Deforestation in an African biodiversity hotspot: Extent, variation and the effectiveness of protected areas

Highlights

• Generalised additive models built to describe forest and woodland loss in the Eastern Arc Mountains of Tanzania.

• Rates of evergreen forest loss, at −0.2% per year, are lower than rates of miombo woodland loss, at −2.3% per year.

• Protected areas are partially effective at reducing forest (40% decrease) and woodland (16% decrease) loss.

• The majority of remaining forest lies within protected areas, while most remaining woodland lies outside protected areas.

• More resources are needed to reduce the loss of forest and, in particular, woodland, both valuable for ecosystem services.

Abstract

The Eastern Arc Mountains of Tanzania show exceptional endemism that is threatened by high anthropogenic pressure leading to the loss of natural habitat. Using a novel habitat conversion model, we present a spatially explicit analysis of the predictors of forest and woodland conversion in the Eastern Arc over 25 years. Our results show that 5% (210 km²) of evergreen forest and 43% (2060 km²) of miombo woodland was lost in the Eastern Arc Mountains between 1975 and 2000. Important predictors of habitat conversion included distance to natural habitat edge, topography and measures of remoteness. The main conservation strategy in these mountains for the past 100 years has been to develop a network of protected areas. These appear to have reduced rates of habitat loss and most remaining evergreen forest is now within protected areas. However, the majority of miombo woodland, an important source of ecosystem services, lies outside formal protected areas, where additional conservation strategies may be needed.

Introduction

Forest area in Africa decreased by an estimated 34–41 thousand km² per year during the 1990s and 2000s (FAO, 2010). Tropical moist forests are amongst the most species-rich terrestrial habitats on Earth, making deforestation a crucial issue for biodiversity conservation (Joppa and Pfaff, 2011, Pimm et al., 2001). Furthermore, they sequester and store large amounts of carbon; tropical deforestation is estimated to contribute 7–14% of global carbon dioxide emissions, resulting in accelerated climate change (Harris et al., 2012, IPCC, 2007). Given forests’ crucial role in conserving biodiversity and mitigating climate change, as well as in sustaining local livelihoods, understanding the drivers and threat mechanisms of forest conversion and finding ways to reduce rates of loss is high on the conservation agenda (Balmford et al., 2009).

An increasing research focus on the rates and predictors of habitat conversion has been aided by the advent of satellite remote sensing technologies (DeFries and Townshend, 1999). Regression models have been widely used to identify predictors of habitat conversion and previous studies have investigated the impact of market access (e.g. distance to roads and population centres) and topography (e.g. slope and elevation) at study scales ranging from sub-national (Patarasuk and Fik, 2013, Serneels and Lambin, 2001, Vuohelainen et al., 2012) to regional (Pfeifer et al., 2012, Portillo-Quintero et al., 2012) and global (Scrieciu, 2007). These models have been used to elucidate threat mechanisms of habitat conversion and to investigate the effectiveness of conservation efforts to prevent it. Predictive models have more recently been proposed as a tool for protected area planning: by assessing threats spatially and designating protected areas accordingly the potential impact of reserves can be maximised (Joppa and Pfaff, 2009). Predictive models can also help in conservation resource allocation by suggesting different levels of investment required in different parts of a protected area network (Andam et al., 2008). Lastly, an international scheme to decrease the amount of carbon released into the atmosphere (Reduced Emissions from Deforestation and Degradation; REDD) requires an assessment of baseline (current) levels of habitat conversion and the future trajectories of loss to be estimated (Brown et al., 2007).

Much of our current knowledge of the causes and mechanisms of habitat loss is based upon research undertaken in Latin America and Asia, where deforestation is driven by agricultural exports and urban population growth, largely mediated through agricultural expansion, infrastructure development and resource extraction (DeFries et al., 2010, Geist and Lambin, 2002). However, the drivers and extent of deforestation have been found to vary between continents and Africa, in particular, has had few studies devoted to spatial modelling of habitat loss (Achard et al., 2002, DeFries et al., 2010, Fisher, 2010, Geist and Lambin, 2002, Pfeifer et al., 2012). We focus on the Eastern Arc Mountains in Tanzania where, although anthropogenic pressures are high, they vary across the landscape (Brooks et al., 2002, Burgess et al., 2006). Our objectives are to develop the first spatially explicit, high-resolution model of past evergreen forest and miombo woodland change in the Eastern Arc Mountains, based upon potential predictors of habitat loss and retention. We then use this model to predict likely future changes and to consider how habitat loss varies between protected and non-protected areas.