Land-Use Change and Future Water Demand in California’s Central Coast

,


Introduction
Water availability and human land use are inextricably tied (Stonestrom et al., 2009).In water limited regions, available freshwater supplies can often dictate land use intensity.However water withdrawals and diversions to support land uses, especially for irrigated agriculture, directly impact freshwater supplies (Foley, 2005).Adding to the complexity are the associated feedbacks between land use, climate, and water supplies.Human land use has been attributed to widespread increases in average global temperatures, contributing to global warming (Diffenbaugh et al., 2015;Ellis et al., 2010;Williams et al., 2015), losses in species diversity, (Fischer & Lindenmayer, 2007;Hansen & Rotella, 2002;Januchowski-Hartley et al., 2016;Klausmeyer & Shaw, 2009), changes in water quality (Charbonneau & Kondolf, 1993;Los Huertos et al., 2001;Scanlon et al., 2005), and groundwater depletion (Konikow & Kendy, 2005).Understanding potential future land-use related water demand in a region serves as a first step in assessing prospective outcomes and associated mitigation strategies to address potential vulnerabilities.
California exemplifies these issues with water arguably the state's most contentious resource.
The state boasts one of the most productive agricultural regions in the world, worth ~$50 billion (California Department of Food and Agriculture, 2018), which consumes between 60-80% of all water supplies, while residential and industrial consumption is roughly 17% (Brandt et al., 2015;Cooley, 2014;Maupin et al., 2014).Surface water is over allocated, estimated at 400 billion cubic meters, 5 times the average annual runoff (Grantham & Viers, 2014).The state's Mediterranean climate is highly variable, characterized by long-term droughts and atmospheric river flooding events (Dettinger et al., 2011), contributing to inter-annual water supply uncertainty.Moreover, water demand is highest in the dry, summer months.A statewide extreme drought from 2012-2016 led to water shortages, increased reliance on groundwater pumping, and subsequent well drying (Perrone & Jasechko, 2017), and also contributed to saltwater intrusion in some groundwater basins (Barlow & Reichard, 2010;Hanson, 2003;White & Kaplan, 2017).
Efforts to plan for water resource sustainability are more challenging now than ever, as these drought and flood events increase in frequency and intensity due to a changing climate (Berg & Hall, 2015;Diffenbaugh et al., 2015;Swain, 2015).While the state has long experienced periodic droughts, many climate projections show increased drought occurrence in coming decades (AghaKouchak et al., 2015;Ault et al., 2014;Diffenbaugh et al., 2015;Famiglietti, 2014;Konikow & Kendy, 2005;Trenberth et al., 2014).Reduced surface water during drought often leads to increased groundwater pumping in the state (Famiglietti et al., 2011;McEvoy et al., 2017;Ojha et al., 2018).Recent work also projects a 25-100% increase in extreme wet/dry events by century's end, despite only modest changes in mean precipitation (Swain et al., 2018).
Such extreme events, combined with increased evaporative water demand due to climate warming, as well as future population growth and agricultural expansion, will likely contribute to even greater water demand, posing additional challenges to an already unsustainable situation.
This may lead to a pivotal juncture where water demand exceeds available supply.
Oversight of California's groundwater has historically been limited.While surface water withdrawals require permits, groundwater pumping has gone largely unregulated and is managed locally (Leahy, 2016).Several legislative attempts have been made to incentivize groundwater management and to better integrate land use in water supply planning.In 1992, AB 3030 passed, and was modified in 2002 by SB 1938, providing procedures and incentives for local agencies to voluntarily develop groundwater management plans (Costa, 1992;Machado, 2002).In 1995, Senate Bill (SB) 901 required that local governments conduct water supply assessments during the environmental reviews for large projects (above 500 housing units) (Costa & Setencich, 1995).In 2001, Senate Bills 610 and 221 required local land use authorities