EASTERN AND SOUTHERN AFRICA UNITED REPUBLIC OF TANZANIA Tanzania Land Modeling Background Note World Bank Group November 2024 © 2024 The World Bank Group 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org This work is a product of the staff of the International Bank for Reconstruction and Development (IBRD), the International Development Association (IDA), the International Finance Corporation (IFC), and the Multilateral Investment Guarantee Agency (MIGA), collectively known as The World Bank Group, with external contributors. The World Bank Group does not guarantee the accuracy, reliability or completeness of the content included in this work, or the conclusions or judgments described herein, and accepts no responsibility or liability for any omissions or errors (including, without limitation, typographical errors and technical errors) in the content whatsoever or for reliance thereon. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of any of the organizations of The World Bank Group concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of IBRD/IDA, IFC and MIGA, their respective Boards of Executive Directors, and the governments they represent. The contents of this work are intended for general informational purposes only and are not intended to constitute legal, securities, or investment advice, an opinion regarding the appropriateness of any investment, or a solicitation of any type. Some of the organizations of The World Bank Group or their affiliates may have an investment in, provide other advice or services to, or otherwise have a financial interest in, certain of the companies and parties named herein. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of any of IBRD/IDA, IFC and MIGA, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because the World Bank Group encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given and all further permissions that may be required for such use (as noted herein) are acquired. The World Bank does not warrant that the content contained in this work will not infringe on the rights of third parties, and accepts no responsibility or liability in this regard. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. UNITED REPUBLIC OF TANZANIA Tanzania Land Modeling Background Note COUNTRY CLIMATE AND DEVELOPMENT REPORT Table of Contents Acknowledgmentsiii Acronyms and Abbreviations iv 1. Overview of the Analysis 1 2. Land Scenario Creation 2 2.1. Nationally Determined Contribution (NDC) 2 2.2. Existing government sectoral strategies 3 2.3. Identified areas for improvement in government policies and strategies 3 2.4 Constructing the BAU with climate action and ASP with climate action future maps 4 2.5 Allocating irrigation land 4 3. Landscape Assessment Models 5 3.1 Cropland production value 5 3.2 Forestry production value 5 3.3 Grazing production value 6 3.4 Costs 6 3.5 Carbon 7 3.6 Biodiversity 8 Species richness 8 Other metrics 8 4. Developing the Optimization Frontier 9 5. Results 11 5.1 Land scenario comparisons 11 References14 ii  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania Acknowledgments The Tanzania Country Climate and Development Report (CCDR) was prepared by a multisectoral World Bank Group team led by Diji Chandrasekharan Behr, and William Battaile, under the supervision of Paul Jonathan Martin and Abha Prasad, and the direction of Iain Shuker. This background report on Land Modelling was an input to the Tanzania CCDR. It was prepared by Peter Hawthorne and Saleh Mamun of Natural Capital Insights. The team is thankful for the financing for this work which was provided by the Nature-Based Solutions Invest project using funding from the Least Developed Countries Fund of the Global Environment Facility. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  iii Abbreviations and Acronyms ASP aspirational (scenario) BAU business-as-usual (scenario) CCDR Country Climate Change and Development Report CGE Computable General Equilibrium (model) CO2e carbon dioxide equivalent ESA European Space Agency FAO Food and Agriculture Organization of the United Nations GTAP Global Trade Analysis Project IEc Industrial Economics IMPACT International Model for Policy Analysis of Agricultural Commodities and Trade IUCN International Union for Conservation of Nature kgCO2e kilograms of carbon dioxide equivalent MtCO2e million tonnes of carbon dioxide equivalent NDC nationally determined contribution NPP net primary production PREDICTS Projecting Responses of Ecological Diversity In Changing Terrestrial Systems (database) tCO2e tonnes of carbon dioxide equivalent All dollar amounts ($) are US dollars iv  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania 1. Overview of the Analysis The goal of this analysis is twofold: to assess the economic and environmental outcomes of the business- as-usual (BAU) and aspirational (ASP) economy scenarios as described in the Tanzania Country Climate Change and Development Report (CCDR), and contextualize each of these within the maximum attainable potential for the landscape. The analysis creates several targeted land use maps that reflect the various goals and constraints underlying the BAU and ASP scenarios specified for the CCDR and assess landscapes for economic productivity and provision of environmental goods. Complementary work by Industrial Economics (IEc) for the Tanzania CCDR evaluates the potential climate impacts given these land use scenarios. An optimization analysis creates the “production possibility frontier”, which shows the best possible combinations of agricultural and environmental production under different landscape configurations. This analysis allows decision-makers to see how the BAU and ASP transition pathways move towards this frontier and to identify other possible pathways that strike the desired balance between economic and environmental benefits. This document lays out the methods and data used to carry out this analysis and the findings from the analysis. