This paper presents ecological data and analysis on bobcats (Lynx rufus floridanus) that were captured, radio-collared, and tracked using radio-telemetry equipment and triangulation techniques on or adjacent the St. Johns National Wildlife Refuge Beeline Unit in 1999 and 2000. The information presented here comes from data collected as part of the Foundation’s ecological study of free-ranging bobcats in the fragmented landscape of Brevard county, Florida. The purpose of this study is to examine various ecological and biomedical aspects of bobcat populations in fragmented landscapes in order to understand associated dynamics and to suggest habitat conservation guidelines that offset the effects of large-scale habitat fragmentation. The findings presented here are part of a larger study being done by Coryi Foundation Inc. that covers most of Brevard County.

Though the bobcat is not a listed species, there is great potential for its local loss in Brevard County (Mallow, 1999a). With Florida as a whole growing rapidly in terms of population and new development, humans are quickly encroaching upon and eliminating bobcat habitat at an alarming rate. In Brevard County, out of the 203,858 acres of land that exists within the county (see Figure 1 below), 50% is already developed (Mallow, 1999a). Urbanization rations parcels of prime habitat to humans and poor quality habitat to predators in a pattern that is severely fragmented. This patchwork array of poor-quality land may not offer sufficient functionality for the wide-ranging bobcat. Habitat isolation via fragmentation by development could create small isolated populations that could succumb to the deleterious effects of inbreeding (Franklin 1980). This report discusses the functionality of an existing large block of habitat – the St. Johns National Wildlife Refuge Beeline Unit west of Port St. John, and the way bobcats use the landscape there and adjacent to it. It was deemed a worthy effort to determine how such lands could play in the local future of the bobcat, since it is likely that such places could be all that remains of wild lands in the future, should development trends continue as they are.

For the purpose of habitat selection analysis and home range size estimation, fixes on radio-collared bobcats are located via triangulation with radio-telemetry equipment at distances of less than 500 meters between observer and animal. The difference in bearings at each location is 75 to 105 degrees... and only locations that are temporally independent are used in these analyses (Swihart and Slade 1984).

Home range sizes were estimated using 95% harmonic means (HM) (Dixon and Chapman 1980) methods. The harmonic means (HM) method is a statistical estimation that generates a home range boundary on the basis of the clustering of locations and how the boundary ‘hugs’ the spatial density of locations (Mallow, 1999b). The HM method is more representative of spatial use and excludes unused space, unlike the minimum convex polygon (MCP) method (Mohr, 1947). The MCP method is an over-estimation of range size that does not reflect the true use of habitats within the boundary defined by this method.

The need to rely on the HM method in this study is best illustrated by Figure 2 below.

The MCP method is a non-statistical, quick and simple estimation of range size that is determined by calculating the area formed by the minimum convex polygon that possesses apexes at the outermost locations. That is, one simply establishes a range boundary by drawing straight lines between the outermost locations and then determines the area enclosed by the boundary. Regardless of how locations are distributed within the boundary, the enclosed area is assumed (albeit erroneously) to be used uniformly by the animal. The MCP’s obvious shortcoming is that it does not represent a true use area because it includes appreciable unused regions within the boundary. These unused areas consist of avoided cover types. This bias can be most pronounced in landscapes where severe heterogeneity (chiefly due to habitat fragmentation via developed areas) leads to an appreciably uneven distribution of locations across cover types on which animals strongly differentiate preferences. For instance, where the landscape forces the animal to occupy a home range that is U-shaped or crescent-shaped, the arithmetic mean center will lie entirely outside of the home range, whereas the MCP center will appear inside the MCP boundary. The MCP tends to be most accurate in landscapes where the animal uses areas more uniformly. Such landscapes are largely undisturbed and/or may contain more or less evenly distributed preferred cover types.

