The Incomati River drains parts of Mpumalanga, Swaziland and Mozambique between the Limpopo River system in the north and the Phongolo River system in the south. It is economically one of the most important river basins in South Africa, and it consists of three adjacent sub-basins: the Komati, Crocodile and Sabie (Darwall et al. 2009). The main river descends from the highland plateau in Mpumalanga and Swaziland and flows through the coastal plains of Mozambique to the Indian Ocean just north of Maputo at Villa Laisa. The total basin area is about 46 800 km² of which 63% is in South Africa, 5% in Swaziland and 32% in Mozambique. The average discharge of the Incomati River basin at the estuary is about 100 m³/s to 200 m³/s, corresponding to about 3600 million m³ per year, to which South Africa contributes 82%, Swaziland about 13% and Mozambique about 4% (Darwall et al. 2009).

The study area includes two rivers, namely the Crocodile River and the Komati River, which join to form the Incomati River below the border town of Komatipoort. The Crocodile River flows along the boundary of the KNP, and at the confluence, the border extends across the river to also include the lower reach of the Komati River (Figures 1 and 2). Below the confluence, the Incomati River can be described as a meandering river, incised into a wide sandy river bed, and in some sections, it flows through multiple bedrock channels. The river varies between 40 m and 50 m wide, with mostly large sandy pools and occasional rapids and a few riffles (Roux et al. 1990). Collection and tagging were done upstream and downstream of the confluence between KNP and Tenbosch weirs and the low-water bridge in the Komati River. The choice of the collection and tagging area was motivated by the relative abundance of tigerfish in this river reach. The ability of tigerfish to overcome obstructions and their various home ranges later defined the extent of the study area. Historically, tigerfish distribution data would indicate that tigerfish occur up to an altitude of 300 m in the Incomati River system. Gaigher (1967) previously collected tigerfish in the Crocodile River gauge close to the town of Nelspruit and in the Komati River close to the town of Tonga on the border between South Africa and Swaziland. Consequently, the experimental design made provision for long-distance tracking in relation to historical distribution in the Incomati River system.

Collection and handling of fish were performed in such a manner as to minimise physical and physiological stress to the specimens (Spedicato, Lembo & Marmulla 2005). Tigerfish were caught using two techniques: rod and reel with artificial lures and fly-fishing, both using barbless hooks to reduce injury to fish and to facilitate quick release, thereby reducing lactic acid stress and ensuring survival after handling and release (Gerber et al. 2017).

Flow levels in the Incomati River system were determined from daily readings at the KNP gauging weir in the Incomati River. Water temperature, pH and conductivity were recorded daily in the Crocodile River, Komati River and below the confluence of the two rivers (in the Incomati River) using Eutech portable microprocessor-based water quality instruments. Meteorological data were gathered from a nearby weather station (Transvaal Sugar Board, Komatipoort), including rainfall, minimum and maximum air temperatures.

Two fish that moved out of the study area into Mozambique shortly after tagging were excluded from the analysis. In addition, a third fish showed no movement for an extended period after tagging and was presumed dead and excluded from the analysis. Descriptive statistics for the entire study period (summer and winter) were based on more than 10 fixes per fish for 38 fish. GPS coordinates of the radio-tracked tigerfish were used to calculate longest distances travelled and to determine home range sizes.

Bi-variate Gaussian or normal distribution kernel methods (Seaman & Powell 1996; Silwerman 1986; Worton 1989) were used to plot home ranges. This group of methods is part of a more general group of parametric kernel methods that employ distributions other than the normal distributions as the kernel elements which are associated with each point in the set of location data. Because of the meandering nature and relatively small width and limited available habitat within the Incomati River system during low flow periods at specific sites, an adaptation of the simplified minimum convex polygon (MCP) (Baker 2002; Creel & Creel 2002; Meulman & Klomp 1999) was used. Boundaries of home ranges were drawn using different sets of location data (Planet GIS). This method of using the shoreline as a boundary of the home range is a widely accepted and commonly used method in fish telemetry experiments (Hocutt et al. 1994).

For ease of statistical analysis, a binning algorithm was implemented in which the longest distance travelled, home range size and the radio-tagged fish were grouped in classes according to their magnitude. For longest distance travelled (Økland et al. 2005), fish were organised in classes ranging from 100 m to 500 m, 501 m to 1000 m, 1001 m to 1500 m, 1501 m to 2000 m and > 2000 m travelled. The home range size were classed in groups ranging from 0 m² to 10 000 m², 10 001 m² to 20 000 m², 20 001 m² to 50 000 m², 50 001 m² to 100 000 m² and > 100 000 m².

The IBM SPSS Statistics 18 program was used for basic and inferential statistics which include frequencies, normality, correlations and comparisons (SPSS 2009).

The project proposal was approved with Ethical Clearance by the Faculty of Science, University of Johannesburg and Mpumalanga Parks and Tourism (Permit number MPB 8553.).