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  1 2. Land Scenario Creation 2.1 Constructing the baseline current land use map We first construct a land use and land cover map that represents the current state.1 This map serves three purposes in this analysis: • To provide an estimate of baseline ecosystem service and agricultural production. • As a starting point for the creation of the 2050 land use scenarios and accompanying land-change pathways. • To be summarized into coarser land use and land cover categories to provide the land resource base for the macroeconomic modeling. A major challenge is that the most detailed available maps of land cover come from remote sensing, but these maps are unable to capture land use accurately. Specifically, grazing and forestry, which are important agricultural activities that also impact environmental outcomes, are hard to detect and spatialize. As part of the World Bank’s Nature’s Frontiers work (Damania et al. 2023),2 we use globally applicable methods to estimate the spatial extent and location of these activities for 146 countries (including Tanzania). For this analysis, we apply similar modeling methods, but updated with national data available for Tanzania, as described below. We start with the 2015 land cover map from the European Space Agency (ESA) (ESA 2019). This map was originally chosen for analysis in the Nature’s Frontiers report due to its compatibility with the ecosystem service models and its continuity with a 1992–2015 time-series. Because this dataset only includes land cover (including cropland), but not land use, it does not indicate where on the landscape forestry and grazing are occurring. But these are important agricultural activities impacting ecosystem services, so we employ generalizable methods to estimate where they were likely to occur. For forestry, we use the data published in Lesiv et al. (2022), which uses remote sensing to detect forest management and use. Note that this identifies all forests that are being utilized, not necessarily for commercial use. For grazing, we combine reported pasture meat production statistics with potential productivity modeling to estimate the number of hectares of grazing required, and then distribute this area among eligible land cover classes based on potential economic value. In this application to Tanzania, we do not have spatial data indicating forestry and grazing activities, but we do have national statistics reporting areas for each activity, including cropland. We made the following adjustments to the map produced with the global approach to reflect these data: • Cropland: The estimate of 22 million hectares of total cropland from the ESA map is greater than the 20 million hectares from the agricultural sector report. Some of the ESA categories reflect low- intensity cropping, which is probably not counted in the national statistics. We convert those classes to an appropriate natural cover resulting in a matching 20 million hectares of cropland. • Forestry: The estimate from Lesiv et al. (2022) of approximately 22 million hectares of utilized forest is greater than the national forestry statistic of 5.9 million hectares. This likely reflects the difference between official forestry and informal harvesting and agroforestry. We identify the 5.9 million most profitable hectares to count as forestry and label the other ~16 million hectares as degraded reflecting the stress imposed by harvest. 1 Maps for all constructed scenarios are shown in figure 3. 2 This work (https://www.worldbank.org/en/publication/natures-frontiers presents a methodology that combines innovative science, new data sources, and cutting-edge biophysical and economic models and builds sustainable resource efficiency frontiers to assess how countries can sustainably use their natural capital in more efficient ways. The analysis provides recommendations on how countries can better use their natural capital to achieve their economic and environmental goals. 2  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania • Grazing: The estimated total for grazing based on the production estimate is much larger than the official 3.4 million hectares. Here again, the difference is likely in the use of lands by herders, which could show up as grazing land vs land being designated as pasture. We identify the most potentially profitable 3.4 million to count as pasture and revert the remainder back to the corresponding natural code. After these adjustments are complete, the revised baseline map brings together information from remotely sensed land cover, national statistics, and economic modeling. Finally, we sum the area in each land use/land cover class, then further aggregate these to provide the land area resource base for the macroeconomic modeling used for the CCDR (that is, the MANAGE-WB Computable General Equilibrium (CGE) model,3 hereafter referred to as the CGE model). For the purposes of using land use/land cover in the macro model, we exclude land in protected areas, since protected areas are set aside and not eligible for displacement by agricultural expansion. 2.2 Constructing the BAU 2050 Future Map This map represents the changes under the BAU economy future scenario. Year-by-year land use transitions generated by the CGE model are converted to changes in land use and land cover to produce a time-series of annual land use maps resulting in the final 2050 land use. The macro model specifies, for each year, the expansion or contraction of economic land uses (cropland, grazing, and forestry), and corresponding changes in natural land categories (natural forest, shrubland, and grassland/savannah). To construct the BAU land transition, we allocate these changes incrementally, starting with the baseline map to produce the next year’s map, and then iterating that procedure. In terms of sequencing land use change, for each year, we first allocate additional forestry in order of potential economic value until the annual increase is met. We then allocate pasture and croplands to the remaining natural habitats in order of descending economic value. We repeat this for each year to 2050. In this scenario, protected areas are not converted, but habitat conversion outside of protected areas is allowed. 2.3 Constructing the ASP 2050 future map To construct the ASP 2050 land use map, we assumed that the higher growth and, as a result, revenue available, implies that several specific area-based land use and land cover targets are fully met by 2050. We also assume that agricultural intensification occurs, as it is necessary to meet the increase in agricultural value in the ASP economy future, and that there is greater enforcement of land uses—for example, in forests. Accordingly, the ASP map assumes that Tanzania will: • Restore 3 million hectares of degraded forest and afforest 2.2 million hectares of previously converted forest (in line with AFR 1004) • Increase cropland by 39 percent (7.8 million hectares), as determined in the CCDR Agriculture Sector background note5 • Maintain current grazing area at ~3.4 million hectares • Maintain current forestry area • Neither convert current forest area nor change land use in protected areas. 