Implications of seasonal variation for biomonitoring

Color profile: Disabled Composite Default screen

1411

Implications of seasonal variation for biomonitoring with predictive models in the Fraser River catchment, British Columbia Pamela F. Reece, Trefor B. Reynoldson, John S. Richardson, and David M. Rosenberg

Abstract: Reference-condition models for the Fraser River catchment were developed using samples collected during the autumn of 1994, 1995, and 1996. The goal of this study was to examine applicability to the reference-condition models of samples collected in other seasons and the effect of taxonomic resolution (genus and family) on model sensitivity to seasonal variation in the benthic invertebrate assemblage. Samples from eight streams representing interior, coastal, and large-river habitats were collected in spring, summer, autumn, and winter of 1995 and in spring of 1996. The benthic invertebrate assemblage changed seasonally such that the models could not be used for seasons other than autumn. The models were equally sensitive to seasonal variation when genus-level or family-level data were used. We recommend that test samples (i.e., samples collected from disturbed sites and meant for comparison with the reference database) be collected either during the autumn or over multiple sampling dates to reduce the possibility that seasonal shifts or stochastic events will lead to erroneous conclusions about the state of a test site. Résumé : Des échantillons recueillis aux automnes 1994, 1995 et 1996 ont servi à élaborer des modèles avec conditions de référence dans le bassin hydrographique du fleuve Fraser. Les buts de l’étude sont de vérifier si des échantillons prélevés en d’autres saisons peuvent être appliqués aux modèles avec conditions de référence et d’examiner les effets du degré de résolution taxonomique (genre et famille) sur la sensibilité des modèles à la variation saisonnière de la communauté des invertébrés benthiques. Des échantillons ont été prélevés dans huit cours d’eau représentant des habitats de la côte, de l’intérieur des terres et des grandes rivières, au printemps, à l’été, à l’automne et à l’hiver 1995 et au printemps 1996. Les changements saisonniers de la communauté d’invertébrés benthiques sont tels que les modèles ne peuvent être utilisés qu’à l’automne. Les modèles restent aussi sensibles à la variation saisonnière, que les données soient identifiées au genre ou à la famille. Nous recommandons donc que les échantillons d’essai (i.e. les échantillons provenant de sites perturbés destinés à être comparés aux données de référence) soient récoltés à l’automne ou alors à plusieurs périodes durant l’année pour minimiser la possibilité que des changements saisonniers ou des événements aléatoires mènent à des conclusions erronées sur l’état du site à analyser. [Traduit par la Rédaction]

Reece et al.

1418

Introduction The reference-condition approach to river biomonitoring uses data on benthic invertebrate assemblages and associated physical, chemical, and biological conditions from a set of minimally impacted or reference sites to predict the expected benthic assemblage at possibly stressed test sites (Reynoldson Received March 22, 2000. Accepted March 31, 2001. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on June 12, 2001. J15675 P.F. Reece1 and J.S. Richardson. Department of Forest Sciences, 2424 Main Mall, University of British Columbia, Vancouver, BC V6T 1Z4, Canada. T.B. Reynoldson. Environment Canada, National Water Research Institute, 867 Lakeshore Road, P.O. Box 5050, Burlington, ON L7R 4A6, Canada. D.M. Rosenberg. Department of Fisheries and Oceans, Freshwater Institute, 501 University Crescent, Winnipeg, MB R3T 2N6, Canada. 1

Corresponding author (e-mail: [email protected]).

Can. J. Fish. Aquat. Sci. 58: 1411–1418 (2001)

J:\cjfas\cjfas58\cjfas-07\F01-076.vp Thursday, June 07, 2001 9:34:09 AM

et al. 1997). If a test-site assemblage falls within the range of variability found at reference sites with similar environmental conditions, the site is considered equivalent to reference. If the test site falls outside of the range, it is considered different from the reference condition. Accurate predictions of expected benthic assemblages are limited to the range of variability included within the reference database, which may or may not include seasonal variability. Benthic communities change seasonally (Hynes 1970; Boulton and Lake 1992; but see Death 1995), but the implications for biomonitoring have not been fully established. Furse et al. (1984) recommended seasonal collection of reference samples when using predictive models. Their use of a reference data set that combined seasonal data for each site increased prediction accuracy of test sites. However, the additional time and cost of collecting and identifying samples from a large number of reference sites over several seasons are significant considerations that constrain the scope of biomonitoring programs with limited resources. Therefore, the implication of seasonal changes in the benthic community to the application of the reference-condition approach needs to be established.

