Producing Generalised LIFE Response Curves

Clarke, Ralph; Dunbar, Michael. 2005 Producing Generalised LIFE Response Curves. Bristol, Environment Agency, 79pp. (Science Report SC990015/SR, SCHO0305BIQJ-E-P, SC990015)

Before downloading, please read NORA policies.

Preview

Text SCHO0305BIQJ-e-e.pdf Download (2MB)

Abstract/Summary

The LIFE methodology (Extence et al. 1999) has demonstrated that it is possible to link changes in benthic invertebrate community structure (as sampled routinely by Agency biologists) with indices of historical river flow at a gauge close to the sample site. This report investigates the response of LIFE observed / expected (O/E) score to preceding gauged flows through a linear modelling framework. Data were supplied by the Agency in Excel files, key data were extracted from the three main tables and loaded into a relational database. This initially included all available macro invertebrate sample data up until December 1999 and some data from 2000 and 2001. Later it was agreed to take advantage of the macro invertebrate data available since 2001, thus the LIFE score database was extended to the end of 2003. To match this, all flow indices were re-calculated for water years 1989-2002. Additional data derived by CEH were added to the database. RIVPACS was run for the majority of the sample sites, either using site characteristic data held already in the CEH Dorset Invertebrate Database, or the site characteristics provided with the data. The RIVPACS outputs were most probable RIVPACS group and expected LIFE scores, for spring, summer and autumn samples. The Institute of Hydrology Report 108 low flow quality classification was also added for each gauging station. Following exploratory data analysis, some sites were excluded from the data set on the basis that their characteristics (e.g. deep water and silty bed) make their macroinvertebrate community unlikely to respond to flow in the manner that the LIFE index was designed. Linear modelling demonstrated that autumn LIFE O/E score does indeed vary systematically with flow. Flow variables from the immediately preceding summer are the most important in explaining variation. The relative importance of high and low flow variables can depend on how they are standardised. The simplest models, explaining variation in LIFE score solely on preceding flows, confirm the validity of the LIFE approach, but do not explain a high proportion of overall variation in LIFE score. Various site-specific factors have been used to improve the fit of the model, most notably splitting the sites into categories based on base flow index, and also whether there are significant artificial influences in the catchment. These give model R² values of between 0.1 (high BFI) to 0.2 (low BFI). Adding immediately preceding spring sample LIFE O/E as an explanatory variable increases R² to 0.4. A “site” factor encompassing all unexplained variation in mean LIFE O/E increases R² to 0.6, an encouraging result, indicating a maximum value for R² but still retaining a common slope. A simple approach adding RIVPACS group (two to four categories) as an interaction term (i.e. affecting slope of response) did not add to the model fit. There is a trend for higher BFI sites to show negative relationships between LIFE O/E and winter Q10. These sites are also more strongly associated with Q95 flows from the previous summer, although this trend is weak, and is dependent on the method of flow standardisation. Both these relationships can easily be masked by inter-site differences. Analysis of autocorrelation of residuals from the model suggested that the year to year correlation of autumn LIFE O/E is largely due to the correlation in the explanatory flow variables. Also, there was little evidence for the perceived greater lag of baseflow-dominated catchments, beyond that explainable by the lag in the flows themselves. In general, low flow duration indices are probably not the most sensitive indicators of LIFE response on baseflow dominated catchments, alternative indicators emphasising drought duration could give improved model fits. There is still clearly considerable unexplained site to site variation in the O/E LIFE scores, this is illustrated by the wide variation in slopes of the individual sites’ LIFE O/E vs flow relationship. For the analysis undertaken, the data supported common LIFE vs flow slopes for all sites, although there was evidence of a interaction effect between artificial influences and preceding summer Q10: the more influenced catchments had more depressed LIFE O/E values when summer Q10 was low, but had LIFE O/E scores similar to the less influenced catchments when summer Q10 was high. This was not the case for LIFE vs preceding summer Q95. Analysis of replicate data taken within a season has allowed calculation of an average total within-season standard deviation of LIFE score, consistent with previous work this decreases with number of taxa observed. The average total within-season variance was compared with the total mean squares of LIFE score to give an indication of the maximum potential model R² possible, this was in the region of 0.75: the quoted R² values for models should be viewed with this in mind. Unexplained variation in mean LIFE O/E can hopefully be tackled by improvements to RIVPACS, perhaps by incorporating catchment characteristics from digital datasets. Unexplained variation in the slope of response of LIFE to flow could be tacked in several ways, including more sophisticated application of the RIVPACS groupings. In addition, it is likely that adding additional site factors such as habitat types and / or channel geometry would improve the methodology. This unexplained variation will need to be addressed for single LIFE O/Es to be a useful tool in determining abstraction / flow stress without supporting information. However the models as they stand would be extremely useful in assisting determining response where small amounts of data only are available.

Item Type:	Publication - Report
Programmes:	CEH Programmes pre-2009 publications > Water > WA01 Water extremes > WA01.4 Ecological responses to extremes
UKCEH and CEH Sections/Science Areas:	_ Hydrological Risks & Resources UKCEH Fellows
ISBN:	1844323714
Funders/Sponsors:	Environment Agency
Additional Keywords:	LIFE, RIVPACS, Biological monitoring, macroinvertebrates, low flows, ecological stress, linear modelling, flow monitoring, base flow index
NORA Subject Terms:	Ecology and Environment Hydrology
Date made live:	26 Mar 2008 11:13 +0 (UTC)
URI:	https://nora.nerc.ac.uk/id/eprint/2303

Actions (login required)

View Item

Document Downloads

More statistics for this item...