5. 3 Apoyos dados por diferentes instancias del Gobierno Federal
GRÁFICA 8.- INVERSIÓN CONANP EN TURISMO POR REGIONES 1996-2004
Focus species and dataset
Initially, we obtained presence-only data for 3622 species downloaded from eBird (ebird.org) and this number was then narrowed down to 2460 species, which were the ones that had sufficient records for our analysis (i.e. at least 100 records within each time window), comprising a total of 241,039,916 occurrence records across all species. For each species, we grouped their
occurrence data from all years until 2016, and since we were interested in within-year patterns, we used in the analysis only the variable month — but not day or year.
Obtaining seasonal maps
We mapped the seasonal extent of occurrence (EOO) of species’ records and measured their spatial variation among 12 time-windows across the annual cycle. For example, if the records of a species were clustered in region one from January to April and region two from July to October, this could mean this species migrates annually between regions one and two. To this end, we pooled and mapped records of each species for 12 time-windows of four consecutive months each, each window starting with a different month from January to December (Figure 12). We used these time windows in six pairs, representing couples of seasons opposite to each other in the year calendar to maximise the difference between seasons. For instance, within the couples, each season may represent date ranges for the peak breeding and peak non-breeding seasons, but species migrations might be triggered by other factors too, such as escaping (Chapter 2). The two months between each seasonal range were excluded to minimise the number of data from
individuals on the migratory route (following Lees & Martin 2015).
Figure 12: Time windows showing the months of the year from January to December (clockwise, January on top — J) included in each of the six couples of seasons for which we checked for spatial variation in data. Each couple is represented by one black and one red ellipse with a gap of two months on each side. Exact dates were a) January to April and July to October, b) February to May and August to November, c) March to June and
September to December, d) April to July and October to January, e) May to August and November to February, f) June to September and December to March. These six separate groups of windows in time enabled us to check for geographic variation for each species throughout the whole year on a monthly resolution.
We estimated the EOO maps using 95% minimum convex polygons in R (R Development Core Team 2017), with a customised code (available in the Appendices). A minimum convex polygon consists of ‘the smallest polygon in which no internal angle exceeds 180 degrees and which contains all points of occurrence’ (IUCN Red List 2016). To produce a 95% minimum convex polygon, the algorithm iteratively removes the 5% of locations that have the greatest individual influence on the area of the minimum convex polygon (Hu & Yang 2013). We chose this mapping method for its simplicity to apply on a large multi-species dataset (Laver & Kelly 2008; Burgman &
Fox 2003), and because it is endorsed by the IUCN Red List as one of the measures that can be used in extinction risk assessments (IUCN Red List 2016). We envisaged that using minimum convex polygons for the mapping exercise here may allow future comparisons with (and potentially applicability to) the IUCN Red List data.
The Seasonality Index
To measure seasonal spatial variability, we created an index (Equation 1) for each couple of seasonal maps that is one minus the product of the percentage of overlap in the maps and the size ratio between the maps. The index ranges from 0 to 1 and values closer to 1 indicate less overlap and more difference in size, thus more spatial variation in each species’ seasonal maps which
could be an indication of patterns of cyclical migration or bias in data collection. Values closer to 0 indicate more overlap and less difference in size, which means less spatial variation. In total, we obtained 6 indexes for each species (one for each of the 6 couples of seasons). We considered only the largest index of each species for the rest of the analysis.
Equation 1:
Index = 1-(x*y)
Where x is the maximum intersection of the range polygons s1 in s2 or s2 in s1. The range polygons s1 and s2 are potential breeding and non-breeding ranges in each couple of seasons;
and y is the minimum size ratio between range polygons s1 and s2.
Threshold
Using sensitivity analysis, we defined that all species with an index above 0.4 (n=1188) presented
‘substantial’ spatial variation in their seasonal ranges (Figure 13), which could signify that the species is a to-and-fro or irruptive migrant.
Figure 13: Some range configurations detected with our method and data that could be linked with species to-and-fro and irruptive migrations. The movements of some of these migrants may be: (a) low-dispersive, where range polygons display a shift in space, which results in a varying size of overlap of the two polygons depending on the dimension of the shift; or (b) high-dispersive with varying amounts of overlap, where migrants display
contraction/expansion of their range, with or without a shift in their seasonal range centroid location.
To define the threshold, firstly we aimed at discovering the index value above which all species in a group of ‘well-known to-and-fro migrant species’ would be included. We assumed that if Birdlife International had mapped a species’ seasonal range, there was little doubt that the species migrates. To this end, our group of ‘well-known to-and-fro migrants’ was represented by all the species in our dataset with seasonal ranges in BirdLife International’s distribution maps (n= 67). In these data from Birdlife International, each species’ range is represented by one or more polygons, each polygon classified as either ‘resident’, ‘breeding’, ‘non-breeding’, ‘passage’ or ‘uncertain’. Our group of well-known migrants contained all species which exclusively had polygons characterised as ‘breeding’, ‘non-breeding’ and/or ‘passage’. Considering a threshold of 0.4 in our seasonality index, where species presenting an index equal or larger than 0.4 could be migrants, 96% of the well-known migrant species were positively detected as migrants in the eBird dataset. There were three species of well-known migrants with an index of less than 0.4: Long-billed curlew Numenius americanus (index=0.33), Straneck's tyrannulet Serpophaga griseicapilla (index=0.37) and Galapagos penguin Spheniscus mendiculus (index=0.26). A visual inspection of these species’
range maps revealed that much of their breeding ranges were spatially contained within the 95%
minimum convex polygon of their breeding ranges, even when in fact the breeding and non-breeding ranges, in theory, do not overlap at all in space (Figure 14). Minimum convex polygons are expected to produce overestimates of area from certain configurations of data, for example when the data is spread in a horseshoe or doughnut shape (Burgman & Fox 2003), which explains the high overlap of seasonal ranges in these cases. Excluding these three species, accuracy is 100% and errors of omission and commission increase as we move away from the 0.4 threshold (Figure 15).
Figure 14: Long-billed curlew Numenius americanus’, Straneck's tyrannulet Serpophaga griseicapilla’s and Galapagos penguin Spheniscus mendiculus’ extent of occurrence maps provided by BirdLife International of their breeding (dark grey) and non-breeding (light grey) ranges. The three species2 are well-known migrants and their non-breeding ranges spread around their breeding ranges in a horseshoe shape fashion. This configuration of their seasonal ranges jeopardised the measures of spatial and temporal variation of their ranges using our method.
2 Photo credits:
N. americanus: Ingrid Taylar, wikimedia.org [CC BY 2.0]; S. griseicapilla: Cláudio Dias Timm, wikimedia.org [CC BY-SA 2.0]; S.
mendiculus: putneymark, wikimedia.org [CC BY-SA 2.0].
Figure 15: Cumulative percentage of the 67 migrant species that we considered for the sensitivity analysis within different thresholds of the index to measure seasonality in species’ data. A threshold of 0.4 was used for the rest of the analysis because it included 96% of the well-known to-and-fro migratory species (n=64), while the remaining 3 species yielded a low index value because of the unusual spatial configurations of their seasonal distributions (Figure 14).