THE VALTELLINA CASE STUDY: DEVELOPMENT OF HABITAT SUITABILITY CURVES AND THE PHABSIM EXPERIENCE ON THE ADDA RIVER

Azzellino A., Vismara R.

Politecnico di Milano, Italy

1. Introduction

Among the experimental methodologies that secure a quantitative appraisal of minimum flow by using biologically transformed variables PHABSIM, (Bovee, 1986; Milhous et al., 1984) is currently the most widespread and advocated one. For his basic application, PHABSIM requires Habitat Suitability Curves (HSC), for selected aquatic species; HSC are normally expressed as univariate functions with respect to physical parameters characterising the habitat, such as current velocity, water depth and substrate size values. A huge number of standard univariate HSC are available from literature and from these curves, generic HSC have been developed and broadly applied. However, experimental evidence suggests that such generic curves can be less accurate than site-specific HSC in predicting the most favourable conditions (Heggens and Saltveit, 1990; Greenberg et al., 1996.). Prior to this investigation, no Italian site-specific or regional HSC were available and the application of PHABSIM was limited to the use of literature HSC which can bias the final results. To get around this criticism primary objective of this study was to determine the HSC for Brown trout (Salmo trutta fario L.) in a Northern Italian river with respect to current velocity, water depth, substrate class size and cover.

Even if the univariate approach is still the most widely used several authors (Orth and Maughan, 1982; Morhardt 1986; Lambert and Hanson, 1989) suggested that considering independently derived univariate suitability curves could lead to a misleading interpretation of PHABSIM results. Mainly it has been argued that treating each hydraulic variable independently may be questionable and could induce a bias due to an overlooking of possible interactions between the variables. We therefore included in the study a comparison between multivariate and univariate suitability models.

2. Study area

Data were collected in four sites selected within a reach of the river Adda located near the city of Morbegno. The river reach is 32 km long and 25 m wide, with an elevation range of 274 -198 m a.s.l., a mean gradient of 0.2% and a mean annual flowof 7-8 m3 s-1. The studied area is mainly characterised by runs spaced out by short riffles with higher gradient; substrate results composed of cobbles, boulders and gravels.

3. Materials and Methods

Brown trout is common if not dominant in the Adda river and for this reason it was selected as target species. Two life stages recognisable from body length were studied: juvenile (1+) 12-20/22cm and adult >20/22 cm.

3.1 Data collection

Used and available habitats were measured during spring and summer 1996. Brown trout habitat utilisation was found measuring current velocity, water depth, substrate and cover data at each fish location in 4 sampling areas (amounting to 10,000 m2) studied during April and May for a total of 528 observations (213 adults and 315 juveniles). To locate fish position we used backpack electrofishing, trying to minimise the disturbance caused by the sampling method.

Habitat availability was quantified with measurements at 1 meter intervals spaced along four to six transects placed perpendicular to the flow direction in each sampling area. Water column depth was measured using a graduated pole; mean current velocity was measured with a digital current meter at 0.6 of the total depth when total depth was less than 80 cm and at 0.2 and 0.8 of the total depth for deeper locations. Substrate was classified according to PHABSIM code.

3.2 Univariate suitability curves development

Univariate suitability curves were defined for each life stage according to the procedure outlined by Bovee (1986):

i) each variable was divided into classes and frequencies of utilisation and availability were computed from the experimental data;

ii) the preferences for each class interval of the measured variable were computed from estimated relative frequencies of utilisation and availability as follows:

Pi = Ui/Ai

where:

Pi = relative preference value of a target species for a specific interval of the measured variable, Ui = % of utilisation of a specific interval of the measured variable, Ai = % of availability of a specific interval of the measured variable in the studied river sector at the time the organisms were sampled;

iii) to express the suitability curves, polynomial regression models were finally calculated using the relative preference values (Pi). For each life stage and variable, several polynomial functions of different orders were examined. The best model for

each data set was selected according to the coefficient of determination R^2 (>0.6) and to the significant level of the, function coefficients (p<0.05).

iv) the best fitting model was then normalised to the maximum value of 1.0;

  1. 3  Bivariate suitability curves development

We followed many of the same steps as the univariate regression for fitting data to a bivariate function.

i) The data collected were formatted into a 25 25 depth and velocity matrix.

ii) Velocity data were grouped in 20 cm s -1 intervals, from 0 to 100 cm s -1 ; depth data were grouped in 20 cm intervals, from 0 to 100 cm.

iii) For the curve-fitting we used a polynomial model of the general form:

P = aV + bD + cDV + dD2 +eV2....

where:

P = preference value for each combination depth-velocity,

V = mean column velocity,

D = water column depth, a,b,c, ... = equation coefficients.