DOI: 10.1139/cjfas-58-7-1411

© 2001 NRC Canada


1412

The purpose of this study was to determine the effect of seasonal changes in benthic invertebrate assemblages on the accuracy of predictions made with the BEnthic Assessment of SedimenT (BEAST) predictive model (Reynoldson et al. 1995). The model has been developed for bioassessment in the Fraser River catchment, Canada (Reynoldson et al. 2001). Specifically, we tested the effect of seasonal variation on the assessment of samples collected outside of the autumnal reference-sample period. A second goal of this study was to determine the effect of taxonomic resolution (genus versus family) on the sensitivity of the model to seasonal variation. Seasonal variation should be manifest primarily at the genus level because representatives of families are likely to be present throughout the year. The most suitable level of invertebrate identification for biomonitoring depends on the study and the questions being asked (Resh and McElravy 1993). Zamora-Muñoz and AlbaTercedor (1996) found that the family level of identification was sufficient for monitoring water quality in streams, but species data were required to determine the exact biological responses to stress. Monitoring at the family level is used by the Australian River Assessment Scheme (AusRivAS) and has been recommended for marine (Warwick 1993) and lake (Jackson and Harvey 1993) ecosystems.

Can. J. Fish. Aquat. Sci. Vol. 58, 2001 Fig. 1. Location of sampling sites in southwestern British Columbia. All sampling sites ( ) are located in the Fraser River catchment.

Methods Study sites To ensure that the results of this study applied to streams from different ecoregions and to small streams as well as large rivers, seasonal change in assemblage structure was assessed for eight streams. This study was limited to eight streams for logistical reasons. All eight streams are located in the Fraser River catchment (Fig. 1) and include three continental streams (Mellin, Glimpse, and Beak creeks), three coastal streams (Spring Creek, Mayfly Creek, and the North Alouette River), and two large-river sites (Fraser and Thompson rivers). These study sites will be referred to as seasonal test sites. Six of the streams used as seasonal test sites were considered reference sites because they were not visibly disturbed, and their autumnal data were incorporated in the reference database used in developing the predictive models described in Reynoldson et al. (2001). However, during sample collection, it was decided that the Fraser River and Mellin Creek sites were possibly impaired. As a result, they were treated as disturbed sites for data analysis. A detailed description of the study sites along with a discussion of the spatial variation of the benthic assemblage can be found in Reece and Richardson (2000). However, a few details of the study sites that highlight differences between the stream types are given here. Continental climate streams receive very little precipitation through the summer, and highest discharge occurs in the spring as a result of snowmelt. Air temperatures range from 35°C in the summer to –30°C in the winter. These are hardwater streams with high alkalinity, high conductivity, and a slightly basic pH (Table 1). Agriculture is a dominant land use in continental regions of B.C. Mellin Creek was heavily used by cattle in the autumn and was therefore treated as a disturbed site and excluded from the reference database. Beak Creek was inaccessible in the winter. The coastal climate streams receive heavy winter rains and correspondingly high and variable winter discharges. Air temperatures can range from 30°C in the summer to –10°C in the winter. The streams have soft water, low alkalinity, low conductivity, and are slightly acidic (Table 1).

The two large-river sampling sites were the Thompson River at Spences Bridge and the Fraser River at Agassiz. Air temperatures at these sites range from 30°C in the summer to –20°C in the winter. The Fraser and the Thompson rivers have intermediate alkalinity and conductivity and pH similar to the those of the interior streams (Table 1). Land use around Spences Bridge and Agassiz consists of agriculture, some rural development, and forestry. There are major urban centers and pulp mills upstream of both sampling sites. Long-term studies on the invertebrate assemblage at Spences Bridge have indicated that the Thompson River is not heavily impacted at this point (J. Culp, Environment Canada, National Hydrology Research Institute, Saskatoon, Saskatchewan, personal communication), so it was considered suitable for inclusion in the reference database. The state of the Fraser River at Agassiz was unknown, so it was treated as a disturbed site and was excluded from the reference database.

Sample collection The purpose of this study was to test the implications of seasonal variation on the use of predictive models created for biomonitoring in the Fraser catchment, so the same sample collection techniques were used as for collection of the reference database (Reynoldson et al. 2001). However, samples were collected over five seasonal sampling dates: late spring, summer, autumn, and winter of 1995 and early spring of 1996. The autumnal sampling date corresponded with the timing of the 1994 to 1996 collections of reference samples from throughout the Fraser River catchment (Rosenberg et al. 1999). Benthic invertebrate samples were collected from riffles using kicknets (400-mm mesh). In the large rivers, kicknet samples were © 2001 NRC Canada

J:\cjfas\cjfas58\cjfas-07\F01-076.vp Thursday, June 07, 2001 9:34:10 AM


Reece et al.