On the other hand, as alluded to above, the harmonic means method is a statistical estimation that generates home range boundary contours that are not restricted to the convex shape of outermost locations. This method evaluates location clusters as spatial representations of activity intensity (Dixon and Chapman 1980). Range boundaries are established as isopleth contours based on the location clusters and their harmonic means in an areal distribution. These contours represent activity cores that are correlated with areas of "equal" activity. The final products are a home range size and boundary that are more representative of actual use and are a greater approximation of true activity because unused space is more effectively excluded. Indeed, in some cases, the HM home range may actually be represented by two or more activity cores that are separated by "hostile" (or largely unused) areas through which the animal quickly travels. The cores tend to represent the preferred, and therefore, most used areas. Because unused (avoided) space is more effectively excluded from the home range boundary, HM ranges are generally smaller than MCP ranges. The HM method is very useful in fragmented landscapes where land cover diversity is great and habitat preferences strongly differentiated between preferred cover types and avoided types (i.e., developed portions) of a home range.

Comparing estimates of both methods can demonstrate the severity of internal home range patchiness and fragmentation. Generally (though not rigidly true), one may deduce that the greater the difference in range size estimates between the two methods for an "urban" bobcat, the greater the level of fragmentation and patchiness due to development (or other avoided cover types) within its home range. Though this comparison is not made in this report, the reader is referred to Mallow (1998).

Habitats were indexed according to Florida Land Use Cover Classification Scheme system established in May 1993 (May 1993). The classifications used were: upland forests, wetland forests, shrub and brush, range land, agriculture, water, non-forested wetlands, barren land, transportation, communication, utilities, urban and built up, and low density residential. Land cover maps were based on 1995 aerial photographs.

The habitat selection index for a given cover type was determined as the ratio of the number of observed locations in that type to the number of expected locations in that type. The number of expected locations was calculated as the total number locations multiplied by the percentage of a habitat type that was enclosed within the home range. An index greater than one indicated that a habitat type was preferred. An index of less than one indicated that a habitat type is avoided. An index equal to one indicated that the animal does not care one way or the other about the habitat type - in other words, it randomly selects to use that type. Digitized maps were used in the selection index computation process.

Results

Reference is made to Figure 1. Seven bobcats that occupied space on and adjacent the Beeline unit were captured and radio-tracked. The sex, age, and annual home range size data for the years 1999 and 2000 for these cats are in Table 1 below. These are not all the bobcats that have been captured and collared by us, but only those on or in the vicinity of the Beeline Unit.

During the period, the harmonic mean home range sizes of the seven bobcats averaged 10.52 and 4.58 square kilometers for males and females, respectively. Averages for males and females, on and off the Beeline Unit were as follows:

Home range maps of the seven bobcats included in this report is below. Pink ranges are female and blue ranges are male.

Bobcats in the study used the following habitats in the order of descending selection:

Habitat types that were either forested (wet or dry) or dry shrub and brush were preferred (indices greater than 1.0), whereas those avoided (indices less than 1.0) consisted of non-forested wetlands (herbaceous or shrubby), open range land or residential.

Below is a table of the percent composition of preferred versus avoided habitats in the home ranges of the seven bobcats.

Table 4. Cover Composition of Home Ranges

Documented Corridors on the Refuge

A number of corridors were documented in all areas by virtue of the preference that the bobcats showed for the use of linear tracts of land that course through unfavorable habitat. The below figures show the corridors that were documented.

Figure 7. Documented corridors in the south Titusville area around the TICO airport, northwest of the Beeline Unit. These corridors are heavily used by B045.

Road Deaths and Mortality

Since 1993, Coryi Foundation, Inc. has been documenting road kills in Brevard County. The following tables are sourced from the mortality database.

Home range sizes were highly variable among the bobcats. With the exception of B043 (10.05 sq km) and B045 (26.7 sq km), home range sizes are smaller than those reported from other areas in Florida (Mallow 1999b, Maehr 1997, Wassmer et. al 1988). Wassmer et. al (1988) reported averages for male and female bobcats to be 25 and 14 square kilometers, respectively. Most of the bobcats in this area possess ranges far smaller than this.

It is speculated that the variation noted in this county is due to a highly heterogeneous nature of the landscape – one that is extremely fragmented by development. In some parts of the county, there does exist appreciably large and intact blocks of upland and forested lands in which one or more bobcats can establish normal size home ranges. In others, the landscape is so limiting in its expanse of natural and contiguously preferred habitats that bobcats have to ‘piece’ together what is left in order to meet nutritional needs via predating upon existing natural prey populations such as rabbits and rodents (Maehr and Brady 1986).

