Elsevier

Ecological Indicators

Volume 109, February 2020, 105763
Ecological Indicators

Original Articles
Repeatable and standardised monitoring of threats to Key Biodiversity Areas in Africa using Google Earth Engine

https://doi.org/10.1016/j.ecolind.2019.105763Get rights and content

Highlights

  • Demonstrates potential of Google Earth Engine for remote monitoring of KBAs.

  • Easily repeatable methods with shared code adaptable to other datasets and regions.

  • Fire frequency increased on 12.4% of African KBAs and 15.9% of ecoregions.

  • Rates of forest loss increased on 24.3% of African KBAs and 22.6% of ecoregions.

  • Night-time lights increased on 53.3% of African KBAs and 39.6% of ecoregions.

Abstract

Key Biodiversity Areas (KBAs) are sites that make significant contributions to the global persistence of biodiversity, but identification of sites alone is not sufficient to ensure their conservation. Monitoring is essential if pressures on these sites are to be identified, priorities set and appropriate responses developed. Here, we describe how analysis of freely available data on a cloud-processing platform (Google Earth Engine) can be used to assess changes in three example remotely sensed threat indicators (fire frequency, tree loss and night-time lights) over time on KBAs in Africa. We develop easily repeatable methods with shared code that could be applied across any geographic area and could be adapted and applied to other datasets as they become available. Fire frequency was found to have increased significantly on 12.4% of KBAs and 15.9% of ecoregions, whilst rates of forest loss increased significantly on 24.3% of KBAs and 22.6% of ecoregions. There was also evidence of significant increases in night-time lights on over half (53.3%) of KBAs and 39.6% of ecoregions between 1992 and 2013, and on 11.6% of KBAs and 53.0% of ecoregions between 2014 and 2018.

Introduction

Key Biodiversity Areas (KBAs) are sites that make significant contributions to the global persistence of biodiversity (IUCN, 2016). They are identified using globally standardised criteria, with sites qualifying as global KBAs if they meet one or more of 11 criteria, clustered into five categories: threatened biodiversity; geographically restricted biodiversity; ecological integrity; biological processes; and irreplaceability (IUCN, 2016).

The implementation of criteria to identify sites qualifying as KBAs is ongoing (the global standard for KBA identification was agreed in October 2016), but over 15,000 KBAs have already been identified globally using previous criteria. Identification of a KBA does not in itself afford any form of legal protection and many sites remain unprotected (Butchart et al., 2012, Waliczky et al., 2018). Therefore, monitoring is essential if threats to biodiversity at KBAs are to be identified, and appropriate responses developed (Jones et al., 2013).

Monitoring of sites of conservation importance such as KBAs may take place locally, with field-based surveys or assessments of change. Such data are most useful if they are collected according to a standardised protocol, therefore allowing comparisons to be made over time and between sites. To this end, multiple field-based monitoring protocols exist. One of the most widely used is the Management Effectiveness Tracking Tool (METT), for protected area monitoring (Stolton and Dudley, 2016). Additionally, BirdLife International developed a simple, globally standardised protocol for monitoring Important Bird and Biodiversity Areas (IBAs; BirdLife International, 2006), which as sites of global significance for bird conservation, represent a high proportion of identified KBAs. Although there has been reasonable uptake of the IBA monitoring protocol, with over 25% of IBAs globally having been assessed at least once by 2012 (Buchanan et al., 2013), the majority of IBAs remain unmonitored using this framework. While estimates of the number of protected areas that have some form of monitoring data exist (Schulze et al., 2017), the degree to which KBAs are monitored is unknown, not least because their identification is still ongoing.

However, capacity to undertake such on-the-ground monitoring is limited, particularly in the poorest countries, which often have the richest biodiversity and most severe threats (Buchanan et al., 2013). In addition to field-based monitoring, it is now well established that remote sensing has the potential to be a valuable tool in assessing environmental changes in sites of conservation importance (e.g. Turner et al., 2003, Buchanan et al., 2009a, Rose et al., 2015). While not a panacea for monitoring, use of remote sensing data can compliment traditional methods of data collection in the field in a number of ways. It covers extensive geographic areas, including remote, inaccessible and dangerous regions; it allows repeated measures at short intervals and over long time-spans; it allows the user to go back in time through an archive of historic data; it can be highly centralised, allowing a single user to monitor multiple distant sites; and it can be very cheap, in terms of both time and money, relative to fieldwork. Of course, there are also limitations. Not all types of change can be detected using remote sensing, with small-scale changes and changes occurring under the canopy being particularly difficult to detect (Schulze et al., 2017).