Mean water temperature results in the Incomati River system indicate that the minimum is reached in July (18.02 °C) after which temperatures gradually increase to a mean temperature of 24 °C during October. The highest mean monthly river water temperature during this study (30.61 °C) was recorded in the Crocodile River during January (Figure 3). The highest mean monthly river water temperature in the Komati River (30.17 °C) was recorded during February. Temperatures in the Incomati River, below the confluence, were influenced by both tributaries, and consequently, the highest mean monthly temperature for the Incomati River (28.88 °C) was recorded during February.

For the tigerfish active summer period, November to April, the mean monthly pH values varied between 8.1 and 8.5, whereas the conductivity fluctuated between 274 µS/cm and 622 µS/cm in the Incomati River. Summer conductivity values were lower than winter values, but summer pH values were higher. During summer, the turbidity levels increased as a result of the higher summer flows. Although not measured, turbidity was observed to be closely associated with rainfall events in the catchment during the summer period. The highest rainfall recorded was during the months of November (115.4 mm) and February (191.7 mm).

The mean monthly flow (Figure 4) for the winter period (May–October) in the Incomati River, when tigerfish are less active, varied between 0.44 m³/s and 1.89 m³/s compared to 0.82 m³/s and 13.12 m³/s for the summer period (November–April), when tigerfish are active. The highest flow spikes were recorded during the spawning season (October–February) in the summer period (Steyn 1993; Steyn & Van Vuren 1991; Steyn et al. 1996). On three occasions, flow spikes in excess of 25.00 m³/s, with the largest of 51.76 m³/s, occurred in January (Figure 4).

In total, 41 fish were radio-tagged with a mean length (SL) of 62.7 cm (range 47 cm – 76 cm) and a mean weight of 2418 g (range 910 g – 4990 g) (Table 1; Figure 2). Of the 41 radio-tagged fish, 11 (26.8%) were males and 30 (73.2%) were females in a 1:3 sex ratio. For the radio-tagged males, the length (SL) varied between 47 cm and 60 cm (mean = 55.4 cm) and the weight varied between 910 g and 2040 g (mean = 1605.5 g). For the radio-tagged females, the length (SL) ranged from 57 cm to 76 cm (mean = 65.4 cm) and the weight ranged from 1810 g to 4990 g (mean = 2717 g) (Table 2).

The total distance of the river where adult fish were captured and equipped with radio tags measured 5.2 km. After capture, tagging and the associated disturbance to a fish when released, the fish normally moved upstream or downstream and normally only returned 2 to 5 days later to the original tagging site, thereby suggesting site fidelity. The distance moved directly after tagging varied over the 2- to 5-day period from 48 m to 1038 m. In total, 35 (85.4%) of the fish tagged returned to the original tagging site within the mentioned period, but 6 (14.6%) never returned, 3 of which moved downstream into Mozambique and were not recorded again. This showed angling in the form of catch and release may be a major disturbance, but this also confirmed site fidelity of tigerfish to a specific home range. The GPS coordinates of each sample or release site, tag number, type of tag and size, weight and sex of each fish are presented in Table 1. Over time, a movement pattern emerged for each of the 41 radio-tagged fish, and the longest distances travelled and home ranges could be determined (Table 1).

On average, fish were tracked 72.5 times (Table 2) and the total number of fixes was 2971 for the period 04 January 2003 to 22 December 2003. Some individuals were tracked up to 161 times. The maximum total of fixes (n = 161) per individual was associated with a tag life of 10 months. For the summer period (January–April, November and December 2003), there were 1322 fixes, and for the winter period (May–October 2003), there were 1649 fixes. For the summer period (or part thereof), there were 40 active radio-tagged fish, but only 32 active radio-tagged fish for the winter period (or part thereof). The mean number of fixes for females was 78.4 (n = 30) per fish with a range of 6–161. The mean number of fixes for males was 56.2 (n = 11) per fish with a range of 7–110. The reason for the lower amount of fixes for males (56.2 fixes) in comparison with females (78.4 fixes) can be ascribed to the differences in radio tag types used. As males are generally smaller than females, smaller F2040 radio tags, with a much shorter lifespan (94 days), were used to stay within the 2% rule.

For the statistical analysis, data were obtained from 38 radio-tagged tigerfish with more than 10 fixes. The mean longest distance travelled (n = 38) was 729.66 m (Table 2) with a range from 74.5 m to 3200 m. When analysing the longest distance travelled by the different radio-tagged fish, 47.4% (18 out of the 38 fish) travelled between 100 m and 500 m, 34.2% (13 fish) between 501 m and 1000 m, 10.5% (4 fish) between 1001 m and 1500 m, 5.3% (2 fish) between 1501 m and 2000 m and 2.6% (1 fish) travelled more than 2000 m (Table 2; Figure 5).

