Inferring ecological relationships from occupancy patterns for California Black Rails in the Sierra Nevada foothills

Occupancy models provide a useful tool for examining relationships between species' occurrences and environmental or ecological covariates when detection probability is less than one. This research is focused on the secretive and rare California Black Rail ( Laterallus jamaicensis coturniculus ) and its wetland habitats in a newly discovered part of its range in the Sierra Nevada foothills of California, USA. In order to examine the Black Rail's distribution, residency, density, relationship to a larger conspecific, the Virginia Rail ( Laterallus limicola ), and relationship to livestock grazing, three classes of occupancy models were utilized: single-species/single-season models, single season/two-species models, and multi-season/single-species models. In order to develop a wetland classification procedure for identifying potential Black Rail habitats, we evaluated the utility of combining topographic features with spectral and geometric features using high-resolution satellite imagery and a digital elevation model (DEM). The secretive California Black Rail has a disjunct and poorly understood distribution. After a new population was discovered in Yuba County in 1994, we conducted call playback surveys from 1994-2006 in the Sierra foothills and Sacramento Valley region to determine the distribution and residency of Black Rails, estimate densities, and obtain estimates of site occupancy and detection probability. We found Black Rails at 164 small, widely scattered marshes distributed along the lower western slopes of the Sierra Nevada foothills, from just northeast of Chico (Butte County) to Rocklin (Placer County). Marshes were surrounded by a matrix of unsuitable habitat, creating a patchy or metapopulation structure. We observed Black Rails nesting and present evidence that they are year-round residents. Assuming perfect detectability we estimated a lower-bound mean Black Rail density of 1.78 rails ha -1 , and assuming a detection probability of 0.5 we estimated a mean density of 3.55 rails ha -1 . The probability of detecting occupancy from a single call playback survey at a marsh was high (mean = 0.84), and the estimated proportion of marshes occupied (across all years) was 0.58. The proportion of sites occupied by Black Rails in the foothills remained relatively stable from 2002-2006 despite turnover from year to year of specific sites. Irrigation ditches were the primary water source at 75% of the marshes that had Black Rails. Approximately two-thirds of marshes with Black Rails were on private land. Black Rails are more widespread in the Sierra foothills than was previously known, and the foothills distribution appears to be discontinuous with populations in the San Francisco Bay-Delta Estuary. Occupancy surveys may be an improved method for monitoring population trends of this secretive marsh bird where habitat patches are highly fragmented. Two-species occupancy models that account for false absences provide a robust method for testing for evidence of competitive exclusion, but previous model parameterizations were inadequate for incorporating covariates. We present a new parameterization that is stable when covariates are included, the conditional two-species occupancy model, that can be used to examine alternative hypotheses for species' distribution patterns. This new model estimates the probability of occupancy for a subordinate species conditional upon the presence of a dominant species. It can also be used to test if the detection of either species differs when one or both species are present, and if detection of the subordinate species depends on the detection of the dominant species when both are present. We apply the model to test if the presence of the larger Virginia Rail affects probabilities of detection or occupancy of the smaller California Black Rail in small freshwater marshes that range in size from 0.013-13.99 ha. We hypothesized that Black Rail occupancy should be lower in small marshes when Virginia Rails are present than when they are absent, because resources are presumably more limited and interference competition should increase. We found that Black Rail detection probability was unaffected by the detection of Virginia Rails, while, surprisingly, Black and Virginia Rail occupancy were positively associated even in small marshes. The average probability of Black Rail occupancy was higher when Virginia Rails were present (0.74 p however, spring cover was not well predicted by RDM. Accurate, transferable and efficient mapping procedures are needed for wetland inventory, assessment and monitoring. Wetland mapping is typically carried out using two types of inputs: (1) spectral reflectance data from imagery and (2) topographic/hydrologic data derived from digital elevation models (DEMs). Hybrid approaches that integrate remotely-sensed imagery with topographic data have shown improved wetland mapping accuracy in several studies. Here we evaluate the performance of nine topographic features (aspect, downslope flow distance to streams, elevation, horizontal distance to sinks, horizontal distance to streams, plan curvature, profile curvature, slope and topographic wetness index) on freshwater wetland classification accuracy in the Sierra foothills of California. To evaluate object-based classification accuracy we test both within-image and between-image predictions using six different classification schemes (naive Bayes, the C4.5 decision tree classifier, k-nearest neighbors, boosted logistic regression, random forest, and a support vector machine classifier) in the classification software package Weka 3.6.2. Adding topographic features had mostly positive effects on classification accuracy for within-image tests, but mostly negative effects on accuracy for between-image tests. The topographic wetness index was the most beneficial topographic feature in both the within-image and between-image tests for distinguishing wetland objects from other green objects (irrigated pasture and woodland) and shadows. Our results suggest that there is a benefit to using a more complex index of topography than simple measures such as elevation for the goal of mapping small palustrine emergent wetlands, but this benefit, for the most part, has poor transferability when applied between image sections. Occupancy models provide a robust method for examining species-environment relationships when detection probability is imperfect. The Black Rail study system in the Sierra foothills provides a unique and valuable opportunity for examining the effects of interspecific competition and grazing on a threatened subspecies at a regional scale. Further development of wetland mapping procedures will allow for a more complete description of the distribution of this rare and enigmatic marsh bird.