to demonstrate longterm water supply availability before approving new, large development projects (Costa, 2001;Kuehl et al., 2001).Despite these restrictions, none of these laws regulated groundwater pumping.By 2014, rapidly falling groundwater tables combined with ongoing extreme drought led the state to pass the Sustainable Groundwater Management Act (SGMA; AB 1739, SB 1168, and SB 1319) (Dickinson, 2014;Pavley, 2014aPavley, , 2014b)).Passage of SGMA marked the first time local agencies were required to regulate and sustainably manage groundwater resources of critically over-drafted groundwater basins.The implementation of SGMA is ongoing, with local agencies actively designing their groundwater sustainability plans.However, many of these agencies lack the ability to quantify sustainable groundwater yield driven by future land use related water demand.
California's Central Coast is an ideal system for examining the linkages between land use change and land use driven water demand over time and exploring the long-term impacts of water laws and policies on this process, as well as impacts on groundwater supplies, and resource and community sustainability.The region has major agricultural and residential areas that are entirely reliant on local groundwater.There is limited imported surface water, primarily in San Benito and Santa Barbara counties and groundwater overdraft (extraction exceeding recharge) occurs in an estimated 40% of basins in the region (Martin, 2014).Many of the coastal aquifers have seawater intrusion, exacerbated by the recent droughts, rendering local groundwater unsuitable for drinking or irrigation (Barlow & Reichard, 2010;Hanson, 2003;White & Kaplan, 2017).
Many of its valley floors overly groundwater basins and support extensive agriculture, while the vast majority is largely undeveloped natural land, creating the potential for substantial new development.It is home to some of the wealthiest and poorest communities in the state, including several disadvantaged communities (annual median household incomes <80% of statewide MHI; California State Legislature, 2002).The city of Salinas is currently the largest city at 156,259 people (U.S. Census Bureau, 2018).By 2060, the Central Coast is projected to add nearly 300,000 more people to its population (State of California, Department of Finance, 2018), likely increasing water demand.Water supplies may not be able to keep pace, which could exacerbate water insecurity in already vulnerable communities and potentially spark social conflict.
To assess the trajectory of land use driven water demand for California's Central Coast and explore whether the 1992 -2001 water laws and policies were correlated with the pattern of demand for the region, we ran two scenarios based on historical, empirical datasets of land use changes sampled.The first was a business-as-usual (BAU) scenario fit to land use change rates from the entire historic period, 1992-2016, while the second, recent-modern (RM) scenario only sampled from 2002-2016 rates (i.e., after the second set of laws were put in place in 2001).We simulated projected land use change and associated water demand for the years 2001-2100 at 270-m across 10 Monte Carlo simulations across these two scenarios.Our model was based on the Land Use and Carbon Scenario Simulator (LUCAS) (Sleeter et al., 2015(Sleeter et al., , 2017(Sleeter et al., , 2019;;Wilson et al., 2014Wilson et al., , 2015Wilson et al., , 2016Wilson et al., , 2017)), a stochastic, spatially-explicit state-and-transition simulation model.Spatial patterning of land use change was parameterized using local zoning datasets, identifying where land change would and would not occur giving current zoning ordinances and local mandates.Our goal was to understand the region's unique potential water demand, assisting local water resource and land managers in understanding the impacts of past policies to better identify and mitigate for possible future vulnerabilities as they continue to develop and revise new groundwater sustainability plans for SGMA.While SGMA is too new to definitively determine its impact on future water demand, viewing an unregulated future with and without existing policy provides an important baseline for more targeted mitigation planning.

