Modeling trade-offs among ecosystem services for agriculture in the “sisal belt” of Kilosa, central Tanzania

Exploring ways to maintain a biophysically functioning environment while improving human welfare based on competing stakeholder land uses is critical for sustainable development, especially under the context of a surging “global land rush”. This research (1) integrates different stakeholders’ perceptions of human-environmental conditions and dynamics in the “sisal belt” of Kilosa, Tanzania, in terms of three alternative development visions or scenarios of land uses and covers, and (2) demonstrates the trade-offs and synergies among several ecosystem and economic outcomes at a landscape level. Two spatially explicit modeling tools, Future Land Use Simulation (FLUS) and Integrated Valuation of Ecosystem Services and Trade-offs (InVEST), are combined to assess future land-use and -cover patterns and project changes in four ecosystem services, including provisioning commodity production, under the three stakeholder-defined scenarios for the study area up to 2030. Each scenario had higher commodity production values relative to the baseline conditions of 2018 but lower levels of ecosystem services addressed at the level of the Kilosa sisal belt landscape. Carbon and water services may generate synergistic effects provided specific mitigation and payment mechanisms are installed. The spatial distribution of the changes in these services is projected. Our approach provides an empirical-based platform to inform landscape management and planning. It provides a co-designed means to address possible futures of human-environmental conditions affecting sustainability, in this case for food production, resource use, poverty alleviation, and environmental conservation.

Objectives This research (1) integrates different stakeholders' perceptions of human-environmental conditions and dynamics in the "sisal belt" of Kilosa, Tanzania, in terms of three alternative development visions or scenarios of land uses and covers, and (2) demonstrates the trade-offs and synergies among several ecosystem and economic outcomes at a landscape level. Methods Two spatially explicit modeling tools, Future Land Use Simulation (FLUS) and Integrated Valuation of Ecosystem Services and Trade-offs (InVEST), are combined to assess future land-use and -cover patterns and project changes in four ecosystem services, including provisioning commodity production, under the three stakeholder-defined scenarios for the study area up to 2030. Results Each scenario had higher commodity production values relative to the baseline conditions of 2018 but lower levels of ecosystem services addressed at the level of the Kilosa sisal belt landscape. Carbon and water services may generate synergistic effects provided specific mitigation and payment mechanisms are installed. The spatial distribution of the changes in these services is projected. Conclusions Our approach provides an empiricalbased platform to inform landscape management and planning. It provides a co-designed means to address possible futures of human-environmental conditions affecting sustainability, in this case for food production, resource use, poverty alleviation, and environmental conservation.

Introduction
Land estates, smallholders, and sustainable landscapes The onset of the twenty-first century has escalated competition over global arable land (Meyfroidt 2018). While the increased agricultural demand in the latter half of the previous century had been met by increasing the productivity and intensification of agriculture, various assessments indicate that an expansion of agricultural land will be needed to meet the nearfuture global demands (e.g., Williams et al. 2020). Moreover, the recent global crises in food, energy, finance, and the environment compound this challenge, translating into the acquisition of large tracts of land internationally by those actors sufficiently powerful to do so to provision various products (Cotula 2012;Müller et al. 2021).
In such circumstances, large-scale farmland acquisitions in the global South have surged, creating a "global land rush," commonly referred to as land grabbing (Messerli et al. 2013;Yang and He 2021), that may also involve water grabbing (Rulli et al. 2013). According to the Land Matrix (2021), most of these land-based investments involve the establishment and operations of transnational agricultural estates for commercial production. A major share of such estate-driven land grabbing has been undertaken in Sub-Saharan Africa (SSA) (Hufe and Heuermann 2017;Baterbury and Ndi 2018;Ashukem 2020;Neef 2020).
The implications of this wave of land rush across SSA have been controversial, especially when intertwined with local smallholder livelihoods (Herrmann 2017). On the one hand, estate investments may enhance smallholder welfare through employment, technology and expertise transfer, and new opportunities for commercial farming, potentially leading to rural economic transformation (Deininger and Byerlee 2012;Hall et al. 2015a, b). On the other hand, the concerns about the loss of smallholder land access, disrupted subsistence food provisioning, and increasing risks of environmental degradation are widespread (Edelman et al. 2013;Hall et al. 2017;Sulle 2020). Different outcomes, both perceived and actual, often involve the trade-offs between smallholder production, estate development, and ecological functionalities.
Such trade-offs are apparent in the Kilosa "sisal belt" area, central Tanzania (hereafter, KSB), where Chinese firms have resurrected former colonial sisal estates and offered wage opportunities to neighboring communities, and where smallholders have increased commercial rice production, while preserving basic subsistence production. The combination of estate labor, subsistence, and rice cultivation has enhanced the livelihoods of many households but increased land pressures from rice expansion and arriving migrants' subsistence activities, with consequences for the land uses across of sisal belt landscape.
