Natural Barriers and Internal Sources for the Reproductive Isolation in Sympatric Salmonids from the Lake–River System

Abstract

Within the sympatric evolution framework a range of ecological variables are considered as potential initiators and controllers of the diversification process. Here, we identify the proximate factors providing the reproductive isolation among seven sympatric ecomorphs of the genus Salvelinus charr dwelling in the Lake Kronotskoe basin (North-East Asia). We demonstrate that the slope profile of the lake tributaries determining the water flow velocity and position of the groundwater discharges serves as a barrier between the reproductive sites of the ecomorphs and provides the basis for selective pressure affecting the Lake Kronotskoe fish during spawning. The main characteristic under selection is a migratory ability, which is determined by the swimming performance and the amount of energy reserved in the body and depends on fish morphology and physiology. The flex points indicating abrupt slope changes along the spawning watercourse restrict the upstream migration of the groups having a comparatively low swimming performance and energy reserve. A thorough analysis of the ecomorphs’ migratory and spawning activity indicated two energy expenditure strategies: the fish could invest most of the energy either on migration or spawning. Our findings supported with data on fish ecology, morphology, and thyroid hormone status allow us to put forward a following hypothesis. We suggest that the interplay of spatially heterogeneous environmental variables affecting life history decisions via ecomorphological and physiological traits could serve as a trigger for the reproductive isolation among the ecomorphs in a single ecosystem.

Appendix

Satellite Data Processing

The above sea level (a.s.l.) marks were obtained from the global digital elevation model (Global DEM). These data were granted by TanDEM-X Science Service System (12 m (0.4 arcsec) tiles N54E159, N54E160 and N55E160 hydrology DEM) and processed in ArcMap v.10.1 (ESRI). The relative elevation marks were measured with an accuracy of ± 0.5 m, and ten measurements were processed for each point at each locality and averaged. Such marks were acquired for the (i) upper, middle, and lower course of spawning site; (ii) for contiguous 1 km sections starting from the downstream margin of the spawning site to the river mouth; and (iii) for 100–200 m sections from the confluence to 1 km upstream for each tributary. Elevation marks were verified with the elevation benchmarks from the 1:100.000 topographic maps. Finally, these data were combined to estimate the slopes along morph’s spawning migration path. Stream velocity was calculated for each section (see below) and verified with the field measurements. River section length and width were obtained from precise satellite images in SASPlanet (GoogleEarth, BingMaps and HereMaps data combined). The width was measured every 100 m for each section and averaged for the 1 km sections.

Calculation of the Riverbed Slope and Water Flow Velocity

The slope of each section i (I_i) was calculated as:

$$I_{i} = \frac{{\Delta H_{i} }}{{L_{i} }}$$

where L_i is the length of the section varied between 100 m in proximity of the branching points and 1 km in other parts, and ΔH_i – the difference in the absolute altitude marks. The average stream velocity in each section (V_i) was calculated as:

$$V_{i} = k_{Chezy} \cdot \sqrt {I_{i} \cdot h_{i} }$$

where k_Chezy is the Chezy-coefficient relating the slope (I) and some effective average depth (h) with the average water flow velocity at the given i-th river section (Baryshnikov, 2003). The Chezy-coefficient was calculated as follows:

$${k_{Chezy} = \frac{{h_{{av_{{i^{16} }} }} }}{{k_{{roug_{i} }} }}}$$

where k_roug is the roughness coefficient reflecting the friction of flowing water on the river bed composed of the different-sized floor grounds. The values of the roughness coefficient are empirically measured and tabulated (Baryshnikov, 2006; Sribnog, 1932).

The average depth of each section (h_i) is tough to measure experimentally, not only due to the large amount of work but also due to the “effective” nature of this parameter describing the mean properties of a rather long river section. Therefore, we have assessed the average depth of each i-th river section as follows (Baryshnikov, 2003):

$${h_{i} = \left( {\frac{{Q_{i} \cdot k_{roug} }}{{B_{i} \cdot \sqrt {I_{i} } }}} \right)^{{\frac{{3}}{{5}}}} }$$

where Q_i is the average water discharge, and B_i – the average river width. The latter was directly measured using the satellite images, while the former was assessed, assuming the discharges to be approximately constant at the sections lacking the tributaries and sum up at the confluence of two watercourses. The discharges were estimated at the control sections near each confluent point:

$$Q_{i} = V_{i} \cdot h_{i} \cdot B_{i}$$

basing on the directly measured averaged parameters: actual average river depth (h_i), actual mean river width (B_i), and average velocity of the water flow (V_i).

The values of the slopes and integral velocity obtained for 16 spawning sites were statistically tested by H-test and post-hoc Van der Waerden test for multiple comparisons of small-numbered samples in PMCMR module of R (Pohlert, 2014).