Over the last decade, the availability of free satellite data coupled with advances in computing power, including cloud computing, have greatly increased the practicality of using remote sensing to undertake large-scale, long-term analyses (Leidner and Buchanan, 2018). At the same time, satellite technology has also been advancing, producing images with increasingly high spatial and temporal resolution that can be used in biodiversity conservation (e.g. Brink et al., 2018). In addition, the availability of regularly updated and synthesised global remote sensing products, e.g. Hansen’s Global Forest Change map (Hansen et al., 2013), represents another advancement, with such derived products being much more user-friendly and requiring less computing power and technical skill than manipulation of raw data.

Together, these developments mean that there is now real potential to move from one-off assessments towards developing a system of near real-time monitoring of sites. Although not all threats to sites can be quantified remotely (Schulze et al., 2017), the most prevalent threats to IBAs in Africa, namely deforestation and agricultural expansion, involve changes in land cover that can be detected from satellites (Buchanan et al., 2009b). Similarly, a recent global study of the pressures on protected areas also found that in Africa, the most prevalent threats were agriculture, logging and fire and fire suppression (Schulze et al., 2017), all of which could potentially be measured remotely.

Tools and methods are available for detailed monitoring using remote sensing on a site by site basis, and these have been applied to look at changes on multiple sites of conservation value in Africa (e.g. Bastin et al., 2013, Beresford et al., 2013). While these have the potential to deliver high spatial resolution information about individual KBAs, it would require a considerable investment of time and resources to apply them at a continental scale (albeit still much less time than field-based surveys).

There are, however, some readily available global datasets that can be linked to IUCN level 1 or level 2 threats (Salafsky et al., 2008), are intuitive to interpret in terms of site management, and which could be used to make rapid assessments of some of the most prevalent pressures on KBAs across Africa, or indeed the globe. Here, we present three such examples for tracking changes in fire frequency, forest loss and night-time lights. These indicators by no means provide a comprehensive assessment of the overall pressure on sites, but demonstrate how remotely-sensed variables could add to the information available to stakeholders, for example to inform management decisions or priority setting.

Monitoring of fires, to identify changes in natural patterns of burning, has been underway for many years (e.g. Butler, 2008, Davies et al., 2009). Fire data have been supplied at regional and national levels as well as to individual protected areas through the internet (https://firms.modaps.eosdis.nasa.gov) and satellite broadcasts. Fire alerts received in near real time have been used to identify problem fires, and to track the effectiveness of management plans and prescribed burning to manage vegetation (Palumbo et al., 2018). The availability of these data to all users through the Fire Information for Resource Management System (FIRMS; Davies et al., 2009) means that it would be relatively easy to track patterns of burning across a network of sites.

Tree loss on IBAs has already been assessed using remote sensing data (Buchanan et al., 2013, Tracewski et al., 2016). Tracewski et al. (2016) provides a good example of how derived products can be used to undertake assessments of change that can be rapidly updated. Importantly, their analysis also incorporated the opinions of conservationists within country to ‘validate’ the results derived from remote sensing data. This national level knowledge enabled the filtering out of results that were of no conservation concern (e.g. the removal of a commercial plantation from a wetland site in order to improve the wetland habitat).

Several studies have used night-time lights (anthropogenic light pollution in the form of street lights, car headlights and other urban and domestic lighting) to map urban areas (e.g. Small et al., 2005, Zhou et al., 2014, Zhou et al., 2015), to characterize urbanisation dynamics (e.g. Zhang and Seto, 2011, Liu et al., 2012, Ma et al., 2012, Li et al., 2016) and to study the ecological consequences of light pollution (Gaston et al., 2012, Mazor et al., 2013). To date though, there have been no large-scale studies that have looked at nightlights as an indicator of infrastructural intrusion into a network of sites of known significance for biodiversity.