What are these consequences and sustainability implications for the estate-smallholder conditions--as a coupled human-environmental system at the landscape level--now and projected into the future? These questions are central to informing global environmental change and sustainability sciences and serve as a backdrop to development decisions (e.g., Wu 2013Wu , 2019Lambin et al. 2021;Turner et al. 2021). Ecosystems generate a range of goods and services essential for human wellbeing--collectively called ecosystem services (Kumar 2010). Understanding the socioeconomic and environmental consequences of these services under plausible land uses of smallholders and the estate is instructive and may inform local resource management and environmental decision-making relevant to the problem.
This study integrates public perceptions and experiences across different stakeholders by way of empirical field data to create three landscape-based, decision-making scenarios in KSB, based on future estate-smallholder visions of alternative land uses influenced by NGO and state environment and development concerns. The application of three environmental functions and economic valuation models, integrated into InVEST (Integrated Valuation of Ecosystem Services and Trade-offs), demonstrates how these scenarios affect carbon storage, hydrological services (water quality and annual yield), and agricultural provisioning as measured by the value of several marketed commodities (agricultural products). Importantly, we also explore the spatial patterns of these services across the landscape under the scenarios, highlighting synergies and trade-offs between multiple ecosystem services and market returns.

Challenge of coupled human-environmental assessments
Large-scale estate investments in agriculture play an increasingly critical role in raising pressure on the landscapes of Tanzania. Whether these investments promote quality livelihoods and vibrant rural economies without compromising environmental losses remains a key question. Despite the existing development and policy guidelines developed at the international (e.g., Committee on World Food Security 2014) and state and regional levels (e.g., The Tanzanian Development Vision 2025; Kilimo Kwanza--Agriculture First; Southern Agricultural Growth Corridor of Tanzania--SAGCOT) to hold foreign-owned estates accountable for sustainable land use and investments, a paucity of empirical evidence exists on how the integrated goals can be reconciled. The challenge is twofold.
First, nature provides a wide range of benefits to people. There has been increasing consensus about the importance of factoring these ecosystem services into local and regional resource use decisions (Nelson et al. 2009;Tallis and Polasky 2009;Ronchi 2021). These services, however, are spatially explicit and temporally dependent--not only does the amount of a service matter, but also where and when it is provided, rendering difficulties for policy and planning assessments. Second, different stakeholders may seek different demands from the services (Li et al. 2021). In the case of KSB, the international development agencies and local governments emphasize a robust landscape, presumably maintaining a variety of ecosystem services, whereas the Chinese sisal estates and neighboring smallholder farmers, the two most important producers, land users, and stakeholders in KSB, focus on the service of provisioning high agricultural returns. Reconciliation and compromises of these competing demands require novel methods and datasets to identify potential trade-offs and synergies between different services (Reed et al. 2016), presumably to optimize outcomes.
Various frameworks employ social-environmental assessments to integrate stakeholders' distinctions regarding ecosystem services (e.g., Whitfield and Reed 2012;Malinga et al. 2013;Villamor et al. 2014). Many of them are largely heuristic, however, with minimal capacity to provide quantitative and spatially explicit information about the likely consequences in question. Biophysical and economic models, in contrast, quantify the impacts on multiple ecosystem services derived from different policies and actions, allowing for various simplifications and uncertainties (e.g., Bagstad et al. 2013;Duarte et al. 2016;Xu et al. 2018). The model influence on actual practice is, however, commonly muted in cases where local stakeholders have not been party to the model development or the policies that follow. Their involvement constitutes a co-design practice, one championed by the sustainability sciences (Turner et al. 2016). Efforts addressing multiple ecosystem services by way of co-design (e.g., stakeholders, scientists, models) approaches have been few (Goldstein et al. 2012), but are increasing (e.g., Karrasch et al. 2017;Saito et al. 2019;Asah and Blahna 2020;Li et al. 2021).
The key assumption in sustainable development is the need to balance the trade-offs among the targeted ecosystem services (Reed et al. 2020;Wu 2021), identified by various stakeholders, which includes produce among agriculturalists. In the case of KSB, or under the prevailing circumstances of the land rush underway across SSA in general, the services supporting estate development, local livelihoods, and international/state/regional concerns for landscape conservation are at issue, with the aim of producing a land-use strategy that serves the needs and wants of all parties.
Based on existing conditions and processes, our approach employs co-design scenarios--plausible descriptions of future states of the KSB developed with local stakeholders--applied in modeling exercises providing spatially explicit, quantitative outcomes. To represent realistic future possibilities, three overarching scenarios of estate-smallholder nexuses, land use, and local development in KSB are explored through the details of internally consistent storylines grounded on the competing interests and expectations among stakeholders, especially the Chinese sisal estate and local smallholders. Specific attention is given to the services regarding carbon, water, and commodity production to inform local land-use decisions and policy implementations. This assessment constitutes a relatively underdeveloped approach and the first quantitative-scenario-based attempt to address the human-environmental futures of KSB.