Materials and Methods
The LUCAS state-and-transition simulation model (STSM (Sleeter et al., 2017(Sleeter et al., , 2019;;Wilson et al., 2016Wilson et al., , 2017) ) was developed and modified for our study region.The STSM divides the landscape up into spatially discrete simulation cells, each with assigned state classes and transition types.Each state class has pre-defined transition type pathways allowing or preventing cells to move between different state classes over time.What follows is a description of the model parameterization steps for the Central Coast region of California.For more comprehensive information on STSMs, see Daniel et al. (Daniel et al., 2016) We held three stakeholder meetings with individuals from regional municipal governments, water agencies, and community groups while developing our models.Meetings were held at the start of model development, the midpoint, and when presenting a draft version of the final model results.Stakeholders provided information on local spatial planning datasets that were assimilated into the models (see section 2.5) as well as interpretation of results in the context of local concerns about water sustainability and land use.

State Variables and Scale
The current study area encompasses 28,534 km 2 of the 5-county region in California's Central Coast (Figure 1a), covering Santa Cruz, San Benito, Monterey, San Luis Obispo, and Santa Barbara Counties.The region was divided into 270-m x 270-m simulation cells (391,421 total cells).Each cell was also assigned an initial LULC state class (Figure 1b) and three additional spatial identifiers including its 1) county, 2) groundwater sub-basin (Figure 1c; n = 61) (California Department of Water Resources, 2018) and 3) water service agency(s) (Figure 1d; n = 107), described below.Scenario simulations were initiated in 2001 and run through the year 2100.The model tracks changes in state class, age, time-since-transition, and state attributes (i.e., water demand).For each scenario simulation we ran 10 Monte Carlo iterations to capture model variability and uncertainty in our projections.
We utilized the National Land Cover Dataset 2001 (NLCD01; Homer et al., 2007)  The NLCD01 does not contain a perennial orchard and vineyard class.We used a 2001 perennial cropland cover map (Sleeter et al., 2019) which generated orchard and vineyard cover using a gradient boosting machine algorithm framework.Any NLCD01 pixel classified as agriculture which overlapped the 2001 perennial cover estimate was classified as perennial cropland.1 and 2, respectively.
The water agencies map (Figure 1d) was created by combining the Groundwater Sustainability Agency (GSA) Service Area dataset (California Department of Water Resources, 2019b) and the Water Districts dataset (California Department of Water Resources, 2019c).Because polygon boundaries did not line up precisely between the two shapefiles, polygons were manually edited to remove small slivers or gaps.Multiple agencies can also have overlapping jurisdictions (e.g., local city water systems and basin-wide GSAs), so each polygon in the final dataset was assigned 0-2 GSAs and 0-2 water districts each.If GSAs were formed from pre-existing water districts with the same boundaries, we included them only as GSAs.Four county-wide water districts were not included as counties are already represented in the LUCAS model.Lastly, water districts servicing <20 km 2 were removed, unless they were the only agency servicing that area.
If so, they were included and labeled "other small water district."This resulted in 107 unique jurisdictional combinations covering 29 GSAs and 40 water districts as well as "other small water district."

Model Formulation
The LUCAS model was formulated to simulate changes in state class variables for pathways associated with urbanization, agricultural expansion and contraction, and agricultural change (i.e.intensification associated with conversions of annual to perennial cropland).

Land Change Transitions Targets
Data from the Farmland Mapping and Monitoring Program (FMMP) (California Department of Conservation, 2017) dataset was used to supply LULC transition targets for agricultural expansion, agricultural contraction, and urbanization.The FMMP gathers bi-annual land change data using aerial photography and human interpretation.We updated the existing historical land change record (1992-2012) from Wilson et al. (2016) with newly available data, extending the record to span 24 years , from which future scenarios could be sampled.
To calculate the rangeland to annual cropland transition targets, we subtracted the rangeland to perennial transition target from the overall agricultural expansion targets from FMMP.Where more rangeland to perennial occurred than was reported as agricultural expansion, it was assumed that 0 km 2 of rangeland was converted into annual cropland.We recognize this approach introduces some data loss, however lacking wall-to-wall spatial and "from classto class" conversion information at higher temporal resolution, it is the most defensible approach to capture the large scale, notable shifts of natural lands into perennial production, a trend uncommon for annual cropland in this region.

Perennial Transition Probabilities
Conversions out of the perennial cropland class are also challenging to quantify.Perennial crops are expensive to plant, cannot be fallowed, and take several years post-planting to reach maturation (Johnson & Cody, 2015).The average lifespan of vineyards and orchards in California is 25 years (Kroodsma & Field, 2006), after which productivity often declines.In order to capture this lifespan, we extracted age values for our 2001 perennial cropland from an age class map available from Sleeter et al. (2019).Since the LUCAS model can track pixel age and time since transition, we set the following model rules: 1) a perennial pixel must reach a minimum age of 20 years before it eligible for removal or conversion, in any model year or iteration, 2) the annual transition probability for orchard removal was sampled from a cumulative probability of 0.95 for ages 20 and 45, and 3) after removal pixel age is reset to 1 and the cell is free to be converted into new development, agricultural contraction, or annual cropland (with annual probability set at 0.05) .If the cell does not convert in this age reset year, the model assumes it is replanted as perennial.Any perennial crop over 20 years in age has a 0.05 probability of transitioning back to annual cropland.

