Read Me File to support analyses in manuscript "Dynamic Species Co-Occurrence Networks Require Dynamic Biodiversity Surrogates" by Tulloch, Ayesha; Chades, iadine; Dujardin, Yann; Westgate, Martin; Lane, Peter; Lindenmayer, David in Ecography 2016 doi: 10.1111/ecog.02143 Please cite this paper when using the data or code. ############################################################ Datasets ############################################################ 1. Southwest Slopes [see file "SWslopes_plantings_birdsurveysPA_2002to2013.xlsx"] The Southwest Slopes is a region of Australian temperate woodlands in southern New South Wales that has been heavily modified due to clearing for agriculture. Revegetation through either new plantings on cleared land or as enhancement plantings of existing remnants has occurred on 28 farms across a broad band over 6800km2 long since the 1990s, with the objective of restoring endangered Box Gum Grassy Woodland communities, resulting in increased woody vegetation cover and changes to key hollow and food resources. We are interested in finding the best surrogate species for all bird species responding to revegetation. We use an extensive longitudinal dataset gathered over 12 years from 2002 to 2013 from repeated surveys of birds on 65 patches of revegetated woodland (plantings; 708 surveys), in which 150 bird species have been detected (see Supplementary Material for details). Surveys are conducted in spring, and additional winter surveys conducted during five of the 11 years. 2. Booderee National Park Heathland, Jervis Bay [see file "Booderee_heath_birdsurveysPA_2003to3013.xlsx"] Booderee National Park (NP) is a 75km2 IUCN Category I reserve located on the south-east coast of Australia (~35°10'S, 15°40'E). It is co-managed by Parks Australia (a section of the Australian Federal Government's Department of the Environment) and the Wreck Bay Aboriginal Community. The area has a temperate climate with vegetation types ranging from dry heathland to woodland to rainforest, and there is a well-documented fire history, with wildfires burning the Park on average once every 15-20 years as well as controlled burns for biodiversity management. A large wildfire burnt 52 % of the Park in 2003, reducing vegetation cover and changing the composition of the plant community (due to the dependence of many Australian plants on fire for flowering and reproduction; D. Lindenmayer, unpublished data). We are interested in finding the best surrogates for all bird species responding to fire and its effects on vegetation. We select the heathland for this study as it is regularly burnt and is the stronghold of the nationally Endangered Eastern Bristlebird Dasyornis brachypterus. We use a longitudinal dataset monitoring 26 heathland sites during spring annually over 11 years from 2003 to 2013 (excluding 2008), detecting 90 bird species over the course of the surveys. Note: The full species lists for each case study dataset, plus detection rates and threat status, are in the file "Species_lists.xlsx". ############################################################ Description of methods for calculating co-occurrence matrices ############################################################ We adapt the approach of Lane et al. (2014) to calculate the surrogacy value of each species, sij, which represents the amount of information that surrogate species i provides on target species j. The input is a presence/absence matrix of species (m) by surveys (q). The final output used for optimising decision-making is an m-by-m surrogacy matrix of values sij for each species in the range [0, 1] quantifying the strength of any positive relationship between species i and j. When sij = 0, the presence of species j is not associated with the presence of species i, whereas a value close to 1 means that species i is a good surrogate for species j. The surrogacy value of a species for itself, sii, is 1. To derive the m-by-m surrogacy matrix, we follow the following steps: Step 1. We first calculate the odds ratio between each pair of species using the R package sppairs. Odds ratios provide information on the strength and direction of associations. An odds ratio rij of 1 means that the presence of species i and species j are not associated in the set of surveys, while an rij < 1 means that the presence of species i is associated with the absence of species j. An odds ratio rij > 1 means that species i is a potential surrogate for species j (i.e. an indicator of presence). Note that unlike correlations, these odds ratios are not symmetrical: rij may be larger or smaller than rji, depending on the relative frequency of occurrence of species i and j. Step 2. Our second step is to set all negative associations and all insubstantial species co-occurrences (i.e. those associations with an odds ratio of between 1/3 and 3) to zero; this excludes small effects. An alternative approach for removing insubstantial co-occurrence relationships is to only include statistically significant co-occurrence relationships (this method is described in the Supporting Material for the paper; see also Gotelli et al. 2015 and Gotelli and Ulrich 2010). Step 3. We then make a final step of converting each positive odds ratio to a value between 0.5 and 1 using the formula sij = rij /(1+ rij). This allows all values to be standardised before optimisation, and ensures that the optimisation is not dominated by large odds ratios, which may derive from fortuitous co-occurrence of some moderately rare species. Gotelli, N. J. et al. 2015. EcoSimR: Null model analysis for ecological data. R package version 0.1.0. http://github.com/gotellilab/EcoSimR. Gotelli, N. J. and Ulrich, W. 2010. The empirical Bayes approach as a tool to identify non-random species associations. Oecologia 162: 463-477. Lane P.W., Lindenmayer D.B., Barton P.S., Blanchard W., Westgate M.J. 2014. Visualization of species pairwise associations: a case study of surrogacy in bird assemblages. Ecol Evol 4(16), 3279-3289. (doi:Doi 10.1002/Ece3.1182). ############################################################ Co-occurrence matrix names and data used to derive them ############################################################ 1) Southwest Slopes SW_matrix_s1_1to11yrs_alldata.txt: All data SW_matrix_s2a_1to3yrs.txt: First 3 years of monitoring dataset (2002 to 2005) SW_matrix_s2b_1to5yrs.1sthalf.txt: First half of monitoring dataset (5 years) SW_matrix_s2c_1to7yr.txt: First 7 years of monitoring dataset (2002 to 2009) SW_matrix_s2d_1to9yrs.txt: First 9 years of monitoring dataset (2002 to 2011) SW_matrix_s3_6to11yrs.2ndhalf.txt: Second half of monitoring dataset (6 years) SW_matrix_s4a_spring.txt: Only spring surveys SW_matrix_s4b_winter.txt: Only winter surveys SW_matrix_s5_1in2yrs.txt: Only the surveys from every 2nd sampling year SW_matrix_s6_random.txt: Random selection of half of all surveys 2) Booderee National Park BNP_matrix_s1_1to11yrs_alldata.txt: All data BNP_matrix_s2a_1to3yrs.txt: First 3 years of monitoring dataset (2003 to 2005) BNP_matrix_s2b_1to5yrs.1sthalf.txt: First half of monitoring dataset (5 years) BNP_matrix_s2c_1to7yr.txt: First 7 years of monitoring dataset (2003 to 2009) BNP_matrix_s2d_1to9yrs.txt: First 9 years of monitoring dataset (2003 to 2011) BNP_matrix_s3_6to11yrs.2ndhalf.txt: Second half of monitoring dataset (6 years) BNP_matrix_s5_1in2yrs.txt: Only the surveys from every 2nd sampling year BNP_matrix_s6_random.txt: Random selection of half of all surveys