As a contribution to developing a global remote monitoring protocol for sites of conservation importance, including KBAs, we make use of open access data on a free cloud-processing platform (Google Earth Engine) to assess changes in three remotely sensed threat indicators (fire frequency, tree loss and night-time lights) over time on KBAs in Africa. We describe patterns of change and identify sites and ecoregions where the greatest changes have taken place. We thereby develop easily repeatable methods that could be applied across any geographic area or network of sites, and could be adapted and applied by conservation end-users to other datasets as they become available.

Section snippets

Data extraction

Analysis was carried out using Google Earth Engine, an online cloud processing environment for spatial data (https://earthengine.google.com/). Commands were submitted in JavaScript with code available at https://github.com/AlisonBeresford/KBA-Monitoring.git.

Digitised boundaries of KBAs were obtained from BirdLife International (2018). We selected all KBAs for which ‘region’ was listed as Africa: a total of 1784 sites. Boundaries were converted into Keyhole Markup Language (kml) format and

Comparison of remotely-sensed threat indicators with field-based monitoring scores

There were significant positive correlations between the metrics of fire frequency, forest loss and stable night-time lights derived from remote sensing data and the corresponding pressure scores recorded in field-based assessments (Table 1). However, no significant relationship was found between average radiance in 2018 and the corresponding pressure score for residential and commercial development. Country explained a significant amount of variance in all of the models (Table 1).

Trends in remotely-sensed threat indicators

Across all

Discussion

Having examined patterns of change in indicators of three different threats to KBAs across Africa, our results suggest a complex pattern of pressures acting to different extents on different sites and ecoregions across the continent. Overall, fire frequency was found to have increased significantly on 12.4% of KBAs and 15.9% of ecoregions, whilst rates of forest loss increased on 24.3% of KBAs and 22.6% of ecoregions. On average, KBAs containing forest in the year 2000 had lost 5.80% of their

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper

Acknowledgements

We are grateful to Stuart Butchart for his helpful advice and comments on this manuscript, and to Łucasz Tracewski for his technical help and advice on Javascript coding for Google Earth Engine.

Data Sharing

All satellite remote sensing data used are freely available on Google Earth Engine (https://earthengine.google.com/). KBA boundaries can be requested from http://www.keybiodiversityareas.org/site/requestgis. JavaScript code is available at https://github.com/AlisonBeresford/KBA-Monitoring.git. Annual values of each of the indicators used in this study are presented in Table S1, trends in these indicators are summarised by KBA in Table S2 and ecoregion-level trends are summarised in Table S3.

References (51)

  • W. Turner et al.

    Remote sensing for biodiversity science and conservation

    Trends Ecol. Evol.

    (2003)
  • W. Turner et al.

    Free and open-access satellite data are key to biodiversity conservation

    Biol. Conserv.

    (2015)
  • Q. Zhang et al.

    Mapping urbanization dynamics at regional and global scales using multi-temporal DMSP/OLS nighttime light data

    Remote Sens. Environ.

    (2011)
  • Y. Zhou et al.

    A cluster-based method to map urban area from DMSP/OLS nightlights

    Remote Sens. Environ.

    (2014)
  • F. Achard et al.

    Determination of tropical deforestation rates and related carbon losses from 1991 to 2010

    Glob. Change Biol.

    (2014)
  • Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R.H.B., Singman, H., Dai, B., Scheipl, F., Grothendieck,...
  • A.E. Beresford et al.

    Protection reduces loss of natural land-cover at sites of conservation importance across Africa

    PLoS One

    (2013)
  • BirdLife International

    Monitoring Important Bird Areas: A Global Framework (version 1.2)

    (2006)
  • BirdLife International, 2018. Key Biodiversity Areas (KBA) digital boundaries. Developed by the KBA Partnership:...
  • A.B. Brink et al.

    Introduction to remote sensing for conservation practitioners

  • G.M. Buchanan et al.

    Delivering a global, terrestrial, biodiversity observation system through remote sensing

    Conserv. Biol.

    (2009)
  • G.M. Buchanan et al.