Adjacency & Spatial Multipliers
For each potential LULC transition, adjacency multipliers were applied where the relative probability of any transition increased linearly with the number of existing, neighboring "from class" cells within a 405-m x 405-m moving window.A cell would be eligible to transition if it contained at least one neighbor of the destination class (or transitioning "to class") within a 405m radius of the cell to be transitioned.The more neighbors of the "to class" increases the likelihood of transition which was linearly scaled between 0-1 based on the number of "to class" neighbors present.This parameter was updated every 5 timesteps for every possible LULC transition pathway.
We developed region-specific LULC transition spatial multipliers for the each LULC transitions: 1) urbanization 2) agricultural expansion and 3) agricultural change.Spatial multipliers are raster-based, probabilistic surfaces that either increase or diminish the likelihood of the specified LULC transition type.A probability of 1 ensures a transition will occur in that specified raster space if a transition target or multiplier is supplied, whereas a probability of 0 will prohibit the given transition from occurring in a cell.What follows is a discussion of the datasets used in the development of the LULC transition spatial multipliers.
Overall, we used national and state level land protection data from PADUS (U.S. Geological Survey, 2016) to prohibit any land change on protected lands and land owned by the Department of Defense.In addition, we incorporated available county-level land use zoning data to improve the regional accuracy of projected land change.This information was used to identify areas where LULC conversions are not currently allowed or where future development is already planned and zoned for.Land use zoning has been shown to be a strong predictor of urban growth and more accurately represents land change (Onsted & Chowdhury, 2014).For land change modeling, inclusion of spatial planning information generates better informed analyses (Dieleman & Wegener, 2004;Hersperger et al., 2018;Poelmans & Van Rompaey, 2010).Such an approach has been used by land change modelers to test alternative zoning scenarios (Geneletti, 2013) and as factors in LULC transition decision rules (Abdolrassoul & Clarke, 2012).We acknowledge that zoning data can and will change over time and land area can be rezoned with new designations.However, many zoning designations are likely to persist into the future, including open space and resource conservation areas.Alternatively, planned development areas are not likely to remain undeveloped for decades.Table 1 shows the additional zoning datasets used in the development of the spatial multipliers and their unique zoning designations.Zoning categories listed as No Conversion in Table 1 were applied as 0 values in all LULC spatial multiplier probability surfaces.We next describe each spatial multiplier in detail.

Urbanization
Additional constraints on the placement of new developed lands were derived from U.S. Census Bureau (U.S. Census Bureau, 2015) data and county-level land use zoning information (Table 1).
For conversions into new developed lands, we used the Urban Areas in 2011 dataset (U.S. Census Bureau, 2015), with areas designated as core urban areas (population > 50,000) assigned a probability of 1 for urbanization transitions, while secondary urban areas or clusters (population 2,500 < > 50,000) were assigned a probability of 0.5.All remaining areas not classified as 0 were given a 0.25 probability of conversion.See Table 1

Agricultural Expansion
Areas designated as protected in the urbanization multiplier were also considered unavailable for transitions into new agricultural lands.For county-level zoning datasets, this included open space, public recreation facilities, parks, protected lands, preserves, and more.See Table 1 "No Conversion" category for all areas prohibited from conversion into agricultural land uses for more detail.Agricultural expansion transitions into new perennial croplands were supplied the spatial multipliers described in Section 2.5.3.).We combined this with parcel-level orchard and vineyard data, aggregating avocado groves, citrus groves, orchards, and vineyards into a single perennial class with a probability of 1 for conversion into perennial cropland during this timeframe (Corelogic, 2018a).All other pixels were set with a probability of 0 to force new perennial crops into known locations.

Conversions to
In 2019 or the first projection year (i.e.year for which we do not know where new perennial crops occurred), we developed a "To Perennial 2019" multiplier, based on the 2018 multiplier to include probabilities of 0 for the "No Conversion" regions identified in Table 1, and 1's for the known historical locations.In addition, all other pixels classified as annual cropland or rangeland in 2001 were assigned a probability of conversion into perennial cropland.We calculated these probabilities of perennial conversion for each county based on the proportion of historical conversion from each class, based conversion rates defined in Section 2.3.

Water Demand
In

Projected Land Use and Land Cover Change
General LULC change trajectories were similar between scenarios but the overall magnitude of change was markedly different (Figure 2).In both scenarios rangelands and annual cropland declined, being outcompeted by development and perennial cropland expansion through 2100.
The declines were dramatic with BAU annual cropland declines averaging 80.0% (1,029 km 2 ) across Monte Carlo simulations, while the RM lost 81.4% (1,046 km 2 ).The BAU projected greater increases in developed land, yet lower losses of rangeland overall.In comparison, the RM scenario projected lower rates of development and greater increases in perennial cropland.
Perennial expansion in the region continued its robust historic trend, with planting of these specialty crops nearly doubling in the BAU and nearly tripling in the RM scenario.On average, the BAU was projected to gain 710 km 2 of new perennial cropland by 2100 with the RM scenario gaining 1,084 km 2 (Figure 2).Overall cropland totals-the sum of both annual and perennial cropland-increased slightly (37.4 km 2 ) in the RM scenario but declined an average 19.3% in the BAU (Figure 2).Developed lands increased in both scenarios across simulations but were approximately 11.7% higher in the BAU (843.3 km 2 ) than in the RM (666.6 km 2 ) (Figure 2).At the county scale, the greatest declines in annual cropland were projected in Monterey and Santa Barbara counties (Figure 3).The greatest increases in both developed and perennial cropland occurred in Monterey and San Luis Obispo counties, predominantly at the expense of rangeland (Figure 3) which declined between 181-186 km 2 (BAU-RM) and 365-479 km 2 (BAU-RM) respectively.In Monterey County, developed land increased between 21.6% (RM) and