The study area, data, and methods
The Kilosa sisal belt, Tanzania As one of the six districts of the Morogoro Region in Tanzania, Kilosa has a long history of sisal production. At the time of independence in 1961 (Tanganyika), there were 14 sisal estates in operation in Kilosa (Floor and Kaihura 1990). These estates were either abandoned or discontinued by the late 1980s. In 1999, China Sisal Farm (CSF) consolidated two colonial sisal estates (Rudewa and Kisangata) with a total area of 6900 ha to produce and export roughlyprocessed sisal fibers to China. By 2018, CSF had rehabilitated 2300 ha sisal plantation, creating the second-largest sisal fiber producer and exporter in Tanzania (Fig. 1). The CSF can only operate within its defined boundaries (Fig. 1).
Located in the central-northern part of the Kilosa District, the KSB area comprises three wards, Moswero, Rudewa, and Madoto, covering 1433 km 2 at elevations ranging from 368 to 1802 m above sea level. Three major landscapes constitute KSB's territory, each supplying different ecosystem services supporting distinct livelihoods (Fig. 2). A forested highland dominates the northwestern part of KSB, descending southeastward into a sparsely-vegetated lowland plain in which the majority of the KSB population practices subsistence activities and the estate operates. This plain stretches to the east and south from B127 Road, where it meets a vast woody savanna that provides support for Maasai nomadic herding, one of the few human activities undertaken in this space. In addition, the forested and lushly vegetated floodplains of the Wami River and its branches are exploited for various purposes, such as timber, firewood and charcoal, and hunting.
The CSF sisal fields intersect with the three wards, and sisal operations provide smallholders from 12 neighboring villages with an additional means of employment. Smallholders typically cultivate maize and beans on their land for subsistence, and increasingly, rainfed rice for the market. The CSF, by contrast, relies on smallholder labor to maintain its operations. Wage labor poses a distinct attraction for migrant laborers, causing the KSB population to have nearly tripled over the past two decades. To expand sisal production, CSF plans to enter into outgrowing schemes in which the adjacent smallholders are encouraged and subsidized to produce sisal on their own land that is near roads to facilitate the transport of sisal to the CSF. The outgrowers are paid through a division of proceeds after the CSF sells the fibers. Some pilot smallholder outgrowers have planted sisal in their fields since 2016. This scheme is expected to propagate The estate-smallholder interactions also have significant ecological impacts beyond the provisioning for crop production. In the past two decades, lowland and mountain forests have been converted into cultivated land, posing threats to maintaining multiple ecosystem services, especially those regarding water, soil, and carbon. Hence, fostering the local economy while maintaining a healthy landscape has been emphasized in various local policies and development goals (e.g., Bergius et al. 2020). Within KSB, however, inadequate farming practices, conversion of native land covers to cultivated land, and large-scale estate investments are underway, generating ecosystem service trade-offs with socioeconomic consequences among different stakeholders.

Data
The key dataset used in our research is a 2018 landuse land-cover (LULC) base map created from Google Earth Engine (GEE, https:// earth engine. google. com) remote sensing and classification algorithms based on the synergistic use of Sentinel-1 and Sentinel-2 image collection ( Fig. 2: 2018 baseline map). The creation of the 2018 base map is detailed in Li (2022: 38-44), but briefly illustrated through Table S1 and S2. Nine LULC classes are identified in Table 1, informed by local knowledge about phenology.
InVEST (below) contains a suite of modules that use LULC maps to evaluate the environmental and economic values of ecosystem services provided by a landscape (Sharp et al. 2018). Our research ran a subset of InVEST modules to project changes in land uses with a specific focus on water quality and annual yield, carbon storage, and the value of several marketed commodities. Other datasets required to run InVEST are listed in Table S3. All data were projected into UTM 37S WGS 1984 geographic coordinate system with the raster datasets resampled to 30 m resolution. Input coefficients for running the carbon and water service modules were derived from multiple sources noted in Table 2 and Table 3. Importantly, climate change is not considered, given the complexity of the model and the temporal proximity of the projection year, 2030, to the current average conditions, foremost precipitation, in Kilosa. All assessments include the entirety of the study region or KSB landscape ( Fig. 1). Estate land practices are restricted to the estate's boundaries. Smallholder practices remain within the limits of the villages in question, with Water Wami River and its tributaries, most of which are seasonal, and some ponds 7 Forest Highland densely-vegetated woodland 8 Abandoned sisal Deserted old sisal fields, usually interspersed with trees/shrubs 9 Outgrowing sisal Sisal planted by smallholder outgrowers under contract Table 2 Input values for the carbon storage/sequestration model We could not find relevant datasets for the parameter estimates of sisal (Agave sisalana). As such, we drew on the data for, e.g., Agave decipiens (false sisal hemp) in southern Africa, America, and Asia and Agave americana (blue agave) in West Indies and Mexico. For instance, the carbon storage in aboveground Agave americana biomass in Mexico is estimated by Cummins (2019) as ~ 33 tC/ha. This estimate fits within the range estimated by Vuorinne et al. (2021) for the fiber on the Teita Sisal Estate of Kenya, as well as with our field assessments. In practice, we assumed that carbon was 50% of total biomass and used root-to-shoot ratios to estimate belowground biomass based on specified values for aboveground biomass. We estimated biomass values to reflect the total storage capacity for each LULC type based on Tanzanian studies, where available, and otherwise from other sources noted above

Scenario and storylines
The InVEST model is based on scenarios and storylines that stake out, respectively, the broad dynamics driving land uses and covers and the specific socioeconomic and biophysical dimensions in each overarching case. Our approach used a participatory development method to create stakeholder-defined scenarios and storylines to represent alternative estate-smallholder nexuses and the consequent landuse and local development visions in KSB for 2030. This date constitutes the termination of CSF's shortterm plan, the 2030 Tanzania Implementation Agenda for Sustainable Development (United Nations 2015), as well as other state and regional development policies and strategies, for example, SAGCOT (Bergius et al. 2018), and Intended Nationally Determined Contributions (INDCs--The United Republic of Tanzania 2015). Three steps guided the scenario-storyline process.