3 The use and structuring of this model (https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=6845) was discussed with the government of Tanzania’s Planning Commission. 4 https://afr100.org/. 5 This note was developed in close consultation with the CCDR focal person for the Ministry of Agriculture and the Ministry of Livestock and Fisheries. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  3 To do this, we first identify potential afforestation targets. These are pixels where the potential vegetation type is forested, but the current land cover is non-forest. Of these, we select those yielding the greatest possible carbon sequestration potential to fill the quota of 2.2 million hectares. Similarly, for the 3 million hectares of degraded forest, we highlight the pixels identified as degraded that offer the greatest potential carbon sequestration benefits, based on ecoregion and vegetation type. We do this in order of carbon potential so that forest restoration contributes as much as possible towards net greenhouse gas emissions reductions and contributes to meeting the government’s nationally determined contribution (NDC) commitment.6 Next, we allocate cropland to pixels based on potential economic value until the 7.8 million hectares is met, with additional area needed to compensate for any cropland the selected for afforestation (that is, to ensure a net increase of 7.8 million hectares). Finally, we allocate grazing to remaining pixels to try to achieve no net loss of grazing lands.7 However, with the other constraints in place, there is not enough land area to accomplish this, and so grazing land is reduced by roughly 2 million hectares. 2.4 Constructing the BAU with climate action and ASP with climate action future maps These maps are constructed to show the potential contributions of forest restoration and afforestation towards national goals of reducing net greenhouse gas emissions. Based on potential sequestration rates of approximately 10 tonnes of carbon dioxide equivalent (tCO2e) per year, per hectare and available area for afforestation, the maximal possible sequestration from afforestation is somewhere between 70 and 90 million tonnes of carbon dioxide equivalent (MtCO2e) per year. Since the NDC calls for a reduction of at least 138 million tonnes of carbon dioxide equivalent in gross emissions from baseline, and the modeled emissions from other sectors (see SI SE PUEDE, 2024) left a gap between projected and desired emissions larger than 138, we created the BAU with climate action and ASP with climate action scenarios by afforesting all potential forest pixels, which results in an additional 11 million hectares of natural land cover in BAU with climate action and 9 million hectares of natural land cover under ASP with climate action, with an equivalent reduction in agricultural land. 2.5 Allocating irrigation land In the four scenario maps, we perform a second allocation to determine which cropland should be irrigated. The targets are 8 percent of cropland area for the BAU and BAU with climate action scenarios and 25 percent of cropland area for the ASP and ASP with climate action scenarios. The decision is made to maximize economic value—that is, to place irrigation in the areas where it will have the biggest impact on farm value per hectare after accounting for expected changes in crop yields and the cost of installing irrigation. To make these results comparable with the optimization results (section 4) the decision of where to allocate irrigation is made spatially at “parcel” scale. Parcels are spatial units larger than pixels that cover the country but are hexagonal in shape rather than following administrative or ownership boundaries, with hexagons that cross a district bound subdivided into separate portions per district. The logic of parcel optimizations is that it is more computationally efficient and reflects the fact that most the land use decisions are made at scales larger than the land use pixel. For each scenario, we calculate, for each parcel, potential value under rainfed or irrigated yields, assuming future yield trends, and then select the irrigated parcels that maximize cropland value and meet the appropriate area constraint. 6 Tanzania’s NDC commitment is to reduce greenhouse gas emissions economywide by 30–35% relative to the BAU scenario by 2030—this is about 138–153 MtCO2e in gross emissions, depending on the baseline efficiency improvements, consistent with its sustainable development agenda. Forestry is one of the four key sectors for achieving this target. 7 This sequence (urban expansion, afforestation, cropland extensification) implies a certain prioritization between the benefits provided by each the land use and ensures that the maximal available area is afforested. If instead the land changes were sequenced urban, cropland, afforestation, then economic value from cropland would be increased, but less area would be available for afforestation resulting in less overall sequestration. 4  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania 3. Landscape Assessment Models Each landscape is assessed for the following economic and environmental factors: • Agricultural production and net economic value: Production value from cropland, forestry, and grazing; accounting for transportation costs, land share, and transition cost • Carbon: long-term storage; annual sequestration rates • Biodiversity: a combination of different metrics The descriptions below provide information on the methods and data sources used to assess each outcome. We also use these models to parameterize the optimization to create the possibility frontier (see section 4 for a description of this analysis). 