The increasing dominance of the percentage of non-forested (marsh) communities in the county that is the result of continued reduction of forested and upland communities via development is the pervading trend of the area. This is rendering a situation whereby the flood plain habitats are becoming the dominant natural landscape – possibly the last natural area in existence and one in which the bulk of wildlife may only be found at some time in the future. The Beeline Unit is a case in point that serves as an excellent model of study for bobcats in this respect. That is to say, with most upland areas on the brink of conversion to development, places like the Beeline unit, as are all flood plain lands, are becoming true places of refuge for bobcats in the county.

Forested habitats account for a small amount of the total composite of habitats in that system. As a result, bobcats in that matrix may have adapted to smaller home range sizes as a result of conspecific competitive interactions that intimately play in tandem with available resources – limited preferred denning cover (dry ground) and preferred prey, all of which are dependent on associated forest communities. The net result is patchy use – islands of preferred habitat and associated species concentrated within accordingly – not because prey densities are high, but because there is little preferred bobcat habitat in that system to start with. It is interesting to note that subjects such as B058 and B059 were underweight and ‘appeared’ less healthy than their counterparts to the north in south Titusville and to the south. (It must be mentioned that B056 and B035 were located in areas of appreciable forest.) This apparent reduction in fitness may be due to a lower preferred prey base in the flood plain compared with the upland forest communities in which B043, B045, B056, and B035 are located.

However, it is equally plausible that the large ranges of B043 and B045 are due to the fact that much of their range space is highly developed and that they have managed to piece the landscape together in such a way as to render those large ranges. In essence, the fact that the net size of available habitats that are preferred are less than 50% speaks to this account – that is, that the larger range is acting to compensate for the developed areas that exist within. But this could only be done because of the greater availability of preferred types in their landscape that are still connected. More on this will be discussed later.

Range sizes can negatively correlate with the percentage of good quality habitat as noted in Florida panthers in the Big Cypress (Maehr 1997) and bobcats in large  forested regions like the Mallory Swamp of north Florida (Mallow 1999b) or; alternatively, with the 'island' character of small patches of preferred types located within a vast sea of avoided types (e.g., marsh). But they can also positively correlate with the degree of internal fragmentation that forces the animal to defend a larger range. I believe we are seeing all these mechanisms at work in the highly heterogeneous landscape of Brevard County. 

The habitat selection indices show a definite preference and avoidance of particular habitat types. The preferred habitats across the board included wetland forest, upland forest, pine flatwoods and shrub brush (dry type). Those avoided were marsh, shrub/scrub wetland, low density residential, and range land (open grazing land). In essence, ‘forest’ is the key habitat for bobcats, and this has been verified in other studies (Mallow 1999b, Griffith and Fendley 1986).

Bobcats on the refuge and outside of it show a strong preference for any forested community and avoid the non-forested wetlands, range lands, and low density residential areas. However, the northern end of the Canaveral Groves subdivision is occasionally used by B056 – but only because it possesses forested strands that course between individual properties, many of which are appreciably non-cleared, and all of which are 2.5 acres or greater per dwelling. [These individual 'backyard' corridors are small in scale, but when connected, can provide an expansive network of travel corridors that can get an animal as a bobcat deep into a residential area. They can also provide a series of small and connected links that can afford travel under cover of forest from one side of a subdivision to the other. Figure 6 illustrates the various corridors used by B056 in that part of Canaveral Groves.]

There was no indication in the data that bobcats on the refuge had adapted to a more preferred use of marsh just because they are forced to live in a predominant marsh landscape. In fact, B058 and B059 exhibited an average individual selection index of preference for forest and shrub and brush of 1.52 and an avoidance of marsh of 0.71. Thus, bobcats still exhibit the same avoidance to marsh as do the other subjects outside the refuge. The apparent effect is that since they continue with normal selection of habitat use, their ranges are severely reduced in size due to the limitation of preferred habitat types. In other words, marsh or open range may comprise an appreciable portion of their MCP ranges, but it is used little (relative to its availability), thereby rendering HM range sizes (and therefore, true range sizes) very small. And, in light of the island-like character of forest that occupies the flood plain, the HM ranges are extremely small compared to others outside the flood plain.

Table 4 indicates that, overall, there is extreme fragmentation in ranges. This occurrence can be best illustrated by looking at the range habitat maps of B043 below in Figure 8 and of B058 (in Figure 4).

