during the wet season. However, this was apparently replaced not by a shift toward predation on Tenrec, as at Kirindy, but toward Rattus rattus. The reason for this was not clear, but the different combination and relative abundances of prey species at Ankarafantsika may be important.

Propithecus, unlike all other prey types, gives birth in the dry season, and this may be a reason for its higher relative frequency in the scats collected during the dry season. Digestion of a small juvenile Propithecus is likely to give rise to fewer scats than digestion of an adult one, and so any effect that an influx of young animals has on the composition of the diet will be underestimated in analysis of scats. The spiny tenrecs Tenrec,Echinops, and Setifer all produce large numbers of young in the wet season, yet any effect of this on diet composition in the wet season was not evident for Echinops and Setifer in our study.

From the analysis of the scats from known individual fossas, it appeared that there may have been a lower proportion of lemur prey in the diet of juveniles than in that of adults. It seems unlikely that the relatively large size of some species of lemurs was an impediment to their predation by juvenile fossas, given that no significant difference was found in the size of the prey found in scats from juveniles versus adults. Fossas reach maturity at 3–4 A years of age (Albignac 1975), and our finding suggests that the hunting of lemurs is a skilled task that takes fossas some years to learn. This may help explain why no other nonprimate mammals specialize in predation on primates. Male fossas might be expected on the basis of their larger size and canine dimensions to prey on species of larger mean size than females. This difference was not found, with proportions of large to small prey in their respective diets almost equal.

Although fossas have the capacity to specialize in predation on lemurs, they appear to be opportunistic predators, within the limits of a diet composed primarily of vertebrates. The relative frequency of a prey type in the scats, corrected to give an indication of numbers of individuals consumed, was related to the abundance of that prey, whereas no other factors were found to explain either corrected or uncorrected relative frequencies. Prey abundance accounted for 70% of the variation in corrected relative frequencies. Such a response in prey choice to prey abundance has been found in other carnivores, including neotropical rain-forest felids (Emmons 1987), and is mostly supported by other findings for fossas. As Rasolonandrasana (1994) showed, fossas in the montane humid forests of Montagne d'Ambre appeared (from the small sample collected) to prey on higher numbers of rodents than of lemurs, whereas we very clearly showed the reverse in the dry deciduous forest of Kirindy. Calculations of relative densities of rodents and lemurs in Montagne d'Ambre and Kirindy, made using comparable methods (Ganzhorn and Kappeler 1996; Ganzhorn et al. 1996, 1997; Goodman et al. 1997a), show that there is a higher ratio of lemurs to rodents in Kirindy than in Montagne d'Ambre, which is reflected in the diet of the local fossas (Rasolonandrasana 1994). This supports the findings of Goodman et al. (1997a) that fossas appeared to take more birds and rodents in the Andringitra Massif where lemurs were rare. Two studies do not entirely support the conclusion that fossas are primarily opportunistic predators. Rasoloarison et al. (1995) found evidence from scats in Kirindy for a preference for larger prey, disproportionate to relative abundance. Dollar et al. (2006) make a similar suggestion, but in neither case has the relationship been found to be clearly significant. The surface area factor, for which we corrected, may have contributed to this conclusion. Further investigation into the best way to correct for this factor as well as fur length might clarify the issue.

Prey abundance does not equate to prey availability: a highly abundant prey species may be difficult to locate or catch. It is much more informative to investigate relations between changes in scat composition and changes in prey abundance, on the assumption that changes in abundance are the prime source of fluctuation in prey availability. There are, unfortunately, insufficient data currently available on changes in densities of lemurs in Kirindy from year to year; the only available data are on Hypogeomys (S. Sommer, Department of Ecology and Conservation, University of Hamburg, pers. comm.), examination of which indicates that populations of this species remained roughly constant during the period of the fieldwork for the present study.

Emmons (1987) proposed that the only way for felids in the rain forest to acquire prey is to obtain them opportunistically through walking extensively, because the prey is widely scattered, only visible within a few meters, and unpredictably distributed. This strategy results in a diet that reflects the relative abundances of the available prey species, as was found for fossas in the present study. To support her hypothesis, she cited the high diversity of prey reported in other studies of felids as evidence of the unpredictability of encounters with prey in the rain forest. As mentioned above, however, abundance is not the only factor involved: LepilemurPhanerMirza and Microcebus appeared to be taken much more frequently than would be predicted on the basis of their abundance alone, and Cheirogaleus appeared to be taken much less frequently.