Step One was initiated by a review of the literature related to the state and regional land-use and development policies and strategy papers related to the relevant economic and environmental sectors, informed by consultative advice from local experts.
Step Two involved extensive interviews with 80 smallholders engaged in estate labor and household agricultural production and with the entirety of the Chinese managerial staff of the estates. This effort clarified current trends in resource use, livelihoods, and commodity production, and led to the development of visions of the land system dynamics and the resulting socialenvironmental consequences in KSB by the year 2030. These visions were built on plausible underlying estate-smallholder relationships and included three contrasting scenarios: (1) Business-as-Usual (BAU), the conditions and dynamics present today continue into the near future; (2) Formal Wage Labor (FWL), the estate expands, offering suitable and consistent wages to a larger number of fulltime laborers; and (3) Outgrowing Scheme (OGS), the estate modestly grows but increasingly relies on sustained production of sisal from outgrowers.
For each of these three scenarios, several storylines about socioeconomic and environmental dynamics were developed in Step Three. These storylines were discussed in group interviews with representatives of a broad range of stakeholders, including smallholders and estate staff, local policymakers, researchers, and members of NGOs. The informants were asked to evaluate and help rank the likelihood that each storyline would take place, and if occurring, the extent to which each storyline would impact LULC and various ecosystem services across the region. Based on the responses, new storylines were added if necessary, and others were eliminated if deemed unlikely to occur. The top-ranked storylines were compiled and finalized for each scenario, as noted below and detailed in Table S4.
BAU includes continued population growth and insufficient protection of existing natural resources. It assumes that by 2030 development follows its current trajectory, with most smallholder households maintaining a livelihood combination of estate labor, subsistence, and rice cultivation, the CSF adding sisal at the current rate (80 ha annually), and weak local-to-state governance along with few financial incentives for sustainable development in KSB. A rapidly growing population (average annual growth rate for Kilosa, 5%), ongoing resource exploitation, and non-implementation of nature conservation plans at any meaningful scale lead to land-use/cover conversion, environmental degradation, and slow-to-moderate growth in estate revenue and family income.
FWL represents a desire by most landless smallholders who hope to work in the sisal estate full-time under a formal contract. There are two direct results. First, the population grows at a remarkable rate--6%, consistent with that of the past sisal boom periods, largely owing to migrants seeking wage opportunities on the estate. Second, the costs of CSF to operate the estate increase significantly as a result of the rising payment for employment and expedited estate expansion. CSF estimates that it needs to add at least 200 ha of new sisal fields into production every year to provide enough jobs for the growing wage seekers, part of their agreement with Tanzania, and to maintain a profitable operation. Cropland increases extensively to meet the rising subsistence demands, whereas rice expansion is constrained due to the restructuring of intrahousehold labor allocation, where fixed labor forces are directed from household fields to the estate. United Nations Strategic Plan for Forest 2017-2030 (3% increase in forest stock--UNSPF) is conditionally implemented.
Outgrowing Scheme (OGS) involves an optimistic scenario of the future, where KSB meets all its stated policy goals to alleviate poverty and manage natural resources sustainably. Existing forest resources are conserved and increase 10% by 2030, following the aim of AFR100--the African Forest Landscape Restoration Initiative (Gizachew et al. 2020;Owusu et al. 2021). The population continues to grow, but slowly, at the growth rate stated by INDC, 1.5%. Some larger smallholder landholders (households with six or more acres of land; acre is the local metric) are engaged in the outgrowing scheme initiated by CSF, allocating uncultivated landholdings to grow sisal under contract. They are expected to receive monetary returns from the outgrowing sisal produced and sold to the estate five to six years subsequent to seeding. Cropland declines moderately to the quantity that meets the basic subsistence requirements with bare crops remaining. Essentially, per household land uses for rice cultivation largely stagnate, as an increasing number of labor forces are channeled into the outgrowing scheme.