3.1 Cropland production value For cropland value, we assess potential production for the 15 most produced crops or categories of crops,8 expected prices for producers, and the costs associated with production and transport. Current yields (tonnes per hectare) for 2018–19 are taken from Ministry of Agriculture (2020), which provides yields for many of these crops at a district level; for other crops, we use the national averages supplied by IEc. To better capture spatial heterogeneity in crop yields, we spatialize these district average values using average crop productivity from the Nature’s Frontiers report, which come from an updated version of Monfreda, Ramankutty, and Foley (2008) and Mueller et al. (2012) that has been developed for the World Bank’s Changing Wealth of Nations report.9 This approach maintains the district level average but reflects localized productivity differences. We estimate future yields using crop-level trends from International Food Policy Research Institute’s International Model for Policy Analysis of Agricultural Commodities and Trade (IMPACT),10 also provided via IEc. These trends reflect historical growth in yields plus modeled effects of additional available capital. Comparison with projections of historical yield data from the Food and Agriculture Organization of the United Nations (FAO) generally align with the IMPACT projections, with larger differences for some specific crops but agreement on average. We do not attempt to model the specific crops grown in a specific landscape parcel. Instead, we apply the district-level average crop mix to each parcel within the district. These mixes are also based on the Ministry of Agriculture (2020) and were provided by IEc. For prices, we assume 2020 prices as reported in the FAO database. Finally, we account for production costs using data from the Global Trade Analysis Project (GTAP) economic database,11 which include national/regional-level statistics on the fraction of producer returns required to cover costs associated with capital, labor, and inputs (Tanzania is represented nationally in this database). We account for transportation costs in a spatially explicit manner (see section 3.4). 3.2 Forestry production value Forestry production12 is calculated by assuming that forests are managed in Faustmann rotation, which specifies the optimal age at which to harvest a stand to maximize economic value. To calculate yield under Faustmann harvest, we estimate forest growth using results from the global dynamic vegetation model MC2 (Kim et al. 2017). The optimal yield from this model is provided zonally for different forest types distributed 8 The modeled crops/crop categories are: banana, beans, cassava, groundnut, maize, millet, potato, rice, sesame, sorghum, sugarcane, sunflower, sweet potato, tropical fruits, and vegetables. These overlap with most of the crops considered in the climate impact modeling. 9 https://www.worldbank.org/en/publication/changing-wealth-of-nations. 10 IMPACT was developed in the early 1990s. The IMPACT model has been expanded and improved repeatedly to respond to increasingly complex policy questions and state-of-the-art of modeling. A network of linked economic, water, and crop models, it is a partial equilibrium multimarket economic model that simulates national and international agricultural markets. The links to water and crop models support the integrated analysis of the changing environmental, biophysical, and socioeconomic trends, allowing for in-depth analysis of a variety of critical issues of interest including at national level. See https://www.ifpri.org/project/ifpri-impact-model/. 11 https://www.gtap.agecon.purdue.edu/databases/default.asp. 12 Measured in megagrams (representing the amount of net primary production) per hectare, per year. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  5 across different ecological regions (Tian et al. 2018). To better capture local variability in productivity, we use net primary production (NPP) as a scaling factor to transform the zonal estimates into pixel-level estimates. For each of the timber zones, we calculate average per-pixel NPP over 15 years, normalize these values so that they average to 1, and use the result as a scaling map. While NPP is an imperfect estimator for forest productivity, it is better than treating potential yields as uniform. Finally, we convert yields to net returns by multiplying timber volume by average producer prices and subtracting costs (see section 3.4). 3.3 Grazing production value We use the value of meat production from beef cattle as an estimate of grazing production value in each pixel. Because grazing takes many forms—from nomadic pastoralism to intensively managed pasture— and data tracking use and intensity are so sparse, we focus specifically on maximum sustainable grazing productivity assuming utilization of natural vegetation. This practice represents an increase in intensity relative to most grazing utilization that lower in environmental impact that pastures managed with fertilized non-native species. We assume that both BAU and ASP scenarios, as well as landscapes on the frontier, manage pastures in this way. We estimate grazing yield at a 10-kilometer resolution following methods used by Castonguay et al. (2023). We model meat production based on biophysical and socio-economic factors affecting production using the ORCHIDEE-GM model to generate a mean annual potential grazed biomass value over a 30-year period (1987–2016) for each 10-square-kilometer grid cell. We assume no fertilizer or irrigation inputs in the production of this biomass, and do not include feedlots and crop feedstocks in the estimation of potential meat production. To estimate highest sustainable grazing yields, we run the model assuming three different grazing intensities, corresponding to the percentage of above-ground biomass consumed by livestock (25, 37.5, or 50 percent). For each pixel, we assign potential yield as the highest result from each of these simulations, resulting in a potential yield map consistent with previous work (such as Fetzel et al. 2017). We convert the modeled grass biomass production to potential meat production accounting for the caloric value of the vegetation, the live weight gain factor (based on Herrero et al. 2013), and the dressing factor (60 percent, based on FAO 2017). Meat production (kilograms per hectare) in each grid cell is multiplied by the price of meat (dollars per kilogram) for each country, using a 10-year average price for cattle meat (2006–15) as reported by FAO producer