The predictability of prey distribution, as discussed by Emmons (1987), is difficult to quantify from the point of view of a fossa. The species of lemur that seemed to have the most predictable movements was Eulemur fulvus, which, in the dry season, came regularly to drink at a limited number of waterholes; yet this was the species least frequently found in the scats. None of the other measurable attributes of the prey types (body mass, biomass, daily rhythm, or seasonal activity) correlated with their frequency in the scats.

On the basis of relationships described in Peters (1983), the expected mean body mass of prey taken by fossas was 0.7 kg. Our 2 estimates, at 0.51 kg and 1.1 kg, lie on either side of this prediction, with large standard deviations. Peters (1983) generalization that mammals take prey of 10% of their own body mass was presented as approximate. We believe the former estimate (0.51 kg) more likely to be accurate, being derived from our corrected relative frequencies. The most commonly identified prey type of fossas (LepilemurPhanerMirza) consisted of species with mean body masses close to or less than the predicted weight, but body masses of prey ranged widely, from 30 g (Microcebus) or less (individual species of small mammals were not identified) to 3.6 kg (Propithecus). This variety accords with the findings of Dollar et al. (2006), who estimated median body mass of prey to range between 0.08 kg and 0.9 kg for 2 sites in the dry forests of Ankarafantsika National Park.

Our findings indicate that fossas regularly consumed prey weighing more than one-half their own body mass (6.75 kg). This type of predation is likely to be rendered even more difficult by taking place in the trees, but further evidence for such behavior is provided by the report of Wright et al. (1997) of rain-forest fossas taking Milne-Edward's sifakas (Propithecus diadema edwardsi) typically weighing 5.8 kg (Glander et al. 1992; Mayor et al. 2004). However, our study did not investigate the proportion of young animals in the diet of fossas. It is possible that fossas tend to select younger, smaller animals. The lack of correlation between body mass or daily rhythm of a prey type and its relative frequency in the scats runs counter to the findings of Rasoloarison et al. (1995): the large diurnal lemurs did not appear to be the preferred prey for fossas, although, from the uncorrected relative frequencies, it is clear that in terms of biomass Propithecus was an important food source. On the contrary, it appeared that the tiny Microcebus was taken much more frequently than would be predicted purely on the basis of its abundance.

Examination of our data indicated a higher impact of predation by fossas on P. verreauxi verreauxi than estimated by Rasoloarison et al. (1995). They estimated that fossas in Kirindy remove 12 individuals of Propithecus per year from a 7-km2 area, that is, 29% of the yearly population growth of 42 young in that area (based on density estimates of Propithecus from Beza Mahafaly—Richard et al. 1991). From Table 4, however, our estimates suggest that 35 individuals would be taken from such an area. Following the logic of Rasoloarison et al. (1995), but using data on density of Propithecus from Kirindy (Ganzhorn and Kappeler 1996), production of 53 young per year would be predicted in a 7-km2 area. Thus, fossas in Kirindy would be removing 66% of the number produced each year.

Impacts on LepilemurPhanerMirza and Microcebus look more severe, especially on the former; Lepilemur and Phaner are reported only to produce a single young per year (Charles-Dominique and Petter 1980; Petter-Rousseaux 1964), whereas Microcebus murinus typically produce twins (Petter-Rousseaux 1964). Microcebus also experiences predation by owls (Asio madagascariensis and Tyto albaRasoloarison et al. 1995). The impact of predation by fossas on this species depends on whether the density estimate by Ganzhorn and Kappeler (1996—171 individuals/km2) or that quoted by Kappeler and Rasoloarison (2003—712 individuals/km2) was correct at the time of our study.

The finding that fossas appeared to take relatively abundant prey suggests that they may switch preferences if the population density of preferred species decreases. Their apparent preference for abundant prey, with the high impact that they are estimated to exert on some of their prey populations, indicates that in Kirindy fossas may limit at least some of the populations of their prey species, and that their own numbers may be limited by the abundance of their prey.


We thank the Commission Tripartite of the Malagasy Government, the Université d'Antananarivo, and the Département des Eaux et Forêts for permission to work in Madagascar. We thank the Centre de Formation Professionelle Forestière de Morondava and the Deutsches Primatenzentrum for permission to work in Kirindy. We thank our numerous field assistants, in particular R. Zoelisoa, R. Aimé, I. Fazey, and N. Holmes, for their help with all aspects of the work in the field, and the many researchers who assisted with collection of scats. We are grateful to S. M. Goodman for advice and for his independent analysis of a sample of our fossa scats for comparative purposes. We also are grateful to S. M. Goodman and an anonymous reviewer for their comments on an earlier draft.


  • Associate Editor was Martin B. Main.

Literature Cited

View Abstract