Group interviews with major land users, especially smallholders, also helped create rules reflecting constraints for specific land conversions (Table S5). In line with the land-use plans under various scenarios, storylines, and local practices, land users were asked to evaluate the land conversion probabilities. We averaged these evaluations and performed pairwise comparison procedures (Saaty 1977) to derive the best-fit set of weights for each land conversion (Table S6). 1 Lastly, we factored these derived weights into LULC projections by using Future Land Use Simulation (FLUS), an integrated software application based on coupled system dynamics and cellular automata algorithms for land change analysis and prediction, to quantify and map the LULC patterns for the three estate-smallholder relationship and scenarios (see Liu et al. 2017).

InVEST models
Ecosystem services and commodity production values are a function of land characteristics and landscape patterns (Nelson et al. 2009). Using the three scenarios and required datasets (Table S3), we employed InVEST tools to evaluate three critical ecosystem services and the commodity production outcomes, and evaluated each scenario based on four metrics with contrasting beneficiary groups: (i) carbon storage/ sequestration (metric tons C/ha or tC/ha) as a critical global benefit related to climate change mitigation; (ii) water yield (m 3 / year), measured as the flood control capacity, affecting the safety and livelihoods of communities living in the study region; (iii) water quality, focused on the total dissolved phosphorous export from watersheds (kg) as the proxy for pollution, given the proximity of the agricultural lands to water bodies, and (iv) agricultural provisioning as measured by the market value of commodity production (constant year US$2018). These services were selected because measures of them can be approximated with extant, aggregate data and remote sensing data of land cover, and they constitute issues linked to various stakeholders: the state and NGOs seek to reduce carbon emissions; various agencies and smallholders are concerned with flood control and water quality, and of course, with the economy derived from agriculture.
We tracked the carbon stored in above-and belowground biomass, soil, and dead organic matter using standard carbon accounting methods (Lubowski et al. 2006;Nelson et al. 2008;Sharp et al. 2018). The InVEST model aggregated the amount of carbon stored in these pools according to the LULC projections. Land management strongly affects the total carbon stock in the terrestrial system, with implications for soil fertility and CO 2 emissions (Li et al. 2021). The amount of carbon sequestered in an area for a particular period is determined by subtracting the carbon stored in the area at the beginning of the time from that stored in the area at the end time.
The InVEST annual water yield model computes spatial indices that quantify the relative contribution of a parcel of land to the generation of both baseand quick-flow (Kienzle and Mueller 2013). This model estimated the volume of freshwater that runs off in unregulated watersheds, which has significant implications for the equilibrium of local agricultural economy, land use, and hydraulic systems, especially annual flows for surface water.
In this application, we also used the discharge of dissolved phosphorus (P) into the local watershed to measure water pollution. Although this single measure ignores other sources of water pollution, it provides a proxy for non-point-source pollution. Slope, soil depth, and surface permeability were the major indexes used to define potential runoff by location (Nelson et al. 2009). Areas with more potential runoff, less downhill natural vegetation for filtering, greater hydraulic connectivity to water bodies, and LULC associated with the export of phosphorous (e.g., agricultural land, and more significantly, the sisal fields in this case) have greater rates of phosphorus discharge.
The market value of commodities is represented as the aggregate net present value of commodities produced in the area. We focus on the production of three major crops: maize, the primary source of subsistence, commercial rice, and sisal. We excluded the value of local rural-residential housing due to local data constraints. Livestock raising, aquaculture, timber logging, hunting, and charcoal production exist but were also excluded from this assessment because they either account for only a small fraction of the local economy or are largely prohibited in current community resource administration settings. The estimate of the net present value of crops depends on the crop type, productivity, market prices, and production costs. We derived these variables for 2018 estimates and 2030 projections from field surveys and international assessments (Table S7). In cases of missing local data, such as local crop production costs, data for the same product elsewhere in Tanzania were substituted. We used a discount rate of 5% per annum to compute the net present values of commodity production across time (Moner-Girona et al. 2016).

From scenarios to LULC projections
The storylines determined plausible trends and magnitude of LULC changes. For example, the increase in the residential area essentially corresponds to the population growth under all three scenarios, as does household cropland but as determined by per household subsistence requirements. If the present deforestation rate continues (1.2% per annum, 2019 field data), KSB will lose approximately 13.5% of forest for other uses by 2030 under BAU. In contrast, a 3% and 10% increase would follow from the stricter forest management and restoration guidance by UNSPF and AFR100 under FWL and OGS conditions, respectively. Rice expansion may reach its cultivation cap, 125% of the 2018 level, based on smallholders' assessment that 80% of such land was already used in 2018 and could not increase under BAU, whereas FWL and OGS only gain moderate increases in rice fields, primarily constrained by labor capabilities and the emerging new labor division plans. For example, a tendency exists for households to divert more laborers from commercial rice to the estate as the latter has become a more profitable livelihood option since 2016 (2019 field data), and the emerging opportunities for outgrowing sisal production.
