Survey methods
Occurrence data was gathered over a four year period with eight surveys, in both winter and spring. Random points within the smallest convex polygon around Isla Magdalena and Isla Santa Margarita were generated using ArcMap (v. 10.6.1, ESRI, inc.). The random points were used as centers of survey transects (Elzinga, Salzer, & Willoughby, 1998; Bonham, 2013). In addition to occurrence data, we recorded soil type data and other topographical information, and the latitude and longitude of occurrences was marked using a handheld GPS device (eTrex 30X, Garmin Ltd. Olathe KS). A different surveying technique was employed on Cabo San Lazaro due to the much smaller area of that land mass. At that site, a belt transect method was used, with individuals counted within a 300 m radius along a 3 km segment that covered the entire habitat ofCochemiea halei (Bonham). The only known peninsular population, with six individuals, (Gorelick, 2007), was included in all analyses.
Presence points were spatially thinned to a minimum separation of 1 km to reduce sampling bias (Tessarolo, Rangel, Araújo, & Hortal, 2014; Stolar & Neilsen, 2015). A point raster layer of presence/absence locations was generated with 1 km grid squares and multiple records per square were reduced to at most a single record. Point pattern analysis was performed using Ripley’s K statistic to measure the degree of spatial correlation of presence/absence records (Baddely, 2008).