The data were fitted, using a least square regression technique, to several forms of polynomial model by varying the order of the depth and velocity terms and adding or removing the interaction term. From the several models investigated two final best-fit models were selected, one for the adults and the other for the juveniles. As in the univariate case, the selection of the best model was based on the significant level of the coefficients (t-test p<0.05) and on the coefficient of determination (R2) of the regressions.

3.4 Bivariate vs univariate

Since PHABSIM analysis is based on univariate suitability functions for habitat modelling, an aggregation of the preference associated with depth, velocity and substrate class is needed.

This composite suitability can be computed for every single cell by means of three different ways of aggregation:

- multiplicative Ci = Vi Di Si
- geometric mean Ci = (Vi Di Si)1/3
- minimum factor Ci = mm (Vi Di Si)

Once the composite suitability Ci has been determined, then the amount of Weighted Usable Area (WUA) is computed multiplying each area cell (Ai) by the respective composite suitability factor (Ci) according to the following equation:

n

                 WUA = å   Ai Ci

i =1

These criteria of aggregation are based on the assumption that all the parameters act equally in defining the habitat suitability for the target fish species.

We chose to compare bivariate versus univariate in terms of WUA results of the bivariate and combined multiplicative functions, which are the most widely used among the three aggregation techniques available in PHABSIM.

We selected a representative reach, inside the study area, and ran through a MANSQ model for the hydraulic simulation. WUA were then computed for 21 discharges ranging from 0.7 to 62.3 m3 s-1.

4. Results and discussion

4.1 Univariate Habitat Suitability Curves

Current velocity and water depth ranged, respectively, from 0 to 100 cm s-1 and from 0 to 90 cm. The substrate resulted mainly composed of gravel (2-62 mm), cobble (64-250 mm) and boulder (250-4000 mm), these three granulometric classes corresponding to the 5th, 6th and 7th class codes adopted by PHABSIM. A preliminary data analysis showed no correlation between the hydraulic variables (R2 <0.06).

The HSC developed by means of polynomial regression on the preference values are shown in Figure 1. Both adults and juveniles depth suitability curves rose from approximately 5-10 cm to an optimum of 1.0 at 90 -100 cm.

The suitability increment with respect to the increase of water depth is lower in the adults curve than in the juveniles curve, thus suggesting the greater tolerance of smaller individuals for shallow waters. We have no representative observations of depths greater than 90 cm, anyway we assume deep water not to be a limiting factor for brown trout in the studied river.

Water velocity suitability curves calculated for both life stages show optimum values for low current velocities (<20 cm s-1). As velocity increases above the optimum values, the juveniles preference decreases regularly, while the adults curve is characterised by relatively high values of preferences (>0.5) for almost all the sampled velocities above 20 cm s-1.

According to literature these results show differences between the habitat requirements of the two life stages studied, adults preferring deeper and faster waters in order to find protection from visual predators while juveniles select shallower water in which they find protection from bigger aquatic predators. The low number of substrate classes that characterise the studied area makes meaningless the fitting of the preference data by means of polynomial curve. By inspecting the substrate preference values (Figure 2), it is however evident that both life stages show low preference values for substrate class 6 despite the high frequency of its availability. This can be partly explained with the fish selective use of different substrate depending on day-time and activity; for example fines (class 5) can be preferred during feeding activities while boulders (class 7) can offer cover areas.

4.2 Bivariate Habitat Suitability Curves

The best-fit bivariate models we came up with are:

- for the juvenile life stage: the model has a first-order depth term, a first-order interaction term and a coefficient of determination (R2) of 0.55.

- for the adult life stage: the model has a second-order depth term, a first-order interaction term and a coefficient of determination (R2) of 0.70.

The normalised response surfaces are shown in Figures 3 and 4.

Both models show maximum suitability values for depth over 90 cm and velocity of 0 cm s-1, with a consistent decline in preference with increasing velocities.

4.3 Univariate vs Bivariate

As Figure 5 shows, it seems absolutely evident that the WUA-discharge relationship obtained from bivariate models is consistently different from the one obtained from multiplicative aggregation of univariate functions. Since all these   response curves quite well resemble a "break-point type" curve, where WUA increases rapidly with until it reaches a fairly abrupt change of slope, break-point positions are definitely different.

In order to get a better perception of this curve pattern, WUA raw values from bivariate models have been fitted with a second order polynomial function (R2>0.99) (Figure 5.a, 5.b).

As outlined in the graphs, the main difference between the two aggregation methods, the multiplicative univariate and the bivariate, is how depth weights with respect to velocity.

For the multiplicative aggregation depth and velocity S.l. have the same weight and this determines, for both life-stages, a break-point at much lower flows than the bivariate cases. As a matter of fact, WUA goes up fast until the break-point, because with the flow increase, the velocity remaining quite low and therefore suitable, there is a gain in depth; above the break-point velocity starts to be unsuitable and this fact induces the flattening of the curve because every increase in depth is paid with a negative effect due to an increase in velocity.