Projected Future Water Demand
From 2001 to 2100, overall land-use related water demand was projected to increase between 222.7 and 310.6 million cubic meters (Mm 3 ) in the BAU and RM scenarios, respectively (Figure 5).In 2001, the Central Coast water demand estimate was approximately 1.3 billion cubic meters (Bm 3 ) with a projected to rise between 1.5 -1.6 Bm 3 on average across Monte Carlo simulations and scenarios by 2100 (Figure 5).This represents a 16.4% to 22.8% increase in water demand by the end of this century.Continuing trends in perennial cropland expansion led to a projected 222.7 Mm 3 increase in water demand in the BAU (Figure 6).This increase is small in comparison to the near tripling of perennial water demand in the RM scenario over 2001 use levels, rising by an estimated 359.2 Mm 3 , concentrated primarily in Monterey, San Luis Obispo, and Santa Barbara counties (Figure 6).Water demand from developed land uses was projected to increase 290.4 Mm 3 (53.8%) in the BAU and 230.8 Mm 3 (42.7%) in the RM scenario.The only demand declines projected were for annual cropland cover, with dramatic projected decreases from between 339.3 Mm 3 (77.9%) in the BAU and 344.8 Mm 3 (79.2%) in the RM in all counties (Figure 6).Opposite demand increase trends are seen between the BAU and RM scenarios, as the BAU shows increased demand higher for development than for perennial crops, whereas the RM shows higher perennial demand and lower demand by developed land uses.

Potential changes in groundwater basin overdraft
Projections of future land-use related water demand showed some groundwater sub-basins experiencing much greater increases than others.Figure 7 shows the percent change in total water demand per sub-basin, calculated as (Demand -Demand 2001 ) / (Demand 2001 + 10).Table 2 summarizes these results for each groundwater sustainability agency (GSA) and Table 3 summarizes them for other Non-GSA water districts.
Across both scenarios, increased water demand by 2100 was greatest in San Luis Obispo County (Figure 7).This is largely due to perennial agriculture replacing rangeland in many areas, creating unprecedented (percent increases >1000%) new perennial cropland water demand in Carrizo Plain basin and other small basins in the area, and roughly doubling total water demand in the Paso Robles area.In general, increasing urban water demand was uniformly spread across the study area, with median increases of ~50% per sub-basin (range 0-215%).In the major subbasins around Monterey Bay, many of which are already critically overdrafted (Figure 7b), total water demand increased only slightly.An exception was the critically overdrafted "180/400foot" sub-basin of the Salinas Valley, which underlies part of the disadvantaged city of Salinas and experienced a decrease in water demand of -11% in both scenarios.This restrained growth or even reduction in total water demand was due to urban expansion into previous annual agriculture resulting in a net loss of water.The greatest decreases in total water demand was in San Benito county.This was particularly notable in the RM scenario, where dramatically declining annual agriculture coupled with modest increases in urban water demand, led to an overall decreasing water demand in most sub-basins (median decrease of -8% in both scenarios).
Increasing water demand was projected in basins where encroachment of water-dependent human land uses occurred in previously open rangeland (Figure 1b, Figure 7).development and a historic drought, the RM scenario showed a 22.8% increased water demand overall, much higher than the 16.3% increase projected in the BAU.It is important to note that despite an historically unprecedented drought, perennial cropland expansion was projected to nearly double (BAU) and triple (RM), which may be cause for concern in a predominantly groundwater dependent region with already strained water supplies.
These same trends in agriculture intensification have been occurring statewide for decades.
Between 1960 and 2009, while the amount of harvested acreage in California declined by more than a half million acres, the proportion of fruit and nut crops (i.e., not field crops, vegetable, or melons) more than doubled from 14% to 33% of all acres harvested (Johnson & Cody, 2015).
Between 2004 and 2013 alone, statewide harvested acres for almonds, pistachios, grapes, cherries, berries, and olives nearly doubled as well (Johnson & Cody, 2015).Cropland reports for the Central Coast show annual field and row crops dominating the landscape, however, grape acreage between 2002 and 2017 expanded by nearly 25,000 acres (~100 km 2 ) (United States Department of Agriculture, 2019).
Neither the perennial or urban expansion trends are likely to persist indefinitely, particularly given new water limitations under SGMA.Shifts in future development patterns due to other local economic factors, changing dietary preferences, and a warming climate are likely to further deviate future rates from simply continuing historic trajectories.Specialty perennial crops could slow their expansion, as high value annual crops retain their value and market demand.Despite these limitations, these scenarios projections do provide an understanding of the challenges facing the region if current trends persist, providing a baseline from which additional mitigation scenarios can be developed, to explore alternative potential futures.