Mappings shown in Fig. 3 reveal that the KSB would experience substantial LULC changes relative to baseline conditions under all three scenarios between 2018 and 2030. KSB under BAU lost large parcels of forest to cropland, most of which occurred in the bordering area between the core sisal belt and the forestland in the northwest of the KSB due to its relatively lower elevation and proximity to the most populated region. Forest under FWL and OGS gained in alignment with corresponding forest restoration guidelines. 2 The decreases in savanna represented the most extensive native land losses in the KSB area. Under BAU, KSB lost 6.5% of the savanna primarily to cropland and rice fields as the population grew and sought to expand subsistence cultivation and maximize the profits from commercial crop sales. In addition, savanna loss also resulted from increases in the estate-land restoration (FWL) and outgrowing (OGS) sisal cultivation. Despite the substantial land gains for cropland in response to the rapidly increasing subsistence demands under BAU and FWL, cropland under OGS declined in area sufficient to meet the base needs, as household labor switched to rice cultivation and outgrowing sisal production. Rice fields expanded remarkably under BAU and reached its limits of suitable lands. In contrast, FWL and OGS gained moderate rice field increases. While estate sisal expansion only took place within CSF-prescribed territory, sisal outgrowers prioritized rehabilitating abandoned sisal on their smallholder farms before converting other uncultivated landholdings, provided physical conditions (e.g., moisture, slope, and altitude) were suitable. Such land conversion for outgrowing sisal mainly occurred around the northern and southern ends of the core sisal belt, particularly marked in the areas along the major roads, where accessibility to smallholder settlements, CSF fiber processing plant, and other facilities was relatively good (Fig. 3).

Trade-offs and synergy between ecosystem services
Based on the projected mappings, we illustrate the modeling outcomes of ecosystem services and commodity production value under BAU, FWL, and OGS relative to BSL conditions in the metrics shown in Fig. 4 and Table 4. The market value of commodity production increased in many areas under all three scenarios as a result of increased unit present value for both commercial rice and sisal fibers. Although the market value of commodity production declined in some areas under BAU and FWL due to devalued subsistence crops (Table S7), the aggregate market value of commodity production summed over the whole region increased because of the high value of commercial crops (rice and sisal) more than compensated for the losses elsewhere. Consequently, all scenarios considered in this analysis generated positive net present outcomes and substantially exceeded the value of approximately $13 million projected for BSL (Table 4). Specifically, BAU generated the highest net present value of $17.5 million. The FWL generated a net present value of $16.5 million, and the OGS generated $16.6 million. In addition to the rapidly growing estate revenue, most increases in commodity production values came from household monetary wealth growth. Though BAU and FWL outperformed the total net present values, per household income declined relative to BSL because of the increasing number of households; comparatively, OGS generated the largest CSF revenue and per household income. 3 The majority of change in the market value of commodity production occurred in the core sisal belt, with the values outside of the developing areas largely unchanged (Fig. 4).
Increases in land devoted to agriculture exacerbate the loss of various ecosystem services (Fig. 5A-C). The largest losses in the capacity to sequester carbon today follow from the vegetation loss of new lands taken to cultivation, including deforestation (Figs. 3,  4). These impacts are most significant in the BAU scenario, in which carbon reductions relative to BSL are 10.5%. For FWL and OGS, substantial native land losses for carbon-intensive uses notwithstanding, on-site carbon reductions were repaid by following stricter forest restoration strategies, generating a slight decrease (0.5%, FWL) and moderate increase (5.7%, OGS) in carbon sequestration, respectively (Table 4).
Water services scores declined under all three scenarios (Table 4), but OGS exhibited the smallest decline (Table 4 and Fig. 5D-F). Increases in water yield (indicative of increased flood risk at the catchment outlets on Wami River) were greatest under BAU, which had the largest removals of downhill and floodplain vegetation of any of the scenarios. For water quality, sisal field-caring and fiber production are the predominant sources of pollution in the study area because of the use of herbicide and fiber bleach discharge of large amounts of dissolved phosphorous into the water (FAO 2012). 4 The increasing application of phosphate fertilizer in agricultural fields is also of concern. Water quality declined most sharply with FWL (17.3% increase in P export) and BAU (15.5%), as a result of the largest increases in estate sisal production and smallholder agricultural areas, respectively (Table 4). Phosphorous export also increased under OGS, although less steeply (11.2%), mainly because of offset effects by the increments of native vegetation mitigating the water pollution (Table 4 and Fig. 4).