For bivariate models the depth variable is the crucial one, velocity starting to be substantial at flows above 22.6 m3 s-1. Over this flow WUA keeps expanding even if at a much slower rate. This behaviour of the WUA seems to be reasonable if we consider that with an increase in flow there is always a gain in habitat along the river edges where the negative effect, due to velocity, has to be necessarily less significant.

Moreover it seems unrealistic to believe that WUA remains constant, as univariate models show in Figure 5, when the flow increases up to 5 or 6 times, even if these flows are absolutely extreme and definitely unusual in a regulated stream. Investigating the WUA response curves in a flow range more typical in a regulated river (from 0.7 to 28.3 m3 s-1) it should be noted that WUA resulting from bivariate models and univariate aggregation for adults are quite compatible, juvenile curve resulting to be peculiar.

This similarity is due to the initial dominance of the depth increasing effect up to a flow of 12.7 m3 s-1, being velocity very low and therefore suitable.

On the other hand depth demand for juveniles, using a univariate response curve, reveals to be much lower than in the bivariate case while velocity effect starts earlier at a flow of 6.5 m3 s-1. This difference in behaviour between adults and juveniles, delineated by univariate aggregations, seems to be much more reasonable from a biological point of view than the sharp correspondence revealed from the bivariate models. This lack of sensitivity between the bivariate models could perhaps be related to the low coefficient of determination of the juvenile model.

5. Conclusions

Using PHABSIM methodology to assess minimum flow in regulated rivers demands the development of Habitat Suitability Criteria. From literature there is a good evidence of the site-specificity of these suitability criteria therefore developing or testing HSC is definitely crucial critical for final results of a PHABSIM analysis. In PHABSIM habitat modelling, these univariate functions need to be aggregated to produce the WUA-discharge relationship. A multiplicative aggregation technique is generally used for the combined suitability, which is based on the assumption that all the parameters characterising the habitat act equally in defining the habitat suitability for fish population.

Since current velocity and water depth turned out to be the key factors for habitat modelling in our study site, we compared combined multiplicative suitability with a bivariate model, in terms of Weigthed Usable Area (WUA).

Significant differences between the two cases are evident. Even if all the curves are "break-point type", the bivariate ones have their break-point located at higher flows. This is a direct consequence of the crucial role played in these models by the depth parameter. Velocity in bivariate models is limiting only when the flow goes over 30 m3 s-1. Contrasting with the multiplicative combined suitability, bivariate models are actually an aggregation extrapolated by field measurements; as far as our study is concerned this fact should validate as realistic the dominance of depth variable over velocity. Finally it should be stressed that a break-point criterion to define minimum flow is definitely unsuitable for such curves and alternative criteria should be adopted.

REFERENCES

Bovee, K.D. 1978. "Probability-of-use criteria for the family Salmonidae". Instream Flow Information Paper 4, U.S. Fish and Wildlife Service, FWS./OBS-78107. Bovee, K.D., 1982. "A guide to Stream Habitat Analysis using the Instream Flow Incremental Methodology". Instream Flow Information Paper 12, U.S. Fish and Wildlife Service, FWS./OBS-82186.

Bovee, K.D., 1986. "Development and evaluation of habitat suitability criteria for use in the instream flow incremental methodology". Instream Flow Information Paper 21, Rep. 86 (7), U.S. Fish and Wildlife Service.

Greenberg, L., Svendsen, P., and Harby, A. 1996. "Availability of microhabitats and their use by brown trout (Salmo trutta) and grayling (Thymallus thymallus) in the river Vojman, Sweden". Regulated Rivers, 12, 287-303.

Heggens, J., Saltveit, S.J., 1990. "Seasonal and spatial microhabitat selection and segregation in youmg Atlantic salmon, Salmo salar L., and Brown trout, Salmo trutta L., in a Norwegian River". Journal of Fish Biology, 36, 707-720.

Lambert, T.R., Hanson, F.D., and Morhardt, J.E., 1987. "Development of habitat suitability criteria for trout in the Kings river Basin, California". Pacific Gas and Electric Company, California. Report 009.4-87.10.

Milhous, R.T., Wegner, D.L., Waddle, T., 1984. "User guide to the Physical Habitat Simulation System (PHABSIM)". Instream Flow Information Paper 11. Rep. FWS/OBS - 81/43. U.S. Fish and Wildlife Service.

Morhardt, J.E. 1986. "Instream Flow Methodologies". Rep. EPRI EA 48-19, prepared by EA Engineering, Science and Technology, Inc., for the Electric Power Research Instuitute, Palo Alto, CA.

Orth, D.J., Maughan, O.E., 1982. "Evaluation of the incremental methodology for recommending instream flows for fishes". Transactions of American Fisheries Society, 111:413-445.

Orth, D.J., Leonard, P.M., 1990. "Comparison of discharge methods and habitat optimization for recommending instream flows to protect fish habitat". Regulated Rivers, vol. 5,129-138.