prices. We calculate potential meat production costs using the method described above in the agricultural crop section and subtract production costs from gross revenue to determine net revenue from meat production. We subtract transport costs (dollars per hectare per kilogram of beef) to production in each pixel based on travel time to the nearest city from the pixel, minimum wage for truck drivers, a fuel efficiency of 0.4 liters per kilometer, the cost of diesel fuel, and assuming that each truck carries 15 tonnes. We also calculate methane emissions measured in terms of CO2 equivalents—kilograms of carbon dioxide equivalent, per hectare (kgCO2e/ha) per kilogram of beef—from each pixel of grazing land as a function of grazed a biomass and a methane emissions conversion factor (kgCO2e/ha, based on Herrero et al. 2013), manure as function of grazed biomass and a manure emissions factor (kgCO2e/ha, also based on Herrero et al. 2013), and transportation (kgCO2e/ha), which depends on travel time to nearest urban center, costs of diesel, average speed (1 kilometer per minute), fuel efficiency (0.4 liters per kilometer), and road emissions factor (2.7 kgCO2e per liter). Methane emissions based on livestock density are aggregated over 20 years and subtracted from carbon emissions. 3.4 Costs We include costs of land use and land use change for several factors: labor, capital, and transportation costs associated with agricultural land uses, and transition costs associated with changes in land use between the current and future uses. To account for labor and capital costs, we use parameters from the GTAP model 6  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania database which estimates the fraction of returns that go to labor, capital, and inputs. These costs come to approximately 87 percent across the different crop types and land uses, leaving approximately 13 percent as the return to land use. We apply these factors as proportional multipliers to the returns after multiplying price by yield. We calculate transportation costs spatially, based on the costs of getting harvested goods from the production pixel to market. We use a time-to-market methodology described in Weiss et al. (2018) which calculates the least-time path from any pixel on the landscape to a designated set of destination pixels as a function of the expected time to cross each intervening pixel. The time to cross a pixel is determined by pixel features, primarily the land cover type, but also including slope and other factors. Costs are estimated based on the number of truck trips required to deliver the agricultural product and corresponding fuel and wage costs. Finally, transition costs are assessed for the restoration of natural habitats, either conversion from an intensive use such as cropland, or for restoration of degraded but still present forest land, or for establishment of new cropland or addition of irrigation infrastructure. These data were provided by the World Bank. The net value of each agricultural activity is calculated by subtracting all relevant costs from the gross revenues. 3.5 Carbon We model each land use and land management alternative for carbon stored in aboveground and belowground carbon pools. The amount of carbon stored in forest, herbaceous vegetation (grass and shrub), and agricultural land cover classes are taken from Spawn et al. (2020), and geographically differentiated based on more than 800 different carbon zones, such that the same land-use land-cover classes contain different carbon densities in different parts of the world. These values are very close on average to the government of Tanzania’s forest reference emissions levels,13 but with the added advantage of better reflecting spatial heterogeneity. To represent changes in carbon storage for the adoption of best management practices in agricultural crop production, a weighted average is taken for the value in that carbon zone for the dominant natural vegetation type (proportional to 10 percent of the area) and agricultural crop production (proportional to 90 percent of the area). Agricultural crop intensification is not assumed to change carbon storage. While poorly managed agriculture may result in degraded soils that store very little carbon, intensification per se does not necessarily mean poor management, and we do not have global information about soil tillage or other management practices that fundamentally affect carbon storage in agriculture. Therefore, we do not attempt to represent trade-offs of intensification with regards to carbon storage. For land managed as forestry, we estimate long-term carbon storage for only one-half of non-harvested forest to reflect the fact that timber is periodically harvested, which reduces above-ground biomass that grows back slowly while trees mature. For grazing lands, we do not attribute a change in carbon to the land use. While it is unlikely that there is no effect of grazing on carbon, there is evidence that well-managed grazing can increase carbon storage, especially in soils, and that poorly managed grazing can reduce it. So, we do not attempt to represent trade-offs of moving from natural (ungrazed) grassland to grazing. But if applying grazing causes a change in land cover—for example, from forest to grazed grassland—we attribute the change in that land cover according to Spawn et al. (2020) as specified above. We evaluate potential sequestration rates based on data from Cook-Patton et al. (2020), which extend the Spawn et al. (2020) product to assess sequestration rates of extant forests as a function of age as well as potential sequestration rates for the first 30 years of afforestation. These results are used to calculate annual sequestration potential for future land scenarios. On the other hand, we model emissions from land clearing as happening in the year the land is cleared. 