1413

50° 25 28 121° 19 35.3 204.2 54900 0.0102 70–235.7 0.3–0.5 0.152–1.41 409–1630 0.5–18 7.58–8.65 78.5–110.6 29.2–42.2 0.032–0.148 0.069–0.127 0.96–3.41 0.006–0.017

¢

¢

¢

49° 12 20.2 121° 46 7.2 68.7 217000 0.0002 300–400.9 0.4–1 0.225–1.72 1060–4480 0–17.5 6.94–8.08 95.6–218 39.2–56.9 0.005–0.156 0.118–0.39 9.46–62.51 0.028–0.101

¢¢ ¢¢

Fraser

¢

Large rivers

¢¢

Thompson

¢¢

taken from the river edge proceeding into the river channel, whereas the small-stream samples were collected by zigzagging between stream banks. Samples were preserved in 4% buffered formaldehyde. One-minute kicknet samples were collected at each site and date. Three samples per site per date were sorted and identified, and the data were pooled so that the data used represented invertebrate abundance per 3-min kicknet sample. If a sample contained 500 invertebrates were subsampled (Marchant 1989) and at least 200 animals were removed (Rosenberg et al. 1999). Organisms were identified to the lowest practical taxonomic level, usually genus and species. However, Ostracoda, Turbellaria, and Copepoda were not identified beyond class. Identifications were verified by experts (when possible), and a voucher series was deposited in the Royal British Columbia Museum, Victoria. Forty environmental variables, including geographic, site, and channel characteristics and water chemistry, were measured either when the benthic samples were collected or from maps. The environmental variables measured were the same as those collected for the parent study (Rosenberg et al. 1999; Reynoldson et al. 2001).

49° 17 0.8 122° 32 0.6 315 9.58 0.0400 3.8–15 0.21–0.38 0.43–1.01 0.436–1.14 2–11 5.65–6.68 13–21.23 0.1–4.5 0.096–0.147 0.037–0.075 0.06–1.2 0.000–0.004

¢¢

¢

¢

N. Alouette

¢¢

¢¢

49° 19 02 122° 32 0.4 413.5 2.14 0.0020 2.9–3.84 0.11–0.88 0.08–0.67 0.04–0.52 2–10 5.71–7.68 15.22–24.2 2.4–10.2 0.036–0.1 0.04–0.092 0.29–1.76 0.002–0.006

¢¢

¢

Mayfly

¢

Data analysis

¢¢

49° 16 0.3 122° 34 0.4 132.9 1.99 0.0319 3.1–8.7 0.03–0.99 0.23–0.29 0.04–0.3 3.8–10 6.37–6.73 16.4–30 4.9–12.1 0.032–0.199 0.03–0.19 0.38–2.2 0.001–0.005

¢¢

Beak

50° 6 45.4 119° 58 51 1003 86.41 0.0170 6.8–13.42 0.106–0.27 0.15–1.06 0.047–1.405 0.1–11.5 7.44–8.45 75.7–234.2 30.2–86 0.005–0.06 0.21–0.48 1.56–3.56 0.028–0.034

Glimpse

50° 15 2 120° 14 9 1171.1 10.18 0.0025 1.3–2.8 0.08–0.22 0.07–1.18 0.004–0.31 0.1–8.5 7.69–8.25 234.4–534.9 123–201 0.005–0.057 0.286–0.34 1.35–3.05 0.007–0.016 50° 7 1.1 120° 6 28.8 1123.1 52.35 0.0204 0.15–5.8 0.07–0.41 0.09–0.93 0.006–1.44 0.1–14.5 7.21–8.1 72.1–262.7 40.2–97.6 0.005–0.071 0.487–0.674 4.33–32.9 0.045–0.089

¢

m

Latitude (N) Longitude (W) Elevation (m asl) Drainage basin (km2) Channel slope (m·m–1) Channel width (m) Mean depth (m) Max. velocity (m·s–1) Discharge (m3·s–1) Temperature (°C) pH Conductivity ( S·cm–1 at 25°C) Alkalinity (CaCO3) (mg·L–1) NO3NO2–N (mg·L–1) TKN (mg·L–1) Total suspended solids (mg·L–1) Total phosphorous (mg·L–1)

¢

Mellin

¢¢

Continental

¢¢

Stream class

¢¢ ¢ ¢¢ ¢

Table 1. Physical and environmental characteristics for each study stream.

¢¢ ¢ ¢¢ ¢

Note: Ranges are given for variables that change seasonally. TKN = total Kjeldahl nitrogen, asl = above sea level.

Coastal

¢

Spring

¢

Invertebrate abundance data were analyzed at both genus and family levels (exceptions listed above) to determine the sensitivity of the assemblage to seasonal changes. Invertebrate abundance data were used to test the implications of seasonal variation on the predictive model because abundance data were used in the creation and implementation of the model. Abundance data creates more sensitive models than presence–absence data (Armitage et al. 1987; Smith et al. 1999). Genus level was used because not all larvae could be identified to species. Identification to family is time and cost efficient, achievable for most taxa, and can be consistent among laboratories. Both genus and family data sets and reference groups used for the analysis were the same as those used in Reynoldson et al. (2001). Rare taxa (occurrence at