The stakeholders consisted primarily of the Preserve Partners—a consortium of federal, state, and local agencies—in addition to non-profits such as TNC and Ducks Unlimited. The management scenarios were developed with reference to the Preserve Management Plan , created by the Preserve Partners through a 2-year planning process . The management plan gathered information from the public , the Preserve Partners, local municipalities, and other groups. We used a time-frame of ~30 years to 2050 . We considered each of the scenarios in isolation; for example, landscape-wide restoration does not account for development, nor urbanization for set-asides for wildlife. For parcels not affected by the management scenario, we assumed a static landscape with no change of land use occurring, as much of the land has remained pastoral for ~150 years before present. The objective of this scenario was to maximize restoration of agricultural lands to natural riparian habitat, focusing on areas of specific soil type and proximity to the river, based on an analysis in the Management Plan. In addition, Preserve goals, as set forth in the Management Plan, support maximizing the restoration of riparian habitat in the Cosumnes corridor. We applied six decision rules that relate the location of each parcel to existing landscape features. A parcel received a high likelihood of being restored if: the parcel was currently within the Cosumnes River Preserve lands; was managed for other conservation purposes ; was within a historical riparian corridor; was within 1km of standing water; was within 1km of grassland, shrubland, or wetland; or was within 1km of riparian forest.
The 1-km distance threshold was set to be inclusive of remnant riparian forest within the Preserve,flood table coupled with the assumption that areas within this threshold are practical targets for restoration. If a parcel fell within any one of these six categories, it was given a score of one. Scores were summed for each of the six rules, and a composite score was given to each parcel. Using the composite score, the upper quartile of all parcels with a score greater than one, weighted by area, were designated to be restored. We tested for multicollinearity using the variance inflation factor , which did not indicate substantial multicollinearity of the six decision rule variables . We filtered the final layer of restorable parcels to exclude existing urban areas. The restoration of these parcels was to grassland or riparian forest, which was assigned based on the potential natural vegetation . We did not consider areas defined by Kuchler as subtidal marsh within the study area for future restoration because of the current lack of feasibility; consequently, parcels remained under current land cover . The objective of this scenario was to represent a realistic growth outcome for the area projected to 2050. We assigned parcels as urban in 2050 based on whether they were currently urban or projected to become urban using the Preferred Blueprint Scenario for 2050 . SACOG generated the blueprint to help guide local government in growth and transportation planning through 2050 throughout the six-county region. This Preferred Blueprint Scenario promotes compact, mixed-use development and more transit choices as an alternative to low-density development . We filtered the final layer of potentially urbanized parcels by extracting current protected areas and areas of riparian forest, which are unlikely to be developed because they are within the 100-year floodplain. This provided the land cover data necessary to make projections in ecosystem services and disservices for 2050. The objective of this scenario was to maximize high-quality foraging habitat for the Swainson’s Hawk, and was developed based upon the literature and experience of the authors. A parcel received a high likelihood of being Swainson’s Hawk-friendly agriculture if it was designated as or within 1.25km of alfalfa, grain, pasture, or row crop . We selected all non-protected parcels of natural vegetation, such as grasslands, within 1.25km of existing fields under these four agricultural types to identify parcels available for conversion to agricultural types favorable to Swainson’s Hawk.
For practical reasons, certain agricultural types were not considered for conversion. For example, vineyards, given their high economic value , are unlikely to be converted; vineyard expansion remains a dominant trend for the region . We filtered the final layer of potentially enhanced parcels to exclude riparian forest and existing urban areas. Using the composite score, we converted the upper quartile of all parcels for these rules from existing land use to the four agriculture types more favorable for Swainson’s Hawk foraging. These types were randomly allocated in proportion to the number of parcels in which they currently occur: alfalfa 5%, grain 20%, pasture 45%, and row crop 30%. As with the restoration scenario, we tested for multicollinearity using the VIF, which indicated the model did not have substantial multicollinearity of the variables.We quantified the amount of carbon stored within three different land-cover types based on readily available data and literature: agricultural crops ; natural, non-forest vegetation types ; and forest types . We did not include soil carbon storage in this analysis, nor do we account for carbon storage associated with natural habitat within urban areas. We assumed that these steady-state estimates apply to all locations, and that changes in land cover would increase or decrease carbon storage to a new steady state.Above- and below-ground carbon storage for standing agricultural crops was based on a study by Kroodsma and Field . We divided the yield by the harvest index for each crop, and then multiplied the result by 0.45 as the proportion of biomass assumed to be carbon to estimate Mg C ha-1. For row crops not included in Kroodsma and Field , we estimated it using the National Agricultural Statistics Service yield data from the year 2000 and the average harvest index for all row crops . We estimated carbon storage for orchards by assuming the mid-point of the crop’s lifespan, multiplying this by wood accumulated , and again multiplying by 0.45 to provide Mg C ha-1.
