We used NLCD land cover to classify natural habitat types, which were not identified in the SCAG layer. Lands in classes which were essentially open space with substantial human activity and which had less than 10% impervious surface were classified as ‘disturbed’. When land-cover layers from the different sources were inconsistent, we verified classifications with ground visits or visual inspection of air photos .We used motion-activated digital cameras at 38 sites to detect carnivore species from April 2007 to June 2008, resulting in 1,130 trap nights. Cameras were placed in and around 6 orchards and 2 continuous wild lands , with distance to nearest camera between 30– 900 meters . At all sites, cameras were placed along similar-sized dirt roads, near signs of carnivore activity or at trail junctions when possible. We placed scent lure in front of the camera to encourage animals to approach the camera and to stop long enough to be photographed. For each carnivore species at each camera site, we tallied the number of nights in which the species was detected at least once. We considered each 24-hour trap night to begin at 6:30 am, and cameras were active continuously between 12 and 76 nights at a particular site.We began the modeling process by selecting the best detection model for each species while holding occupancy rate constant, as in Negro˜es et al. and Duggan et al.. We expected that season and land cover at the camera site could affect detection rate so we included both as covariates in detection models. Detection covariates, as well as predictors of occupancy described below, were standardized by z-score as described in Donovan and Hines. Next, to determine if spatial clustering affected occupancy, and at what scale, we compared models including site , meta-site , or county as predictors of occupancy while including any detection variables selected in the previous step. We then included the covariate from the top-ranked model of spatial scale as a predictor in the candidate model set for occupancy of that species. Finally, for each species we modeled occupancy while including detection covariates from the top-ranked detection model for that species. Potential predictors of occupancy included land cover at the camera site,vertical farming distance from each camera to the perimeter of continuous wild lands , and season .
We also evaluated the degree to which area of orchards and other landscape variables in the neighborhood of a camera influenced carnivore occupancy. To do so, we used the land-cover map to quantify the extent of orchards and covariates within a 1,935 m-diameter circle centered on each camera, approximately the average size of a bobcat home range in this region and intermediate between range sizes of foxes and coyotes. We had 38 sites, and therefore examined only single- and double-factor occupancy models to avoid over parameterization.We report results for the average model, but also include summaries of the top ranked model and all models within 2 AICc points of this model, indicating substantial support. To compare the selection support for each predictor variable, we also calculated variable importance weights, which are the sum of the model weights of all models that contain a given variable. Averaged models include only models within 2 AICc points of the best model. Variable importance rates are assessed across all models and therefore each variable has equal representation.Cameras were active for a total of 667 trap nights in orchards, 201 in natural vegetation near orchards, and 262 in wild lands. We detected 8 of the 11 native carnivore species in the study region. Seven native species were detected in orchards: coyote , striped skunk , bobcat , gray fox , mountain lion , black bear , and raccoon . Eight native species were detected in natural vegetation: coyote , bobcat , mountain lion , gray fox , raccoon , badger , black bear , and striped skunk . The 3 native carnivore species not detected included ring tail, spotted skunk, and long-tailed weasel.On average, the number of native species detected per site differed among land-cover classes , but differences between individual classes were not statistically significant . The number of native species detected was greatest in orchards , intermediate in sites with natural vegetation adjacent to orchards , and lowest in wild land sites . The top-ranked detection rate models for coyote and gray fox included land cover at the camera station location , with higher detections in avocado orchards or near avocado orchard relative to wild lands. Season was also included in the top-ranked models; for fox, the direction of the effect could not be distinguished from 0 , while for coyote, detection rate was lower in the dry season than in the wet season . These detection covariates were included in all subsequent models. For bobcat, the intercept-only model was the top model, so subsequent bobcat models did not include detection covariates.