13 https://www.ncmc.sua.ac.tz/wp-content/uploads/2018/files/FREL/Tanzanias-forest-reference-FINAL.pdf Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  7 3.6 Biodiversity Following the approach used by Damania et al. (2023), we combine six types of biodiversity data—species richness, habitat for threatened and endangered species, habitat for endemic species, habitat for rare species, forest intactness, and key biodiversity areas—which account for different levels of biological organization (species and ecosystems) and the level of threat to that component of biodiversity, such as endangered species or range-restricted species. This allows us to provide information about biodiversity around many of the points raised in the recent Kunming Montreal Global Biodiversity framework, notably Targets 1, 2, 3, 4, 10, 11, 12, 14.14 Species richness The first biodiversity indicator we include is species richness, which is the number of different species represented in an ecological community, landscape or region. Species richness remains by a wide margin the most identifiable and relatable component of biodiversity to policy- and decision-makers. We use the Projecting Responses of Ecological Diversity In Changing Terrestrial Systems (PREDICTS) database (Hudson et al. 2014; Newbold et al. 2015) to address how changes in land use and land cover affect species richness. We first calculate regional species pools for every land pixel in every country, using the International Union for Conservation of Nature (IUCN) range maps for amphibians, birds, mammals, and reptiles (IUCN 2019). We then modify regional species pools by the local land cover in each pixel, based on the ESA land cover data (ESA 2019) and the PREDICTS database, based on three factors: • General land use type (natural, plantation forest, cropland, pasture, or urban area) • Level of human intensity (minimal, light, heavy) • Age of habitat for natural habitat type (young, intermediate, mature, primary). The general land use type (Type 1) is easily comparable across datasets. We assume that: all truly natural habitats have light land use intensity; all current row-crop agriculture are of moderate intensity; all current pasture and forestry in natural forests are of moderate intensity; and plantation forestry is high-intensity. In our future scenarios, pixels can be converted to more or less intensively managed row crop agriculture. Future timber activities are assigned to either moderate intensity, if in naturally occurring forests, or high intensity, if in plantation forestry. The age of each natural habitat is determined by aligning the pixels with their historical ESA land cover in the years 2005 and 1991. If the pixel is currently in natural habitat but was not in natural habitat in 2005, we assume it is young secondary vegetation. If the pixel is currently in natural habitat, and was in 2005 but not 1991, then we assume it is intermediate secondary vegetation. If it is and was in natural vegetation in all three classes, we assume it is primary vegetation, as mature secondary vegetation scores higher than the primary reference habitat. Any future pixels converted to natural habitat are assumed to be young natural habitats. This approach provides for each scenario, a value of total species richness for every taxon in each pixel. We then weigh these by first dividing by the total number of species globally for each species group and then sum over all four taxa. These values are then normalized within each country between 0 and 1 using min- max normalization such that a higher value indicates a higher value for species richness. Other metrics For the other metrics, we calculate weighted scores that combine a potential biodiversity value (habitat for threatened and endangered species, habitat for endemic species, habitat for rare species, forest intactness, and key biodiversity areas) and the impact of each potential land use type. The potential values are assembled from a range of globally available sources. 14 https://www.cbd.int/article/cop15-cbd-press-release-final-19dec2022. 8  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania 4. Developing the Optimization Frontier The possibility frontier is constructed by generating many individual optimized landscapes, each of which vary in priorities from exclusively maximizing ecosystem services to exclusively maximizing economic production. Between those extremes, each optimization puts a different weight on each objective so that we end up with landscapes distributed along the trade-off curve between the environmental and economic outcomes. These landscapes outline the best possible combinations of environmental and economic benefits. We use an optimization analysis to construct the efficiency frontier, which shows the range of Pareto-efficient15 combinations of sustainable environmental and economic objectives that can be attained through different land-use and land-management choices. The optimization analysis uses a discrete choice model where the country is divided up into parcels and, for a given combination of weights on the objectives, a management decision is made for each parcel yielding a national-scale land use configuration. The parcels are defined as larger spatial units that contain multiple smaller land units (“pixels”), which are defined by the resolution of spatially explicit data, which in our analysis match the ESA land cover map at 300 x 300 meters. Some pixels are precluded from certain land uses or management options. For example, agriculture is excluded from some areas because the land is unsuitable for crop production due to poor soil or steep slopes, or because the land is in a protected area. Each pixel is given a score for each objective under each land use and land management alternative according to the models for climate mitigation, biodiversity, agricultural crop production, livestock grazing production, forestry production, and transition costs. Each parcel’s score for an objective is the sum of its constituent pixels’ values. The optimization procedure is as follows: For each parcel, we calculate the value of the primary objectives by summing pixel-level values for their component quantities: • Net economic value: Sum of cropland, grazing and forestry value minus relevant costs, including restoration costs • Climate mitigation: Sum of carbon dioxide equivalent (CO2e) for carbon storage in above and below ground biomass and 20 years of methane emissions from grazing animals • Biodiversity: The biodiversity index (section 3.6). Then, to generate the frontier, we run a series of optimizations in which we assign a randomly generated weight to each primary objective, calculate the weighted sum of the objectives for each activity in each parcel, and then for each parcel, select the activity with the greatest summed objective value. Formally, this is the same as the optimization. Where p ranges over parcels, m over possible management options, and i over the different objectives; xpm is a binary variable indicating that parcel p is assigned to management m, wi is the weight assigned to objective i, and vipm is the value to objective i obtained by assigning management m to parcel p. This produces a set of landscapes that have been optimized favoring different combinations of the three main objectives. Finally, for each optimized landscape, we calculate the national-level scores for each of the primary objectives (net economic value, climate mitigation, and biodiversity) and their component metrics— 15 Pareto-efficient refers to a combination of outcomes that can only be improved in one dimension if there is a loss in another. These points are “on the frontier”, as opposed to “inside the frontier”, where it is possible to improve in all dimensions together. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  9 crop yields and value, grazing value, and so on—based on the selected activities. This calculation is used to generate the frontier and identify the Pareto-improving landscapes among the solution set. The land management scenarios included in the optimization are: • Restoration: converted land restored to natural habitat or degraded land restored to higher quality • Cropland expansion with current practices • Cropland expansion with intensification • Cropland expansion with conservation agriculture practices • Forestry expansion • Pasture expansion. Each of these scenarios is implemented as a parcel-by-parcel land transformation scenario, which is then scored for each of the optimization objectives (economic value and carbon) using the models and methods described in section 2.2. These results give the potential outcomes for each parcel under each potential management option (that is, the vipm terms in the equation above). Certain spatial constraints, such as non-conversion of land in protected areas, are enforced “pre-optimization” in the construction of the pixel-level maps that correspond with each potential management scenario. For example, in such a case, to evaluate the impact of cropland extensification in a parcel that was partially protected and partially not, we would only convert the unprotected portion, leaving the protected pixels unchanged. Similarly, the transformations do not simply change all land use pixels from their current state to the selected land use; pixel-level characteristics are used to determine whether that pixel is eligible for the transition. The constraints that apply are: • Restoration: all pixels can be restored, but pixels are restored to their modeled potential vegetation type. This means, for example, we do not allow afforestation outside of areas that are in forest-type ecoregions. • Crop expansion: – Economic viability: the pixel must have a net value greater than $0 after accounting for all costs – Biophysical suitability: slope, soil, and so on – Protected areas: no expansion into protected areas – Conservation practices: pixels adjacent to streams are restored to create riparian buffers. • Grazing expansion: – Economic viability – Protected areas. • Forestry expansion: – Must be a potentially forested pixel – Economic viability – Protected areas. Finally, after we identify a specific point on the frontier, these scenario rasters are used to construct the corresponding national-level land cover map by stitching together the land cover tiles associated with each parcel. 10  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania 5. Results 5.1 Land scenario comparisons Focusing on two primary objectives, net economic value from land-based agriculture and carbon storage in biomass, all four of the land scenarios (BAU, BAU with climate action, ASP, and ASP with climate action) result in improvements in economic value and losses in biomass carbon storage. The scenarios without climate action not only give up around 1,000 million tonnes more than those with climate action, but also produce significantly more economic value. The reduction in value for the scenarios with climate action comes from both reduction in agricultural land and from the cost of carrying out the restoration activities outlined in the NDC. While all scenarios move toward the frontier, there remains a gap highlighting additional opportunities to improve economic and environmental outcomes. Figure 1: Frontier and national totals for current and constructed land use maps 14 13 Carbon storage (tCO2e, billions) 12 11 10 9 8 7 6 0 5 10 15 20 Net economic value ($, billions) Frontier Current BAU BAU with climate action ASP ASP with climate action Optimized Notes: Horizontal axis depicts the net economic value attributable to agricultural land uses (cropland, grazing, and forestry) less transition costs, and vertical axis depicts total carbon storage in above and below ground biomass (expressed as CO2e). The frontier extends below zero on the horizontal axis because reaching such a high level of carbon storage requires significant restoration efforts, which have a cost in the model. Figure 2 shows the total area by land use category across the six scenarios depicted in figure 3. One thing that stands out is the similarity between them; this is due to the large expanse of protected areas in Tanzania and because much of the non-protected area is already dedicated to cropland use, so the differences between scenarios are restricted to a small portion of the country’s area. The primary difference between the optimized and BAU/ASP scenarios is that the optimized scenario focuses on increasing agricultural production by the addition of irrigation infrastructure, while the BAU and ASP scenarios focus more on extensification. Because the optimized scenario mostly maintains the current cropland footprint, little carbon is lost from existing natural habitats. The feasibility of the irrigation-focused pathway depends on several factors, including the country’s ability to roll out irrigation infrastructure at a rate exceeding even the ASP scenario’s target, and an assessment of whether water resources could actually sustain the targeted irrigation use. The optimization scenario accounts for irrigation sustainability based on an assessment by Rosa et al. (2019), but this was a global study and not focused specifically