Skip to main content
Dryad

A practice-led assessment of landscape restoration potential in a biodiversity hotspot

Cite this dataset

Wills, Abigail et al. (2022). A practice-led assessment of landscape restoration potential in a biodiversity hotspot [Dataset]. Dryad. https://doi.org/10.5061/dryad.nk98sf7wr

Abstract

Effective restoration planning tools are needed to mitigate global carbon and biodiversity crises. Published spatial assessments of restoration potential are often at large scales or coarse resolutions inappropriate for local action. Using a Tanzanian case study, we introduce a systematic approach to inform landscape restoration planning, estimating spatial variation in cost-effectiveness, based on restoration method, logistics, biomass modelling and uncertainty mapping. We found potential for biomass recovery across 77.7% of a 53,000 km2 region, but with some natural spatial discontinuity in moist forest biomass, that was previously assigned to human causes. Most areas with biomass deficit (80.5%) were restorable through passive or assisted natural regeneration. However, cumulative biomass gains from planting outweighed initially high implementation costs meaning that, where applicable, this method yielded greater long-term returns on investment. Accounting for ecological, funding and other uncertainty, the top 25% consistently cost-effective sites were within protected areas and/or moderately degraded moist forest and savanna. Agro-ecological mosaics had high biomass deficit but little cost-effective restoration potential. Socio-economic research will be needed to inform action towards environmental and human development goals in these areas. Our results highlight value in long-term landscape restoration investments and separate treatment of savannas and forests. Furthermore, they contradict previously asserted low restoration potential in East Africa, emphasising the importance of our regional approach for identifying restoration opportunities across the tropics.

Methods

Here, we develop and apply a systematic approach to inform spatially-explicit forest landscape restoration planning. Our approach prioritizes cost-effective ecosystem recovery for timely achievement of global and regional restoration targets, accounting for biomass accumulation (and thus carbon sequestration and storage) objectives in a strategic region in Tanzania. The approach can be applied to any landscape-scale restoration project, using spatial prioritisation methods for more detailed planning than is possible with existing restoration decision support tools. It is based on direct financial implementation costs of the most appropriate methods for restoring native vegetation and associated biomass, biodiversity, ecological function and livelihood options under different scenarios and investment timeframes. In achieving this, unlike previous studies, we account for direct implementation and community engagement costs, logistics, expected vegetation growth, and estimated uncertainty resulting from incomplete ecological knowledge. The findings are intended to be useful for advancing the science of restoration planning and for inspiring donors, through development of metrics directly useful for attracting and prioritising grant funding.

Our approach comprised four steps to determine the cost-effectiveness of ecological landscape restoration per hectare, using a 5.3 million hectare region in Tanzania as a case study. First, we estimated landscape above-ground biomass (AGB) deficit (and thus restoration potential) from current and maximum potential AGB, predicted by up-scaling AGB measurements in 195 plots using spectral-reflectance and climatic predictors, respectively. Secondly, we assigned the most appropriate silvicultural method for restoring AGB, i.e. passive regeneration, assisted natural regeneration (ANR), or planting native vegetation, to each hectare pixel with restoration potential, basing this on key landscape variables, including: vegetation type, elevation, severity of degradation (using AGB deficit as a proxy), and distance from proximal intact habitat. Thirdly, we modelled the expected rate of AGB gain through application of these methods and number of years to full-recovery of the AGB deficit. This was achieved using a regional dataset comprising cumulative modelled annual estimates of above-ground carbon (AGC, from zero to maximum) in naturally-regenerating African forests, generated based on vegetation plot AGB measurement over time (Bernal et al. 2018). Finally, we estimated the financial implementation costs of restoration, and therein cost-effectiveness (AGB gain per dollar spent), per hectare, accounting for land procurement, labour equipment and transport, community engagement, project management and administrative costs.

We incorporated pessimistic, realistic, and optimistic scenarios into all four stages and estimated AGB gains, implementation costs and cost-effectiveness over two investment timeframes: (1) five years, to represent a typical upper limit of donor investment; and (2) expected time to full AGB recovery. A combination of expert knowledge, pilot data and literature review were used to determine: (a) environmental degradation thresholds for selecting methods; and (b) comprehensive implementation costs.

Spatial variations in restoration potential, AGB gain, cost and cost-effectiveness were evaluated retrospectively in terms of technical implementation and landscape features of use to practitioners, namely: (a) restoration method; (b) landcover class; and (c) governance (protected versus unprotected areas). Means with standard deviations were used to summarise estimates of landscape AGB, which followed a broadly normal distribution, whereas costs and cost-effectiveness were described using medians and inter-quartile ranges. Cost-effectiveness of different methods, landcover and governance types was compared using Kruskal-Wallis tests with Dunn posthoc tests and Holm-adjusted P-values, Padj (Dunn, 1964). We also identified the top 25% most cost-effective sites for restoration across all scenarios (pessimistic, realistic and optimistic) and investment timeframes (five years and to full recovery of AGB deficit) combined. We did this both overall within the landscape and specifically outside protected areas, to identify locations with potential for community restoration schemes. To account for uncertainty in our estimates, we reported restoration potential before and after discounting areas with high “ecological uncertainty” from climate modelling (defined using envelope uncertainty maps, EUMs, following Platts et al. 2008).

Usage notes

All statistical and spatial analyses were conducted using R version 4.0.1 (R Core Team 2020), besides the distance matrices and maps, which were produced in ArcGIS Pro version 2.7.1 (ESRI Inc. 2020). The caret package was used for modelling (Kuhn 2008) and the raster package for spatial up-scaling (Hijmans and van Etten 2012). Statistical analyses were performed using the R base package. Both our R script (https://bit.ly/3KO5Hgz) and all input and output maps (https://bit.ly/3QkIrrI) produced during our stepwise method are available online.

Funding

IUCN Sustain

African Wildlife Foundation

United Bank of Carbon

Australian Research Council, Award: FT170100279

Rainforest Trust

Flamingo Land Limited