The areas of the home range map in Figure 8 that are white are developed and avoided habitat types. The dark shaded areas are of preferred habitat types. The bobcat that maintains this range occupies a space that is severely fragmented. As mentioned above, the size of the range in this case may be positively correlated with the percentage of developed areas within (the more avoided types in the range, the larger the range). This map depicts a MCP range boundary. It is apparent that the HM boundary would tightly hug the dark shaded areas; for such is where radio-locations were always found. Thus, in order to defend a suitable range, this bobcat links up the landscape amidst the developed areas. The net effect is a larger MCP range than a HM range.

To contrast, in the map of Figure 4, the locations of B058 (which are almost entirely found within the black polygons - the preferred habitats (pine flatwoods, upland hardwood forest, etc.)) define a very small range. This is a map of the Beeline Unit. The dark strand is the strand of forest that virtually runs from west to east through the huge marsh and open range landscape that comprises the majority of the refuge. Making reference to the harmonic means home range boundary in the map of Figure 3, one can see how the true (HM) range has a kidney shape to it. This is because the contiguous expanse of preferred habitats of B058’s range is shaped like just like this. This block of preferred forest is located within the vast and largely undesirable wet marsh, shrub wetlands, and open range land that define this region of the St. Johns flood plain.

Now, comparing range size of B043 and B058, one must address the question as to why they radically differ in size, though both are seemingly as affected by avoided habitat types (development in one case and marsh in the other). In the case of B058, her territory makes use of a landscape dominated by avoided types. In the case of B043, her territory makes use of a landscape less dominated with avoided types. I.e., B058’s range occupies a small island of preferred habitat (forest) in a vast sea of avoided type (marsh). The opposite occurs for B043 – she resides in an area (south Titusville) that still retains a great amount of preferred and undeveloped habitats, albeit with severe internal fragmented. Her habitat island is larger, and better connected across the landscape, through the maze of avoided land cover (that avoided type being mostly development land cover types).

Indeed, some of the home ranges in the study are severely fragmented internally as a result of development, as is the landscape at large. Others are akin to isolated islands – where there is little internal fragmentation by avoided types (B058’s range), but nonetheless small because they reside within landscape dominated by avoided cover (e.g., the vast marsh of the flood plain).

Bobcats are restricted in their use of the landscape to the undeveloped or preferred portions of their ranges, chiefly those that are forested or brushy as indicated in Table 3 (habitat selection indices). Locations rarely fall within developed cover types. That the percentage of home range areas that possessed developed cover types ranged from 30.6-45% is indicative of appreciable internal patchiness or the extrinsic domination of preferred patches by vast seas of avoided expanse. The intra-range heterogeneity would result in an appreciable percent difference in home range size estimates between the MCP and HM methods.

The amount of internal fragmentation would appear to partly dictate home range size. If a range is severely fragmented within (by natural and/or developed types of land cover), then it would likely make a bobcat maintain a larger than normal home range in order to obtain sufficient prey year round to sustain it. This is what is speculated to be occurring with the large ranges of B043 and B045. On the other hand, a home range size could be abnormally reduced if it was located in an extremely large and unfavorable landscape that was meagerly filled with preferred types of habitats that function as small islands of refuge. This may the case for B058 and B059 on the Beeline Unit. Such is the theory put forth from the data and analysis.

Despite all these impediments to range establishment and maintenance, bobcats do appear highly adaptable to connecting the landscape via corridors. Some corridors are ‘intra-range’ as depicted in the upper third of the range map of B043 (Figure 8), while other corridors actually define a nearly entire territory, as in the case of B058 (Figure 4).

Other corridors documented included those of B056, which used a shrub/brush and forested corridor that courses along the western edge of Port St. John. This functional corridor actually demonstrated a useful link of passage within relatively high amounts of preferred cover between the north and south sides of Port St.John. That corridor resides on the Beeline Unit. Rather than setting out across the marsh and open range land that comprises the unit at large, B056 repeatedly used this narrow corridor (depicted in Figure 5) as a means to circumvent Port St. John. It must be noted that there are gaps in that corridor – small gaps of open areas that an animal can securely traverse under cover of darkness. 