Land use and water use sustainability implications
Many orchard and vineyard crops have higher water demand than their annual row crop relatives, and most perennial crops require year-round watering.Our estimates show perennial cropland water demand is generally higher than annual cropland water demand in all but Monterey County where it is slightly lower, possibly due to cooler temperatures in the region and lower evapotranspiration loss.Monterey County also relies almost solely (95%) on groundwater (County of Monterey, 2019) and much of the agriculture occurs in the Salinas River Valley, a region prone to saltwater intrusion and significant water limitations.Given limited water supplies, regional growers have had to increasingly rely on advanced technology for watering vineyards, such as pressure chambers to detect water needs through leaf moisture, soil moisture probes, and groundwater moisture meters (Joseph, 2015) as well as water recycling (Shea, 2019).
Implementation of the Sustainable Groundwater Management Act could also exacerbate the situation, creating even greater limitations on groundwater pumping for perennial growers.
Given the 20-30 year lifespan of most of specialty perennial crops, their resilience to a changing climate and shifting water availability is also limited (Lobell & Field, 2011).Central Coast specialty crops show high sensitivity to changing temperature under future climate projections (Kerr et al., 2018).Specifically, wine grapes, strawberries, and lettuce-dominant crops in the Central Coast-had higher relative magnitude of negative impacts from increased temperatures of the top 14 value-ranked specialty crops in the state (Kerr et al., 2018).Yield declines have also been predicted with warmer winters and hotter summers (Lobell & Field, 2011).However, agricultural intensification also has many benefits.It often leads to 1) a higher investment and return per acre, 2) the creation of more jobs and demand for related support industry and housing, 3) the creation of more land use conflicts at the agriculture/urban interface, 4) technological innovation, and 5) improvements in irrigation efficiency (County of San Luis Obispo, 2010).These competing factors could influence a market-driven demand for improved water use efficiency.
New developed lands often generate additional water demand, potentially creating increased competition over ever-limited water resources.Well-drying and self-reported water supply shortages were already reported during the 2011-2016 drought and through 2019, and were highest in San Luis Obispo, with 201 reports submitted since 2014 (State of California, 2019).
By all accounts this represents only a small fraction of the total number households which likely experienced shortages, as vast under-reporting is suspected given limited outreach (State of California, 2019).By contrast, where urban growth was projected to spread into existing cropland, such transitions were demand-neutral and sometimes even led to reduced overall water demand as seen in areas around the Monterey Bay and San Benito County.Unfortunately, such growth patterns conflict with the conservation of prime agricultural lands, a major goal of regional and state land management (California Department of Food and Agriculture, 2015) and also reported by stakeholders.Future development patterns over time may include urban redevelopment and infill with higher density which would better preserve existing farmland.
New upland regions in non-prime farmland could also be targeted for additional housing.
The region's vulnerable populations in disadvantaged communities will be least resilient in a water limited future.The combined pressures of climate variability, water quality, and aging infrastructure which will likely lead to price increases up to four times current rates in coming decades (Baird, 2010).If extreme climate event trends continue with changing climate, additional costs to improve wastewater infrastructure for storm water treatment will be incurred and passed on to consumers.These price increases often disproportionately affect the least resilient communities (Feinstein et al., 2017;Mack & Wrase, 2017), as higher prices consume a larger proportion of monthly income.

Assessment of historic policy impacts
Between 1990-2006, over two-thirds of cities and counties in coastal California's metropolitan areas adopted policies explicitly aimed at limiting urban development by restricting housing growth (Legislative Analyst Office, 2015).Additionally, laws adopted between 1992-2001 required the demonstration of a sustainable water supply for new suburban and urban housing developments.Our projections showed a clear drop in rates of development following the passage of these laws, suggesting that they were effective.
Our scenarios illustrated that while likely limiting development, these policies were nevertheless unable to achieve long-term groundwater sustainability in the Central Coast.FMMP data was not available prior to 1992, and thus the impact of these laws on different LULC rates could not be directly assessed, but they did not prevent LULC from increasing water demand overall in overdrafted basins.Thus, the 1992-2001 water laws restricting urban development, while effective at slowing rates of urban growth, were unable to promote water sustainability because they did not impact the agricultural expansion, particularly of perennial crops.More concerning, stakeholders in meetings expressed a serious concern about local housing shortages, particularly around the critically overdrafted Monterey Bay.These laws may have contributed to this shortage by both throttling the development of new housing units and consequently increasing housing costs.
Our results can be used to inform the development of groundwater sustainability plans by local groundwater sustainability agencies (Table 2) in critically overdrafted basins, as required under SGMA (AB 1739, SB 1168, and SB 1319, passed in 2014;Leahy, 2016).In 2012 the California legislature also passed AB 685-the Human Right to Water Bill-becoming the first state to declare access to safe, clean, affordable, and accessible water adequate for human consumption as a basic human right (Fong et al., 2012).However, this doesn't account for potential impacts from changing supplies, increasing demand, or a changing climate.Our results indicate the previous approach of regulating urban and suburban development is unlikely to address water demand challenge posed by the expansion of perennial agriculture.If perennial water demand projections continue to rise, multi-pronged conservation and technology implementation strategies will be needed to avoid continued groundwater depletion and to meet the sustainability goals outlined in SGMA (Dickinson, 2014;Pavley, 2014aPavley, , 2014b)).