Energy Cost of Spawning Migration and Reproduction

The Stokes friction force for the “spherical fish” traveling through the water thick could be written as:

$$F_{fric} = 6\Pi {\upeta }RV = k_{fric} V$$

where F_fric is the friction force, η—the dynamic viscosity of the water, R – the sphere radius, and V—stream velocity. We have assumed the “radius” of the fish to be one half of the maximal fish section diameter. The total work of fish against the friction force could be divided into two terms: the work during active migration (E_burst) and the work during resting periods (E_coast). Consequently, E_burst and E_coast could be calculated separately. Denoting an average length of a single burst with respect to the ground as L_b and the burst speed as V_b, then the real length (L_w,i) of the i-th burst in respect to the water would be:

$${L_{{\text{w,i}}} = \frac{{V_{b} }}{{V_{b} - V_{{\text{w,i}}} }}L_{b} }$$

where V_w,i is the velocity of the water flow. Thus, the total work (A_burst,i) against the friction forces during the i-th burst can be derived as:

$$A_{burst,i} = F_{fric} L_{w,i} = k_{fric} V_{b} L_{w,i} = k_{fric} \frac{{V_{b}^{2} }}{{V_{b} - V_{w,i} }}L_{b}$$

Thus, the total energy spent for active upstream migration would be:

$${E_{burst} = \sum\limits_{{i = {1}}}^{N} {A_{burst,i} } = k_{fric} V_{b}^{{2}} L_{b} \sum\limits_{{i = {1}}}^{N} {\frac{{1}}{{V_{b} - V_{w,i} }}} }$$

For salmonids, the average burst speed (V_b) equals several meters per second (Brett, 1965; Metsker, 1970). This parameter depends significantly on the fish size and its overall activity; thereby it seems appropriate to measure it separately for each morph. Moreover, the burst speed also depends on the fish’s physiological condition and changes during the spawning migration. Thus, measurement of this parameter is a very complicated experimental task. To overcome this problem, we calculated A_burst for the unbelievable low V_burst of 1 m/s and unbelievable high V_burst of 10 m/s. We used the obtained values as lower- and upper-bound estimates of A_burst. However, getting ahead, we can state that absolute values of A_burst appear too small relatively to other terms, and assumed inaccuracies did not affect the overall conclusions.

The work against friction forces during coasting periods (A_coast,i) can be calculated in a similar way. Let us assume that all resting periods have approximately the same duration (t_coast). As we assume that the fish does not migrate during resting periods, the distance with respect to the water (L_coast,i) during the i-th rest period would be just:

$${L_{coast,i} } = V_{w,i} t_{coast}$$

as the fish do not move downstream on average. The corresponding work against the friction forces (A_coast,i) would be:

$${A_{coast,i} = k_{fric} V_{w,i} L_{coast,i} = k_{fric} V_{w,i}^{{2}} t_{coast} }$$

The total sum of all A_coast,i leads to the lower-bound estimation of the corresponding mean energy expenditure:

$${E_{coast} = \sum\limits_{{i = {1}}}^{N} {A_{coast,i} } = k_{fric} t_{coast} \sum\limits_{{i = {1}}}^{N} {V_{w,i}^{{2}} } }$$

where N is the number of burst and rest periods.

The product of the average length of the burst (L_b) and the number of such bursts (N) equals the river length (L_R). Consequently, the increase of L_b correspondingly decreases N. Exactly the same, the product of the average time of the coast (t_coast) and the number (N) of the coast periods (and also bursts) is the spawning migration time (t_m), which is also known from the natural observations. Thus, t_coast increase correspondingly diminishes N and causes the same increase of the burst length (L_b), as the number of bursts and coast periods (N) equals each other:

$${N = \frac{{L_{R} }}{{L_{b} }} = \frac{{t_{m} }}{{t_{coast} }} \Rightarrow t_{coast} = t_{m} \frac{{L_{b} }}{{L_{R} }},L_{b} = L_{R} \frac{{t_{coast} }}{{t_{R} }}}$$

The exact values of t_coast and L_b are tough to define experimentally. Depending on the riverbed structure and the pattern of the velocity field distribution in it, these parameters can vary in a wide range. However, in our terminally simplified approach, the estimate on the E_coast and E_burst would depend on the values of t_coast and L_b, if only L_b would be comparable with the spatial resolution of the data on the water flow speed along the river. Indeed, if L_b is smaller than the discretization scale, the twofold (for example) decrease of L_b would cause a twofold increase in the number of the same terms multiplied on the twofold smaller value in the expression for E_burst. So, the result would not change. In our case, the best spatial resolution of the slope estimating was 100–200 m in tributaries' proximity, which is far longer than could be even proposed for a single burst. Therefore, the accuracy of the slope and water flow velocity estimation restricts the accuracy of the proposed approach and makes determining the exact values of t_coast and L_b senseless. So, we have assumed L_b to be equal to 10 m and calculated the corresponding t_coast for each migration route and time.

Besides the migration's energy expenditure, fish have to resist the water flow during the spawning itself, also requiring some energy to be spent (E_spawn). The expression for E_spawn could be derived similarly to the E_coast, but the time period (t_coast), in this case would be equal to the total spawning duration (t_spawn):

$$E_{spawn} = k_{fric} t_{spawn} V_{w}^{2}$$

Finally, the total spawning energy cost (E_total) could be assessed as:

$${E_{total} = E_{burst} + E_{coast} + E_{spawn} }$$

The total estimated spawning energy costs for different morphs (totally 16 calculations) were statistically compared using Van der Waerden test in PMCMR module of R.