260 likes | 419 Views
Michael J. Dunbar , Simon M. Smart, Roger Baker, Ralph T. Clarke, Francois K. Edwards, Lindsay C. Maskell, John F. Murphy, Lisa R. Norton, Paul Scholefield. Freshwater Stressor-Response relationships: Countryside Survey.
E N D
Michael J. Dunbar, Simon M. Smart, Roger Baker, Ralph T. Clarke, Francois K. Edwards, Lindsay C. Maskell, John F. Murphy, Lisa R. Norton, Paul Scholefield Freshwater Stressor-Response relationships: Countryside Survey
The way we investigate ecological response to anthropogenic stressors tends to be very one-dimensional Sometimes people go to the other extreme: all available abiotic variables -> multivariate ecological data
Impact of land management (agriculture) Land cover studies • Spatial • Large-scale • Correlative • Lack mechanism Field experiments • Temporal • Water quality focus • Can we generalise?
Impacts arising from “use” of land Multiple mechanisms for impact Often partly correlated Probably interacting
Countryside Survey Monitored since 1978 (~8 year intervals) 600+ 1km² plots • Broad Habitats • Terrestrial vegetation • Soils • Freshwater • Headwater streams (~300 locations): 1990, 1998, 2007 only • Ponds Land cover map of GB
Statistical approach Past: separate “stock” and “change” estimates Stock: what’s there in any one survey, based on all the data Change: from survey to survey, based on common data points • Survey changes over time Alternative: • Mixed-effects approach: consistent
Freshwater: what’s measured? Headwater stream (1st to 3rd order) • Macroinvertebrates • Physical habitat modification and quality (RHS) • Spot chemistry sample • Macrophytes Also Ponds… 1990,1998,2007 1998,2007
Macroinvertebrates Standard net kick sample Identification in CEH labs Metrics used here: • Average Score per Taxon (ASPT)* • Number of scoring taxa (NTAXA)* • Community Conservation Index (CCI) * Reference condition corrected (O/E ratio)
Headwater Streams: macroinvertebrates • England: broad ongoing improvement in biological quality • Low scores in 1990 • Dunbar, M., Murphy, J., Clarke, R., Baker, R., Davies, C., Scarlett, P., 2010. Countryside Survey: Headwater Streams Report from 2007. Technical Report No. 8/07 NERC/Centre for Ecology & Hydrology 67pp. (CEH Project Number: C03259) • Fixed effect for year: multiple comparison problem: glht package in R <- Biological Metric -> <- Survey years ->
Overall approach Metric ~ Stressors + Environmental variables (including interactions) Multilevel (mixed effects) linear regression (all variables) • Separate spatial and temporal patterns Generalised additive models • Identify non-linearities: averaged 1998 and 2007 data
Examples of resectioning <- “Natural” channel Resectioned channel ->
Multilevel (mixed effects) linear regression Metric ~ Stressors + Environmental variables • Samples (years) nested within CS squares • Separate spatial and temporal patterns • Test selected two-way interactions • Has intensive land impact changed over time? • Broad Habitat % * Year • Do you see differing flow effects across different land cover types? • Broad Habitat % * Flow • Are more engineered channels more sensitive to high nutrient levels? • SRP * Resectioning
Results: main effects Negative effects • Soluble Reactive Phosphorus concentration • Extent of resectioning of the channel • % Arable Land Positive effects • Riparian woody cover • Channel substrate diversity Mixture of effects ?? • % Improved Grassland
Space vs Time Figure 2. Set 1 spatial (A) and temporal (B) relationships with land cover characteristics
GAMs on time-averaged data (non-linear) Results using a multi-model inference approach with five main effects and all combinations of two-way interactions (as bivariate smooths)
Conclusions Clear impact of stressors at three spatial scales on stream ecological quality and biodiversity Strongest impacts from local conditions: habitat and water quality Encouraging results given fairly basic biological indices Can pick up a beneficial effect of reduced riparian management
Statistical and data issues Non-linearities common Interactions sensitive to model chosen (mis-specification) Spatial patterns stronger than temporal General lack of data at the high stressor ends of the scale