    An assessment of land cover and threats in Important Bird Areas in Africa

    Bird Conserv. Int.

    (2009)
  • G.M. Buchanan et al.

    Free satellite data key to conservation

    Science

    (2018)
  • S.H.M. Butchart et al.

    Protecting important sites for biodiversity contributes to meeting global conservation targets

    PLoS One

    (2012)
  • Butler, R.A., 2008. Fire monitoring by satellite becomes key conservation tool: an interview with GIS experts at...
  • Cited by (16)

    • Leveraging on technology-driven information systems for conservation through informed decisions in the Hindu Kush Himalayas

      2023, Climate Change in the Himalayas: Vulnerability and Resilience of Biodiversity and Forest Ecosystems
    • What is going on within google earth engine? A systematic review and meta-analysis

      2023, Remote Sensing Applications: Society and Environment
      Citation Excerpt :

      Articles included in the ‘forestry’ category (F) did not show a clear pattern of applications, in spite of the varied utilization of satellite imagery sources (Table 4). The review found a wide variety of approaches, covering issues such as classification of Mediterranean forest habitats on seasonal Sentinel 2 time-series and input image composition optimisation (Praticò et al., 2021), monitoring of temperate forest degradation using Landsat imagery (Chen et al., 2021), identification and monitoring of threats to key biodiversity areas in Africa using MODIS and NOAA imagery (Beresford et al., 2020), and mapping of mangrove canopy phenology (Valderrama-Landeros et al., 2020; Cissell et al., 2021). In contrast, studies reported under the category ‘soil’ (S) presented clearly defined fields of research, which were grouped into the mapping of soil physico-chemical properties (Xiao et al., 2020; Greifeneder et al., 2021; Ye et al., 2021; Aksoy et al., 2022; Luo et al., 2022) and studies on soil erosion processes (Wang and Zhao, 2020; Alexakis et al., 2021).

    • Global occupation of wetland by artificial impervious surface area expansion and its impact on ecosystem service value for 2001–2018

      2022, Ecological Indicators
      Citation Excerpt :

      In this context, we analyse the occupation of wetlands by ISA expansion at a global scale with the help of the Google Earth Engine (GEE) and calculate the ESV loss caused by it, the results of the analysis are discussed at global, continental and national scales. The data analysis work in this paper is mainly based on the GEE platform, a remote sensing big data platform developed based on Google Earth, both GEE and Google Earth have been used in several academic studies for large-scale spatial analysis and accuracy verification (Beresford et al., 2020; Guo et al., 2022; Wu et al., 2021). GEE makes it possible to calculate and analyse big data quickly worldwide, and has provided a convenient new way of multi-scale spatial analysis (Gorelick et al., 2017).

    • Evaluating the conservation priority of key biodiversity areas based on ecosystem conditions and anthropogenic threats in rapidly urbanizing areas

      2022, Ecological Indicators
      Citation Excerpt :

      After the IUCN promoted Key Biodiversity Areas (KBAs), studies on ecosystem services assessment of KBAs and the effect of human activities on KBAs emerged. Increasing studies have conducted assessments for prioritizing key biodiversity areas for conservation by using a series of indices of ecosystem services and human activities (Beresford et al., 2020; Huang et al., 2020). Shrestha et al. (2021) found that Key Biodiversity Areas in the Chindwin River Basin of Myanmar are facing increasingly threats from human activities, the most significant threats to biodiversity in the KBAs were found to be mining, logging, dam construction, and agriculture.

    • Vulnerability assessment of Sierra Nevada de Santa Marta, Colombia: World's most irreplaceable nature reserve

      2021, Global Ecology and Conservation
      Citation Excerpt :

      Natural reserves face a large number of different threats, although biodiversity loss due to habitat degradation and fragmentation, overexploitation, invasive species, climate change, contamination are quite common (Groom et al., 2006). The assessment of these threats is usually carried out employing indicators that can be taken from different sources, including literature reports, field-based monitoring and remote-sensing-related technologies (Beresford et al., 2020). In this work, the authors have used the methods implemented by the Alexander von Humboldt Biological Resources Research Institute (IAvH) of Colombia (Mesa-S et al., 2016).

    View all citing articles on Scopus
    View full text