on water use and hydrology in Tanzania. An advantage of pursuing this approach is that it requires little or no clearing of natural habitats, while the extensification-focused scenarios could lead to losses in biomass carbon and impacts on biodiversity. Even with the significant afforestation Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  11 under the NDC scenarios, this only brings long-term carbon storage back to the current level. Focusing instead on intensifying current cropland is neutral with respect to carbon storage and biodiversity while providing opportunities to increase economic productivity. Figure 2: Land area by aggregated land use classes for 2050 100 80 Hectares (millions) 60 40 20 0 Current BAU BAU with climate action ASP ASP with climate action Optimized Irrigated cropland Rainfed cropland Grazing Forestry Non-forest natural Forest Table 1. Key subsector values and ecosystem service characteristics of each scenario Scenario Carbon Transition Cropland Forestry Grazing Gross Net Carbon Biodiversity Carbon sequestration cost value value value aggregate economic storage loss CO2e/year value value (CO2e) (CO2e) Current 0 0 2,480 240 101 2,822 2,822 10,725 7 0 BAU 0 283 7,895 352 130 4,100 8,095 9,750 6 272 BAU w/ climate 75 967 7,185 156 115 3,587 6,488 10,584 7 212 action ASP 7 111 8,725 240 21 4,101 8,875 9,963 7 246 ASP w/ climate 80 1,405 7,811 123 20 3,609 6,550 10,628 7 226 action 12  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania Figure 3: Maps of each of the constructed land use scenarios Current Optimized BAU ASP BAU with climate action ASP with climate action Irrigated cropland Rainfed cropland Grazing Forestry Non-forest natural Forest Water Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  13 References Castonguay, A C, Polasky, S, Holden, M H, Herrero, M, Mason-D’Croz, D, Godde, C Chang, J, Gerber, J, Witt, G B, Game, E T, Bryan, B A, Wintle, B, Lee, K, Bal, P and McDonald-Madden, E. 2023. “Navigating Sustainability Trade-Offs in Global Beef Production.” Nature Sustainability 6 (3): 284–94. Cook-Patton, S C, Leavitt, S M, Gibbs, D, Harris, N L, Lister, K, Anderson-Teixeira, K J, Briggs, R D, et al. 2020. “Mapping Carbon Accumulation Potential from Global Natural Forest Regrowth.” Nature 585 (7826): 545–50. Damania, R, Polasky, S, Ruckelshaus, M, Russ, J, Amann, M, Chaplin-Kramer, R, Gerber, J, Hawthorne, P, Heger, M P, Mamun, S, Ruta, G, Schmitt, R, Smith, J, Vogl, A, Wagner, F and Zaveri, E. 2023. Nature’s Frontiers: Achieving Sustainability, Efficiency, and Prosperity with Natural Capital. Washington DC: World Bank. http://hdl.handle.net/10986/39453. ESA. 2019. ESA CCI Land Cover. European Space Agency. https://www.esa-landcover-cci.org/. FAO. 2017. The Global Livestock Environmental Assessment Model: Model Description Version 2.0. https://www.fao. org/fileadmin/user_upload/gleam/docs/GLEAM_2.0_Model_description.pdf. Fetzel, T, Havlik, P, Herrero, M, Kaplan, J O, Kastner, T, Kroisleitner, C, Rolinski, S, Searchinger, T, Van Bodegom, P M, Wirsenius, S and Erb, K-H. 2017. “Quantification of Uncertainties in Global Grazing Systems Assessment.” Global Biogeochemical Cycles 31: 1089–1102. Herrero, M, Havlík, P, Valin, H, Notenbaert, A, Rufino, M C, Thornton, P K, Blümmel, M, Weiss, F, Grace, D and Obersteiner, M. 2013. “Biomass Use, Production, Feed Efficiencies, and Greenhouse Gas Emissions from Global Livestock Systems.” Proceedings of the National Academy of Sciences of the United States of America 110 (52): 20888–93. Hudson, L N, Newbold, T, Contu, S, Hill, S L L, Lysenko, I, De Palma, A, Phillips, H R P et al. 2014. “The PREDICTS Database: A Global Database of How Local Terrestrial Biodiversity Responds to Human Impacts.” Ecology and Evolution 4 (24): 4701–35. IUCN. 2019. The IUCN Red List of Threatened Species. https://www.iucnredlist.org/. Kim, J B, Monier, E, Sohngen, B, Pitts, G S, Drapek, R, McFarland, J, Ohrel, S and Cole, J. 2017. “Assessing Climate Change Impacts, Benefits of Mitigation, and Uncertainties on Major Global Forest Regions under Multiple Socioeconomic and Emissions Scenarios.” Environmental Research Letters 12 (4): 045001. Lesiv, M, Schepaschenko, D, Buchhorn, M, See, L, Dürauer, M, Georgieva, I, Jung, M et al. 2022. “Global Forest Management Data for 2015 at a 100 m Resolution.” Scientific Data 9 (1): 199. Ministry of Agriculture. 2020. Basic Data: Crop Sub Sector (Tanzania Mainland) 2018/2019. https://kilimokwanza.org/ wp-content/uploads/2021/01/Basic_Data_Booklet_2020Sep_1-1.pdf. Monfreda, C, Ramankutty, N and Foley, J A. 2008. “Farming the Planet: 2. Geographic Distribution of Crop Areas, Yields, Physiological Types, and Net Primary Production in the Year 2000.” Global Biogeochemical Cycles 22 (1). https://doi. org/10.1029/2007GB002947. Mueller, N D, Gerber, J S, Johnston, M, Ray, D K, Ramankutty, N and Foley, J A. 2012. “Closing Yield Gaps through Nutrient and Water Management.” Nature 490 (7419): 254–57. Newbold, T, Hudson, L N, Hill, S L L, Contu, S, Lysenko, I, Senior, R A, Börger, L et al. 2015. “Global Effects of Land Use on Local Terrestrial Biodiversity.” Nature 520 (7545): 45–50. Rosa, L, Chiarelli, D D, Tu, C, Rulli, M C and D’Odorico, P. 2019. “Global Unsustainable Virtual Water Flows in Agricultural Trade.” Environmental Research Letters 14 (11): 114001. SISEPUEDE, 2024. GHG emissions pathways. Technical note for URT Country Climate and Development Report. Washington DC: World Bank. 14  |  Background Paper for Country Climate and Development Report: United Republic of Tanzania Spawn, S A, Sullivan, C C, Lark, T J and Gibbs, H K. 2020. “Harmonized Global Maps of above and Belowground Biomass Carbon Density in the Year 2010.” Scientific Data 7 (1): 112. Tian, X, Sohngen, B, Baker, J, Ohrel, S and Fawcett, A A. 2018. “Will U.S. Forests Continue to Be a Carbon Sink?” Land Economics 94 (1): 97–113. Weiss, D J, Nelson, A, Gibson, H S, Temperley, W, Peedell, S, Lieber, A, Hancher, M et al. 2018. “A Global Map of Travel Time to Cities to Assess Inequalities in Accessibility in 2015.” Nature 553 (7688): 333–36. Background Paper for Country Climate and Development Report: United Republic of Tanzania  |  15 World Bank Tanzania 50 Mirambo Street P. O. Box 2054 Dar es Salaam, Tanzania Read more on linkages Tel : +255-22-216-3200 between climate and tanzaniaalert@worldbank.org development in Tanzania