The results indicate that policy and land-use decisions based foremost on market-return-maximizing strategies tend to generate a landscape with lower ecosystem services, as expected. The provisioning of agricultural produce (e.g., part of the market value) maintains an important trade-off for carbon storage and water services. Interestingly, we found little evidence of trade-offs between carbon and water services; indeed, a synergy appears to exist owing to the role of vegetation for both services. The native vegetation maintains more carbon storage than does cultivated land, and its landscape "roughness" reduces surface flows and the loss of water down the watershed with its potential flood impacts. As such, native vegetation also reduces the discharge of phosphorous into the water system.

Considerations
Human uses of landscapes invariably increase some ecosystem services but degrade others or generate disservices. Such is the case identified through the use of the InVEST model applied to the KSB. Provisioning services, as measured by the proxy of Table 4 Comparison of ecosystem services and commodity production values between future scenarios and BSL (baseline) conditions a Under the OGS scenario, CSF suggests that the proceeds after sisal sales be allocated between the estate and outgrowing smallholders at a fixed division ratio of 6:4--CSF earns 60% of the proceeds, and outgrowers receive 40%. This is the tentative plan by CSF, and has not yet reached a consensus with the smallholders interested in the outgrowing scheme b mt C/ha = carbon storage in metric tons per hectare c Only includes household income from crop production (e.g., maize, rice, and outgrowing sisal) The most widely used herbicides are 3, 5, 6-trichloro-2-pyridinyloxyacetic acid and N-phosphonomethyl-glycine (Glyphosate) applied to leaves, stems, rhizomes, and cut plants (Weber 2017). production value, increase under the scenarios presented but at the decline of all other services examined. This relationship is common, if not typical, for the loss of native habitats to many forms of agriculture, and increasingly is called into play owing to the sustainability issues, foremost improving human wellbeing while maintaining a functioning environment (Clark and Harley 2020). These trade-offs are at play in the KSB, with substantial variations created by the scenarios examined.
BAU, as the long-term continuation of BSL conditions, generated the greatest carbon reduction, flood risk, second-greatest water quality decline, and the smallest increase in estate revenue. Per household income growth largely stagnated due to the unchanged intrahousehold labor and income structure. These results suggest that BAU is not an outcome favorable to most stakeholders in the KSB.
FWL appears to increase more secure employment but at a substantial loss of ecosystem services. It envisions that the estate provides sufficiently secure employment and realizes the various promised benefits to the local people. Socioeconomically, this scenario involves estate-standardized sisal fiber production at an unprecedented production scale, creating an enlarged group of off-farm smallholder wage laborers. The local rural economy restructures significantly under this scenario. It fosters the emergence of a large class of off-farm wage workers and incubates subsidiary businesses involving labor support and legal services. Adverse effects are conspicuous, however. First, a complete switch from casual to formal employment means the estate confronts the market and economic uncertainties alone, which increases the operational risks and restrains most estates from making such a move. Secondly, the influx of wage work-seekers increases local-level subsistence demands and consequently expedites the land conversion from native landscapes to smallholder farms, exacerbating the issues of local food provisions and ecosystem services losses.
Of the three scenarios, OGS produces the largest gains (or the smallest losses) in ecosystem services and aggregate market value of commodity production. Carbon sequestration, estate revenue, and perhousehold income increase substantially. Water Trade-off comparison of the improvement or decline in ecosystem service and commodity production value metrics (BAU business-as-usual; FWL formal wage labor; OGS outgrowing scheme) relative to the baseline conditions (BSL) for the sisal belt area. A Carbon sequestration vs. commodity production value; B Water yield vs. commodity production value; C Water quality vs. commodity production value; D Water yield vs. carbon sequestration; E Water quality vs. carbon sequestration; F Water quality vs. water yield services decline, but only slightly compared to BAU and FWL, which also increase the net present aggregate value of commodity production with the tradeoff of carbon reductions. The losses of ecosystem services could be reversed by, for instance, de-phosphorizing the wastewater after sisal fiber processing and creating vegetation buffers around watersheds and agricultural areas--well-established practices that have proven useful elsewhere to increase carbon storage and improve water services (e.g., Correll 1997). These actions might come with trade-offs of increased commodity production costs and land taken out of agricultural and other profitable uses, however, resulting in reduced financial return. More so, neither carbon nor water services currently have a direct price in the study region, meaning that decisions about whether to establish wastewater treatment plants and/ or vegetation buffers hinge on the value assigned by decision-makers to the carbon-and water-service improvements relative to a financial penalty.
