Environ Monit Assess (2014) 186:4685–4695 DOI 10.1007/s10661-014-3730-9
Assessing the ecological condition of streams in a southeastern Brazilian basin using a probabilistic monitoring design Juliana Jiménez-Valencia & Philip R. Kaufmann & Ana Sattamini & Riccardo Mugnai & Darcilio Fernandes Baptista
Received: 30 June 2013 / Accepted: 18 March 2014 / Published online: 15 May 2014 # Springer International Publishing Switzerland 2014
Abstract Prompt assessment and management actions are required if we are to reduce the current rapid loss of habitat and biodiversity worldwide. Statistically valid quantification of the biota and habitat condition in water bodies are prerequisites for rigorous assessment of aquatic biodiversity and habitat. We assessed the ecological condition of streams in a southeastern Brazilian basin. We quantified the percentage of stream length in good, fair, and poor ecological condition according to benthic macroinvertebrate assemblage. We assessed the risk of finding degraded ecological condition associated with degraded aquatic riparian physical habitat condition, watershed condition, and water quality. We describe field sampling and implementation issues encountered in our survey and discuss design options to remedy them. Survey sample sites were selected using a spatially balanced, stratified random design, which enabled us to put confidence bounds on the ecological condition estimates derived from the stream survey. J. Jiménez-Valencia (*) : D. F. Baptista Programa de Pós-Graduação em Ecologia, Instituto de Biologia, UFRJ, Ilha do Fundão, Rio de Janeiro CEP 21941-590, Brazil e-mail:
[email protected] J. Jiménez-Valencia : A. Sattamini : R. Mugnai : D. F. Baptista Laboratório de Avaliação e Promoção da Saúde Ambiental—IOC/FIOCRUZ, Rio de Janeiro CEP 21040-360, Brazil P. R. Kaufmann U.S. Environmental Protection Agency, 200 Southwest 35th Street, Corvallis, OR 97333, USA
The benthic condition index indicated that 62 % of stream length in the basin was in poor ecological condition, and 13 % of stream length was in fair condition. The risk of finding degraded biological condition when the riparian vegetation and forests in upstream catchments were degraded was 2.5 and 4 times higher, compared to streams rated as good for the same stressors. We demonstrated that the GRTS statistical sampling method can be used routinely in Brazilian rain forests and other South American regions with similar conditions. This survey establishes an initial baseline for monitoring the condition and trends of streams in the region. Keywords Ecological assessment . Streams . Probabilistic monitoring design . Multimetric index . Relative risk
Introduction Estimating the ecological condition of aquatic ecosystems is a conservation priority for South America’s very diverse ecosystems such as the Brazilian Atlantic Forest (Mast et al. 1997) because anthropogenic pressures have left few unaffected ecosystems (Begon et al. 2006). Scientific inferences about the condition of widespread freshwater ecosystems are more robust when sampling designs are able to aid the assessment of broad concerns (Stevens and Olsen 2000). National and state water quality surveys have shown that appropriate biological indicators and spatially balanced statistical sampling designs that employ randomized site selection are
4686
necessary for minimizing bias in large-scale assessments of freshwaters (Paulsen et al. 1998, 2008; Whittier et al. 2002; Stevens 2002). Spatial balance ensures that there is minimal effect of spatial correlation on parameter estimates (Dobbie et al. 2008; Tobler 1970). In addition, statistical (probability) sampling provides the basis to infer characteristics from a representative sample of waters (Kish 1967; Cochran 1977; Scheaffer et al. 1986) in a basin, state, or ecoregion to the entire population of lakes or streams (Larsen et al. 1994; Kaufmann et al. 1991; Baker et al. 1991) in the region represented. Although probability-based surveys demonstrably improve the quality of regional estimates and avoid misleading biases, they have only gained favor for fresh water monitoring programs in the last two decades (Dobbie et al. 2008), as for example in the aquatic resource assessments of the United States Environmental Protection Agency (USEPA 2009, 2013). We initiated a survey of the ecological condition of streams in a southeastern Brazilian basin. This survey was part of a series of studies (Baptista et al. 2007; Mugnai et al. 2010; Buss and Vitorino 2010; Baptista et al. 2011; Oliveira et al. 2011a, b; Mugnai et al. 2011) meant to create cost-effective tools for biological monitoring programs in streams of the region. Our objectives were (1) to assess the ecological condition of a basin in southeastern Brazil by quantifying the percentage of the stream length (kilometers) affected by human-induced stress, (2) to quantify the extent of stressors and the risk of their association with degraded biological condition, and (3) to address common field sampling implementation issues and some options to deal with them from a design-based perspective. Minimizing bias is critical in any policy-relevant assessment of resource condition. We rejected professional judgment and convenience-based sample site selection because they can consciously or unconsciously deviate results, regardless of the number of sites sampled (Whittier et al. 2002). Probability sampling also allowed us to quantify the bounds of uncertainty of our regional population estimates (Cochran 1977). It provided a basis for determining the feasibility and cost of changes in monitoring design based on our tolerance of uncertainty. Another objective relevant to the selection of monitoring designs was the potential for aggregating and comparing ecological data with neighboring basins (IMST 2009), foreseeing the possibility of collaborative
Environ Monit Assess (2014) 186:4685–4695
programs that would encompass larger geographical areas (e.g., Mulvey et al. 2009). To respond to these objectives and concerns, we chose a probabilistic monitoring design with spatial balance to select the locations we used to assess the basin.
Methods Study area We conducted our assessment in a southeastern Brazil basin that covers 1,260 km2 in the state of Rio de Janeiro (RJ) (Fig. 1). The springs in the Guapiaçu-Macacu Basin emerge from a mountain range at elevations between 0 and 1,400 m above sea level. The basin is approximately 40 % Brazilian tropical rainforest (Atlantic Forest: Mata Atlântica) and is identified as a biodiversity hot spot (Mast et al. 1997). The area varies from tropical wet forest at lower elevations to mild mesothermal forest above 1,200 m elevation, with average February temperatures of 15–28 °C and average July temperatures 77 (11.56–85.71)
Poor
71–77