For all three species, the intercept-only occupancy model was the top-ranked model for spatial variation, thus we did not include spatial variables as predictors of occupancy in our final set of candidate models. Avocado orchards, either at the camera site or in the neighborhood of the camera, were included in at least one competitive occupancy model for all three carnivore species . The area of avocado orchard in the neighborhood of a camera was the most important predictor of bobcat occupancy and was included in all top four models for bobcat occupancy . The area of avocado orchards in the neighborhood of a camera was the third most important predictor for gray fox occupancy . For coyote, the area of orchard in the neighborhood had a weak negative effect , but both avocado orchard and ‘near orchard’ at the camera site had a positive effect . Land cover at the camera site was not included in any competitive bobcat or gray fox occupancy models. Distance to continuous wild land was the most important variable for predicting gray fox occupancy and third most important variable for bobcats , with occupancy increasing closer to or within wild land habitat. Distance to continuous wild land was not, however, included in any competitive coyote models. The area of disturbed land in the neighborhood of the camera was included in competitive models for both coyote and gray fox occupancy , but large standard error values for fox occupancy suggested a weak influence. Disturbed land was not included in models for bobcat occupancy . Woodland, shrub, and grassland/herbaceous vegetation in the neighborhood of a camera had a positive effect on occupancy in all models for all species, except that woodland had a negative effect on gray fox occupancy.Carnivores were detected with surprising frequency in avocado orchards. We detected most carnivore species native to coastal southern California in avocado orchards, and these orchards were used frequently by bobcats, coyotes and gray foxes. Further, we detected more carnivore species in orchards than in wild land sites. Although orchards are often adjacent to wild lands, the presence of carnivores in orchards does not appear to be simply an artifact of landscape context. If this were the case, we would expect to find more carnivores in wild lands than in orchards, which we did not. We would also expect to find that distance to continuous wild land was a more consistently important predictor in our models; although it was the strongest predictor of occupancy for gray fox, it was present in only one competitive model for bobcat occupancy and no competitive models for coyote.The food subsidy value of avocados may explain why omnivores such as bears, coyotes, and raccoons were present in orchards. Indeed, remote cameras have recorded these species eating avocados in southern California , but why obligate carnivores like mountain lions and bobcats would be present in orchards is less clear. Orchards may provide good cover for carnivores; many of these species are habitat generalists, and orchards often replace oak woodlands with structurally similar vegetation.
In our study, we did not find an effect of wet versus dry season on occupancy, as might be expected if carnivores were attracted by water sources. However, irrigation lines, combined with abundant avocados, might simulate year-round wet-season conditions for small mammals, perhaps leading to bottom-up effects in these agricultural systems. Future research could assess whether orchards are providing more food and water for small mammals than native vegetation,vertical garden hydroponic and whether a relative increase in prey might help explain the use of these lands by carnivores. Finally, further study could evaluate whether the presence of infrequently-used dirt roads in orchards might appeal to animals moving across densely vegetated landscapes.There is growing interest in managing for movement of wild animals through agricultural areas. Knowing the value of different land-cover types for habitat or movement can inform conservation decisions regarding which lands should be purchased or put under easements, or which areas are most suitable for the placement of highway crossings. Avocado orchards appear to serve as both foraging and movement habitat for most carnivore species in California, and conservation easements or other incentives to keep land in orchards could offer a cost-effective conservation strategy. Such alternative conservation strategies are particularly important when considering agriculture in Mediterranean-type ecosystems, which are highly threatened.Research is increasingly showing that agricultural crop yields will be susceptible to future changes in temperature, precipitation, length of growing seasons, and carbon dioxide concentrations. While future climate is uncertain, the potential for important effects on major agricultural crop producers such as the U.S. is clear. The quantity and composition of U.S. agricultural crop production, where crops are grown, trade, and economic value of U.S. crop production could be affected. An important challenge in understanding these implications is that agricultural products are traded across the globe. The U.S. is both a major agricultural importer and exporter of agricultural crops, meaning that U.S. agriculture may be affected not only by future climate in the U.S., but also future climate outside of the U.S. via international trade. These international linkages raise questions about the relative importance of the direct and indirect effects on U.S. agriculture; that is, is the potential for changes in temperature and precipitation in the U.S. more or less important than the potential for changes outside of the U.S. to U.S. agricultural crop producers?
Many studies have looked at the global response of agricultural systems to changes in climate, as well as the relative importance of where the agricultural impacts occur. In recent years, attention has begun to turn toward the implications of international trade on the U.S. agricultural sector in particular, finding that international trade effects for the U.S. agricultural sector are comparable in importance to direct, domestic impact. Zhang et al found that considering full, global climate impacts causes significant changes in the projections of U.S. production and exports of crops. Costinot et al and later Gouel and Laborde examined the aggregate impact of widespread local yield changes on global agricultural markets, again finding that the full picture is necessary to understand the future, as comparative advantages among regions shift with particular attention paid to domestic adjustments in land allocation. Finally, Baker et al. extended these findings to a new modeling application, determining that not only does considering global markets cause significant changes in the projections of U.S. agricultural crop production relative to a U.S. only focused study, but also that freer trade may help buffer local productivity shocks. These results suggest the potential for meaningful effects on the financial revenue of the U.S, agricultural sector through changes in both domestic production and international prices. In this paper, we systematically explore the implications of changes to domestic and international temperature and precipitation for U.S. agricultural crop production using a regionally-resolved, global scale model of energy, land, economic, and climate systems: the Global Change Assessment Model . This paper adds to the literature by providing a complementary quantification and replication of the production effects explored in Baker et al. and other modeling efforts with a different model. The analysis in this study is conducted using a modified version of GCAM 5.2. GCAM is a global model that couples representations of the energy system, the economy, agriculture and land-use, water, and the global climate in a single computational platform.