Future directions
Additional scenario development, which includes continued feedback from local and regional stakeholders, including individual land holders and farmers, will be needed to test alternative regional mitigation strategies and their associated outcome on water demand change.Projections of future land change and water demand would also greatly benefit from more advanced, fully coupled modeling approaches, involving climate-driven hydrological models and the LUCAS land change model.Such an integrated system would facilitate more informed, process-based interactions and feedbacks between models during a model run, between timesteps and iterations.
This would enable the direct utilization of established climate projections with hydrologic modeling to examine human-environment system feedbacks and stressors.The LUCAS framework is already based on the open source ST-Sim model platform (Daniel et al., 2016), which includes a module to facilitate information passing between integrated systems using a Python or R code interface (R Core Team, 2017).Such an approach could include more accurate, process-based analysis of cropland water demand, a more detailed cropland classification scheme, and could serve to identify couplings between human land-use related water demand and forcings on the regional hydrologic system.
as our initial state class conditions, modified for our study region as follows: 1) all four developed classes were collapsed into a single developed class and urban core areas defined per Soulard and Acevedo 2017; 2) the three forest classes were combined into a single forest class; 3) the woody and emergent wetlands classes were combined into a single wetlands class; 4) the agriculture and hay pasture classes were combined into a single annual agriculture class; 5) we used data from Sleeter et al. 2019 for the 2001 perennial agriculture class, described in more detail below; and 6) the "Developed-Roads" class from Landfire's Existing Vegetation Cover 2001 was used to designate a transportation class (LANDFIRE Program, 2019) (Figure 1b).All datasets were resampled from 30-m to 270-m and re-projected into NAD 1983 California Teale Albers map projection.

Figure 1 -
Figure 1 -California's Central Coast Study Area including a) counties, b) land use and land cover in 2001, c) groundwater sub-basins, and d) aggregated water district and groundwater sustainability agency jurisdictions.Complete lists of regions included in c) and d) located in the Supplementary Materials Tables1 and 2, respectively.
Perennial -Historical and Projected Historical perennial cropland expansion in the Central Coast has been spatially disparate and has not occurred near existing cropland areas.Most new perennial crops have been planted in previously open rangeland and valley uplands.In order to capture this spatially anomalous historical trend with observed data, we developed a "To Perennial 2018" spatial multiplier for the historical period (through 2018) by combining two spatial datasets.We used the Crop Mapping 2014 dataset from the California Natural Resources Agency for orchard and vineyard classes (California Department of Water Resources, 2019a addition to tracking state class variables, the model was parameterized to track water use by county and state class type using data fromWilson et al. (2016).They calculated average county level applied water use for the annual and perennial cropland classes by reclassifying the USDA Cropland Data Layer (CDL) (United States Department of Agriculture, 2011) by cropland categories associated with the California Department of Water Resources (CDWR) Agricultural Land & Water Use1998-2010  dataset (CDWR, 2014)).These were then aggregated into annual and perennial cropland classes and assigned an area-weighted average applied water use value for each combination of county and state class type.For the developed class, they derived applied water use from a national dataset of water use in 2010 by various sectors(Maupin et al., 2014).Applied water use for the developed state class was calculated as a sum of public supply freshwater and industrial self-supplied water and divided by the total developed area in each county based on the NLCD 2011(Homer et al., 2015).The NLCD 2011 most closely aligned with the 2010 water data to get a use per unit area estimate.2.7 Land Use and Land Cover ScenariosTwo LULC change scenarios were modeled to examine how projections of future land change based on longer term land change would compare to projections based only on modern land change trajectories.The first scenario, referred to hereafter as the Business-As-Usual (BAU) scenario, randomly samples from the full 1992-2016 FMMP land change record beginning in projected year 2017.The second Recent-Modern (RM) scenario which samples from 2002-2016 FMMP record alone.The RM scenario is intended to both capture more restrictive land use policies implemented in 2001 to restrict development in some regions, while also capturing recent drought-related trends.For any simulation year, LUCAS randomly samples from one of these historic years, sampling all associated LULC transitions, preserving LULC change covariance.