We propose that this corridor link be enhanced to allow passage for species between the north and south. The bulk of habitats on the unit (to the west) are too ‘hostile’ to promote passage. B058 may be able to use the east-west strand, but this strand sort of leads to nowhere – it terminates in the deeper marsh areas of the river, and then, eventually the river itself. I.e., it is not a truly connective corridor; for it does not link two disjunct large blocks of preferred habitat. On the other hand, the corridor used by B056 does; for it does connect the forested south Titusville area with the largely forested area that is adjacent Canaveral Groves. Such a link may be critical to gene flow between north and south Brevard.

During the period of this study, there was documented the road death of an adult male on SR 407 within a ˝ mile of the northern boundary of the Beeline Unit. We suspect that he was the dominant male of the unit, and that his death had prevented us from capturing and collaring an adult male while we were trapping. That is to say, when he had been killed at the beginning of the study, we felt we had lost the opportunity to collar a male on the refuge; for despite long term capture efforts, we only captured the two females in the unit.

In light of the statistics of mortality and road kill causes as shown in Tables 6 and 7, we feel that male turnover is rapid in the landscape at large throughout the county. Coupled with the vast amount of poor quality habitat that acts as impediments to long range travel, we also feel that migration and re-occupation rates of ranges may be relatively small, leading to the existence of vacant ranges for extended periods of time. Certainly, young males (those that travel the farthest to find and establish a territorial range) are experiencing the brunt of road deaths. This high rate of death contributes to the impediment factor that results in slow re-occupation. It also limits gene flow in the county because young males are the sex/age class that contribute the highest gene flow rates (Griffith and Fendley 1986); for they travel the farthest – if they die, so too do those genes that would otherwise perfuse the landscape.

Our findings suggest that adaptation to hostile and fragmented landscapes (as in the cases of B058 and B059, in which their home ranges are in a large and vast expanse of avoided habitat that is dotted with islands of preferred habitats) is accomplished by simply maintaining a smaller range that makes use of the available islands of connected preferred habitats. Because these ranges are ultra small, it is speculated that fitness may be slightly reduced in such a condition as a result of reduced nutritional intake from limited prey resources. However, the reduced energy demand placed on animals moving about in very small ranges may be low enough to help balance the energy equation - that is, the animal would need to find less food by a proportionate amount.. 

It is also suggested that adaptation is accomplished in other cases by connecting up the preferred habitats in the fragmented landscape by intra-range corridors (as in the cases of B043 and B045) so that an effective normal size range can be achieved. In this latter case, prey availability is higher, and so too may be fitness. But the added energy cost of patrolling the larger range requires the additional intake of the available prey that would be found in a larger range. Thus, the energy equation would become balanced.

In both cases, one fact is clear – forested habitat is the key to the needs of this species. Even when pushed into extremely hostile landscapes (hostile defined here as avoided types – e.g., marsh or open range land) such as dominantly found on the Beeline unit in which B058 and B059 reside, bobcats do not shift their selection of habitats to these avoided types, though they may skirt their edges or penetrate them shallowly. They still prefer and will largely confine their movements within what small amounts of forest patches and strands exist. 

Be that as it may, one has to wonder what the lower limit is in terms of the minimum size range that could be maintained in such a landscape. 

Unless there are amply connected patches of forest within that affords them adequate dry denning cover and preferred prey densities their long term life-history and energetic needs may be short-changed. For they simply do not appear to adapt to the use of non-forested habitats such as marshes, though (as in the case of B058 and B059) it is highly available to them. Both animals made almost exclusive use of forest types that acted as islands or strands in the otherwise marsh or bare landscape.

Given the marsh character of the flood plain region at large, it is speculated that bobcat density is quite low all the way up and down the floodplain, and may exist only sporadically across the landscape in places where there is enough dry and forested habitat to afford suitable den sites - islands of preferred habitats close enough for traverses across hostile types and strands that act as additional linkages. Lack of forest cover and inundation by water for a good percentage of the year both impede upon the use of the flood plain by forest creatures as bobcats. But these are tentative statements based on the findings (as good as the evidence is at hand) of a small sample of the population and thus, requires more research.

We would recommend a greater sampling over a much greater expanse of the flood plain before we could be confident that what our findings suggest is absolutely true. We are planning to that end to expand our research within the flood plain.