The economic restructuring in the OGS also provides challenges. These include the rigidities in the contract terms, which favor the estate over outgrowers; lack of transparency in the weighing and measuring of products; the fixing of prices of inputs and products and harvesting delays; broken guarantees of a ready market for products, and the loss of time for subsistence production. Even more fundamentally, the outgrowing scheme exposes farmers to the vicissitudes of global markets, while tying them to the bottom of the value chain. These drawbacks are consistent with the critical discussions of the extensive literature on contract farming (e.g., Oya 2013;Smalley 2013;Hall et al. 2015a, b), and constitute most of the concern among potential outgrowers, being repeatedly debated in the group interviews with smallholders. Though we have not included the effects of these adversities in the modeling approach and analysis presented here, such concerns may increase the reluctance among smallholders to implement this scheme. More importantly, even though most potential outgrowers hope to maintain their own farming activities while participating in the scheme, this is only feasible for larger landholders (households with six or more acres of land in this study) and would have excluded most smallholders at the very beginning.

Limitations and challenges
Various challenges limit the robustness of our results and point to implications for future research. First, although our models can potentially include drivers other than LULC change, we have not included them in this analysis due to data issues and the complexity of incorporating such data at the scale of the analysis. Furthermore, there may be essential feedback effects, such as the conserved land, that increase development pressure on land adjacent to the conserved area. Including changes in climate, soil, technology, and feedback effects--all of which are likely to drive the socioeconomic and ecological relationships that determine the ecosystem services and production value in the future--awaits future applications of InVEST for the KSB. It must be recognized, however, that increasing the complexity of the model likely increases the uncertainties involved.
The second limitation is the exclusion of the market value of commodities generated in urban areas in any scenario. Although KSB remains mostly rural, larger villages, such as Msowero and Mvumi, have grown extensively and developed businesses, most of which are linked to the estate sisal production, including credit cooperatives, transportation hubs, and sisal leaves storehouses. 5 Since urban market returns tend to be higher than those for other land uses, we may have underestimated the aggregate value of marketed commodities for scenarios in which urban markets increase by 2030. The development values produced on that land lost to the urbanized area may overwhelm the ecosystem services values generated by conserving that land. Therefore, market evaluation services might not always favor conservation, especially in high-value urbanized and urban-like areas.
Thirdly, the ecosystem services evaluated here do not reflect all the concerns of a particular stakeholder group or may not match the expectations among various stakeholders on the same service. For example, state and international concerns about carbon storage and emissions were not highly important to local stakeholders, and neither the estate nor smallholders identified the decline in water quality as a pressing issue, despite its potential negative impacts on the livelihoods of inhabitants living in the broader catchment areas of the lower Wami River.
In general, choices between development and conservation may involve markets or payments for vital ecosystem services, almost all of which are not in place in most cases. In either case, clear links need to be made between services and their value, which include variations among different stakeholders.

Conclusions
Nature provides a range of ecosystem services essential for human wellbeing and the functioning of the environment (Metzger et al. 2006). Maintaining these services while improving human welfare based on different human-environment relationships, in this case, land uses, is critical for sustainable development, the goal of sustainability science (Wu and Hobbs 2002;Wu 2013Wu , 2019Wu , 2021Clark and Harley 2020). Such efforts require calculations of the trade-offs between human and environmental wellbeing, a quantitative exercise that sustainability science seeks to advance. Such calculations and their projections into the future are difficult to address. The sustainable development of the KSB, Tanzania, is exemplary of these needs and challenges. Here, large-scale estate investments are underway, bringing about significant impacts on local smallholder agriculture, the landscape, and state and regional sustainable development efforts for climate change mitigation, food production, poverty alleviation, and diversifying rural economic opportunities. Like many developing regions globally, KSB is a microcosm of different forces at play that intensify pressure on land for competing uses. Here, the competing needs of major stakeholders, especially those of the estate and smallholders, must be considered to envision the sustainability of the KSB now and for future generations. This paper coordinated major stakeholders' needs and expectations to develop three contrasting scenarios and related storylines based on likely estate-smallholder relationships, policies affecting land uses, and local development visions up to the year 2030. Integrated land-change projections (FLUS) and ecosystem service modeling (InVEST) were employed to convert the storylines through modeling rules to project the amount and location (mapping) of the outcomes in question. In doing so, the trade-offs and synergies among a small set of ecosystem services across space and time are illustrated, providing insights for future land use and policy decision-making.
As expected, substantial trade-offs exist between three ecosystem services and the provisioning of agriculture, measured by the proxy of agricultural values, across the KSB landscape. All scenarios that enhance commodity production have reductions in the other ecosystem services to varying extents. As such, concerns that the co-development of the estate and smallholder agriculture will fail to reconcile conservation goals were mostly supported. A positive correlation between carbon and water services is the one clear synergy observed. The increase in native vegetation, foremost the forest stock, has significant mitigation effects on the decline in water services, as demonstrated in the OGS scenario.
Beyond the specifics identified in the study, our approach offers a means to engage divergent stakeholders to project future social-environmental conditions quantitatively. The projected scale of socioeconomic and environmental change is linked to the locations in which the changes are likely to take place. This kind of co-designed analysis helps to make the trade-offs involved in land uses and their change transparent.