Figure 2 .
Figure 2. Projected land use and land cover change from 2001-2100 under a business-as-usual (BAU; red) and recent modern (RM; blue) scenarios for the California Central Coast, including Annual Cropland, Cropland (sums Annual Cropland and Perennial Cropland), Developed, Rangeland, and Perennial Cropland.Dark center trendline is the mean for each scenario and shaded area represents the minimum and maximum value ranges across 10 Monte Carlo simulations.
28.0% (BAU) by 2100.In both scenarios, development in San Luis Obispo increased an average 28.5%.County-level trends varied greatly between scenarios losses in rangelands.When accounting for overall percent loss from 2001-2100, Santa Cruz County was projected to lose between an average 25.9% (RM) and 27.4% (BAU) of its rangelands.Conversely, San Benito County had projected increased natural lands in rangeland, following recent FMMP trends in agricultural contraction.Figure4shows the mapped LULC projections under the RM scenario to demonstrate spatial placement of change.

Figure 3 .
Figure 3. Projected change in land use and land cover from 2001-2100 under a business-as-usual (BAU) and recent modern (RM) scenario for each county in the California's Central Coast region, expressed as average net change in annual cropland (orange), perennial cropland (brown), development (blue), and rangeland (yellow) across the modeled period and 10 Monte Carlo simulations.

Figure 4 .
Figure 4. Projected land-use and land-cover (LULC) change from 2001-2100 in 50-year increments for California's Central Coast region under the Business-As-Usual (BAU) and Recent Modern (RM) scenario.Each map represents one out of 10 possible Monte Carlo simulations modeled for each time step.

Figure 5 .
Figure 5. Projected land-use related water demand in billions of cubic meters (Bm 3 ) from 2001-2100 in California's Central Coast under a business-as-usual (BAU; red) and recent modern (RM; blue) scenarios.Darker center lines represent the mean and shaded area represents the maximum and minimum values across 10 Monte Carlo simulations.

Figure 6 .
Figure 6.Net change in water demand in millions of cubic meters (Mm 3 ) from 2001-2100 by land use and land cover class and county for the business-as-usual (BAU) and recent modern (RM) scenarios.

Figure 7 .
Figure 7. Projected change in water demand for groundwater sub-basins from the a) business-asusual (BAU) by 2050, b) BAU by 2100, c) recent modern (RM) by 2050, and d) RM by 2100.Hatched lines shown in b) represent existing state-regulated groundwater basins already experiencing overdraft.

4. 1
Projected water demand trends Projections show increasing land-use related water demand by 2100 of between 222.7 and 310.6 Mm 3 in the BAU and RM scenarios, respectively.Increased demand was driven by continued agricultural intensification (i.e., increasing perennial cropland) and urbanization, even as annual cropland water use declined.Additional increased demand was driven by continued urbanization, generating additional per capita use needs.For the BAU scenario development-related increases in water demand outpaced increased demand from perennial cropland, while the opposite was the case in the RM.This difference illuminated trends noted in the historical FMMP dataset, showing marked declines in urbanization beginning around 2003.The RM scenario only sampled from FMMP-based LULC change years 2002-2016, thus capturing land use changes likely associated with legislative mandates which imposed water use restrictions for new development.We sought to capture this declining urbanization trend as well as the unprecedented 2011-2016 drought in our RM scenario projections.Despite slower rates of

Table 1 .
for a full list of data used to prohibit urbanization transitions (i.e."No Conversion) or promote urbanization transitions (i.e.Spatial datasets and zoning categories used in land use and land cover transition spatial multipliers for designating regions as "No Conversion" and regions of potential development or "To Developed" with data sources listed for the Central Coast study region and for each county.

Table 2 .
Projected percent (%) change in water demand for SGMA groundwater sustainability agencies of the Central Coast by 2050 and 2100 under two scenarios, a Business-as-Usual (BAU; fit to 1992-2016 land use change rates) and Recent-Modern (RM; fit to 2002-2016).

Table 3 .
Projected percent (%) change in water demand in water districts of the Central Coast (excluding GSAs and county agencies) by 2050 and 2100 under two scenarios, a Business-as-Usual (BAU; fit to 1992-2016 land use change rates) and Recent-Modern (RM; fit to 2002-2016).