For perennial crops not listed in Kroodsma and Field , we used the mid-point of the average lifespan and the average wood-accumulation estimates for crops within the same category as defined by the CDWR. For our broad categories of grains, orchards and row crops, we calculated an average value of Mg C ha-1 based on data available for each crop type within the category. We multiplied the area of each of the seven agricultural classes in the study area by the estimated carbon to provide total Mg C ha-1 . Carbon storage for non-forest natural vegetation types used estimates from the literature: pasture, grassland, and shrubland , and freshwater emergent wetlands . We divided forested vegetation types into three main types of riparian forest: valley oak , Fremont cottonwood , and willow . We improved the estimates of carbon storage associated with these types of riparian forest with plot data collected on the Cosumnes Preserve. Plot data included the measurement of all trees >10cm diameter at breast height within a 0.04-ha plot , applied allometric equations to calculate the amount of above-ground carbon , and summed these amounts to report a total Mg C ha-1 for riparian forests . These are in line with other estimates of live biomass from riparian studies in California . Using this variety of techniques,rolling benches we assigneda coarse estimate of total Mg C ha-1 for each parcel in the study area . Since this compilation of varied data includes a mix of both above- and below-ground carbon estimates for different classes , our estimates need to be considered conservative. We assessed the effect of different landscape management scenarios on the Swainson’s Hawk and also on a suite of 15 focal bird species. First, we used Boosted Regression Trees modeling techniques to fit the baseline land-cover data to presence and absence points of Swainson’s Hawk nest locations. We used known nest locations, identified using comprehensive field surveys of the area , to generate presence points . We generated absence points by randomly placing pseudo-absence points within the study area. We used 75% of the points to train the landscape suitability model, and 25% to test the predictive ability of those points.
We generated models by calculating the proportion of each land-use type contained within a 25-ha square that surrounded each presence and absence point. This threshold utilized research which found that 50% of Swainson’s Hawk foraging occurs within 25 to 86 ha of nest sites . We also tested model sensitivity using a 100-ha core area, and noted no significant changes n model results. Once we fitted the current land-cover type to the Swainson’s Hawk nest presence and absence data, we used the BRT to spatially project the probability of landscape suitability onto each of the three future scenarios. We assessed model performance using area under curve of the receiver operating characteristic curve scores . We converted model results to raster grids and assigned each parcel a landscape suitability score based on the average score of all grid cells contained within a parcel . Second, we assessed the effects of the three management scenarios relative to baseline on a suite of 15 other focal bird species identified as indicator species for natural habitats in the Central Valley . In contrast to the Swainson’s Hawk approach, we used existing suitability models developed for each of these focal species in the Central Valley, with suitability scores ranging from zero to one . We assigned suitability values to our baseline and three alternative scenario parcels using two steps. First, we estimated the average suitability for each bird species within each of our land-cover types by overlaying the spatial suitability surfaces onto our land-cover data and calculating the area weighted average suitability of each land-cover type for each bird species . Second, we assigned an area-weighted suitability value for each of the 15 species to each parcel in our baseline and future management scenarios, according to the parcel’s land-cover type. Based on these scores, we calculated the average suitability score for each landscape scenario across all 15 focal bird species, using a 5% increase or decrease as the threshold for meaningful change.We calculated nitrous oxide emissions for the agricultural land-use types in a manner consistent with International Panel on Climate Change Tier-1 guidelines . The key input parameter was nitrogen fertilizer use. We acquired estimates of nitrogen fertilizer application rates from a compilation of California data and summarized these by our seven agricultural types . For grain, orchard, pasture, and row crops, which contain multiple types of crops, we averaged emission rates of these individual crops to provide a single figure for the class. We used IPCC emissions factors to convert nitrogen fertilizer application to nitrous oxide emissions. This was 1% of nitrogen fertilizer applied for all crop groups except rice, for which we used an emissions factor of 0.3% . We excluded estimates for alfalfa since this crop rarely receives inorganic nitrogen fertilizer application and only accounts for a small proportion of the study area . For each parcel, we estimated the amount of nitrous oxide emissions per year under baseline and the three alternative management scenarios based on the land use type within each parcel. We calculated the amount of nitrate–nitrogen leaching for the agricultural land types based on the difference between nutrient inputs and nutrient losses. We compiled nutrient inputs from crop specific fertilization rates and based nutrient losses on the amount of nitrogen harvested in crops . We assumed atmospheric losses to be 10% of the fertilization rate, which is a conservative estimate developed to reflect the total N gaseous emissions . We assumed all surplus nitrogen was leached from soil into the groundwater in the form of nitrate , and for crops where the nitrogen harvested exceeds the nitrogen inputs, we assumed leaching loss was zero. As with emissions, we estimated the amount of nitrogen currently leached per year for each parcel and for the three alternative scenarios.Natural vegetation increased slightly in the restoration scenario from a baseline of 44% to 46% of the study area, and decreased in the urban and enhanced agriculture scenarios to 40% and 21%, respectively . Under baseline conditions, natural vegetation consisted primarily of grassland in the eastern portion of the study area , and riparian forest along the Cosumnes River accounts for 4%. Cover of riparian forest increased to 12% under the restoration scenario .