When distinguishing between the different sexes and longest distance travelled, 46.4% of females (13 out of 28 fish) travelled between 100 m and 500 m, 39.3% (11 fish) between 501 m and 1000 m, 7.1% (2 fish) between 1001 m and 1500 m, 3.6% (1 fish) between 1501 m and 2000 m and 3.6% (1 fish) travelled more than 2000 m. For the males, 50% (5 out of 10 fish) travelled between 100 m and 500 m, 20% (2 fish) between 501 m and 1000 m, 20% (2 fish) between 1001 m and 1500 m and 10% (1 fish) between 1501 m and 2000 m (Table 2). The furthest movement recorded was 3200 m over a 3-day period. This female moved out of its known home range (18 fixes) and established a new home range approximately 3018 m upstream (Table 2).

For females, the mean longest distance travelled was 734.4 m (n = 28) with a range of 74.5 m to 3200 m, and for males, the mean longest distance travelled was 716.3 m (n = 10) with a range of 148.9 m to 1800 m. No significant differences were found between males and females for longest distances travelled (Mann–Whitney U test, mean ranking males 19.8 and females 18.6, p = 0.753).

Three different tigerfish movement patterns were recorded (Figure 6). Movement patterns were obtained from a combined effect of distance travelled and home range sizes (Figure 7). Although all the fish displayed some degree of site fidelity within a specific activity zone, movement pattern 1 represents fish that moved 100 m to 500 m within a well-defined home range, and movement occurred only within this specific home range. Movements of fish number 8 serves as example for this type of movement pattern (Figure 8). The majority (47.37%) of the radio-tagged fish displayed characteristics of movement pattern 1 (Figure 6, Cluster A). Movement pattern 2 represents fish that displayed site fidelity for two or more areas within a larger well-defined home range, spanning a distance of 501 m to 1500 m. Movements of fish number 15 serve as example for this type of movement pattern (Figure 9). This group was represented by 44.7% of radio-tagged fish (Figure 6, Cluster B). Movement pattern 3 represents fish that showed little site fidelity and would temporarily occupy small areas within a large undefined home range that spans more than 1500 m. Movements of fish number 23 serve as example for this type of movement (Figure 10). Fish within the latter group can be seen as vagrants without established home ranges for a specific period. Most of these fish were also later lost as they moved out of the study area and could not be relocated. Fish in this group were large females of weight ranging between 2720 g and 3580 g and represented 7.89% of the radio-tagged fish (Figure 6, Cluster C).

For a detailed account of the movement patterns and demarcation of the home ranges of each of the 41 radio-tagged fish, see Roux (2013). The dots indicate individual fixes during tracking and the contours around the fixes indicate the defined home range.

The home range size varied between individual fish with 38.2% (13 fish) localising within an area between 0 m² and 10 000 m² (mean = 5567.95 m²) and 18.42% (7 fish) localising within an area between 10 001 m² and 20 000 m² (mean = 14 435.53 m²). Furthermore, 18.42% (7 fish) occupied a home range area between 20 001 m² and 50 000 m² (mean = 31 092.2 m²), whereas 10.5% (4 fish) occupied an area between 50 001 m² and 100 000 m² (mean = 79 809.55 m²) and 18.42% (7 fish) utilised an area > 100 000 m² (mean = 163 692.90 m²) (Table 2; Figure 7).

On average, the fish (n = 38) stayed within a defined home range of 48 846.36 m². The home range size for males and females compared favourably with a mean of 53 296.52 m² (n = 28) and a range from 331.6 m² to 236 496 m² for females and a mean of 36 385.9 m² (n = 10) and a range from 1338.8 m² to 135 982.6 m² for males. No statistically significant differences were found between the sexes for their home range size (Mann–Whitney U test, mean ranking females = 20.71 and males = 16.10, p = 0.260).

None of the 41 tagged fish crossed the Tenbosch weir. Three individuals, namely numbers 7, 12 and 18, moved upstream in the Crocodile River to be briefly recorded in the vicinity of this weir. The Tenbosch weir has a crest height of 2 m and a fish ladder constructed at the side of the weir. This ladder is of the vertical slot type and appears to be non-functional to fish migration in general.

Only two radio-tagged fish (fish numbers 15 and 39) ventured into the lower Komati River, close to the confluence with the Crocodile River, where they were confined in a pool below the low-water bridge for a few days. They were not able to overcome this obstacle. This low-water bridge at Komatipoort was constructed on a natural dolerite intrusion that stretches across the river.

Contrary to the above, a total number of 16 crossings, both upstream and downstream, were recorded at the KNP weir. This gauging weir has a crest height of approximately 1.2 m with a well-designed fish way to facilitate fish movement at medium to high flow conditions. Tagged fish with allocated numbers 1, 4, 6, 20, 27 and 37 crossed the KNP weir downstream and upstream over the period January to March, whereas fish 19 crossed the KNP weir downstream during July and returned upstream three days later. Fish numbers 1 and 27 each crossed on three occasions, whereas fish number 20 crossed the KNP weir on four occasions with only a few day intervals between upstream and downstream crossings. Numerous visual observations were made of untagged tigerfish jumping over this weir over the duration of this study. Successful crossing at the KNP weir occurred at flow velocities between 1.78 m³/s and 16.2 m³/s (Table 3).