By: Tor-G. Vågen and Leigh Winowiecki (World Agroforestry Centre (ICRAF))
Introduction
The Sentinel Landscapes (SL) initiative is comprised of geographic areas or sets of areas with a broad range of biophysical, social, economic and political conditions. A core component of the SL network is a set of Land Degradation Surveillance Framework (LDSF) sentinel sites, where intensive data collection is taking place using co-located biophysical and socio-economic surveys in order to collect information on a number of social-ecological indicators. The initial set of SL sites were established in 2012, growing into a network of 10 landscapes by the end of 2014.
The interactive map below shows the different SLs outlined in red.
Land Health Mapping
Indicators of land health
An important consideration in the choice of indicators of land health and land degradation is whether they can be measured readily or not, allowing for assessments that are operational while not oversimplifying the complex processes leading to land degradation. Indicators that integrate or reflect multiple aspects of land health are particularly important in this regard, as are assessments of interactions between different indicators of land degradation.
To operationalize land health baseline assessments and monitoring, analytical tools and approaches are needed that simultaneously summarize variations in multiple measures of ecosystem function and provide critical spatially explicit evidence to land managers and decision makers at different scales. Examples of such indicators include soil erosion, soil organic carbon (SOC) concentrations and stocks, soil compaction, soil acidity, soil fertility parameters, vegetation dynamics (rather than just productivity), infiltration capacity, and biodiversity. All of these indicators can be readily measured on the ground and/or from remote sensing.
Soil erosion
Accelerated soil erosion is arguably the most important indicator of land degradation and also one of the most widespread forms of degradation worldwide. Interactions between soil erosion and a number of other ecosysem processes, including vegetation dynamics, carbon cycling and hydrological functioning make the detection of soil erosion hotspots in landscapes a priority for targeting interventions to reverse land degradation, or restore already degraded areas.
One of the outputs from the LDSF is therefore predictive maps of soil erosion prevalence at multiple spatial scales, based on remote sensing imagery and ensemble prediction models [@Vagen2013d].
Soil organic carbon (SOC)
Given the heterogeneity of landscapes, spatial patterns of SOC concentrations tend to be complex. Spatial information on the distribution of SOC and other soil properties therefore need to be made at relevant spatial scales (i.e. at the farm or local farming system level). Most models and estimates of SOC only allow for coarse-scale assessments of SOC sequestration potential and are not able to predict the possible fate of carbon due to land-use change at scales relevant to management interventions. Alternative approaches have been suggested that enable these types of assessments to be conducted using moderate to high resolution satellite imagery to predict SOC [@Vagen2013d; @Winowiecki2015a], as well as cumulative soil mass (CM) measurements and direct calculations of SOC stocks [@Vagen2013], without the use of bulk density, which is prone to error.
Vegetation/land cover
The most commonly used indicators of vegetation or land cover attempt to estimate the vigor of the vegetation, either expressed as Net Primary Productivity (NPP), Gross Primary Productivity (GPP) or simply vegetation greenness. The latter is generally indirectly associated with NPP and biological properties of the vegetation, such as chlorophyll, nitrogen, phosphorous, and moisture content. In remote sensing, spectral unmixing methods are often used to estimate both fractional and green vegetation cover using a range of sensors, the most common ones being Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat.
Remote sensing
Monitoring of ecosystem health requires data, methods, and technologies that span 10 to 30 year periods to separate the effects of land management activities from those of major climatic events. Satellite remote sensing meets the temporal and spatial scale requirements for monitoring of ecological indicators and can also be used to derive indicators of land degradation. Over the last 30+ years, the Advanced Very High Resolution Radiometer (AVHRR) and the Landsat program’s Multispectral Scanner, Thematic Mapper ™, Enhanced TM and recently Landsat 8 (LC8) provide such data.
Remote sensing analysis is therefore a key component in the assessment of ecosystem health in the Sentinel Landscapes, for example in identifying land degradation hotspots, land cover, and the spatial distribution of soil organic carbon in a given landscapes. A number of different remote sensing platforms are applied, including MODIS, Landsat and RapidEye, providing mapping outputs at scales ranging from 500m to 5m spatial resolution.
Data sharing in the Sentinel Landscapes
The various datasets generated as part of the Sentinel Landscape inititiative are shared through Harvard Dataverse. Data that are currently avaiable and can be requested at present include:
Salazar, Angel; Albornoz, Marco; Castillo, Rider; Cusi, Edgar; Odicio, Manuel; Mamani, Jose; Robiglio, Valentina; Reyes, Martin; Vågen, Tor-Gunnar; Winowiecki, Leigh, 2016, “Land and Soil Health Assessment in the Western Amazon Sentinel Landscape”, http://dx.doi.org/10.7910/DVN/2FTPZ5, Harvard Dataverse, V1 [UNF:6:dv7xUc0Dg3fBawMwyPYh7g==]
Tondoh, Jerome; Doumbia, Oumar; Simbélé Koura, Paulin; Vågen, Tor-Gunnar; Winowiecki, Leigh, 2015, “Land and Soil Health Assessment in the Western Africa Sentinel Landscape”, http://dx.doi.org/10.7910/DVN/9ZMZWR, Harvard Dataverse, V2 [UNF:6:x75HVo7sYeAazoQThUBQEQ==]
Sepulveda, Norvin; Ordonez, Jenny; Uloa, Noel; Vega, Ricardo; Canales, Alfredo; Peralta, Jaime; Cruz, Melvin; Zavala, Ariel; Zavala, Angel Acosta; Calderon, Angelica Leonor Ruiz; Montoya, Lilia; Vågen, Tor-Gunnar; Winowiecki, Leigh, 2016, “Land and Soil Health Assessment in the Nicaragua-Honduras Sentinel Landscape”, http://dx.doi.org/10.7910/DVN/OTSSRA, Harvard Dataverse, V1
Takoutsing, Bertin ; Vagen, Tor-Gunnar ; Winowiecki, Leigh ; Sonwa, Denis; Nna, Denis ; Tchoundjeu, Zacharie , 2016, “Land and Soil Health Assessment in the Central Africa Humid Tropics Transect (CAFHUT) Sentinel Landscape”, http://dx.doi.org/10.7910/DVN/V8M94D, Harvard Dataverse, V1
In addition, you may access data and maps through the ICRAF Landcape Portal. Please register to view and explore maps of the various indicators produced, such as the erosion map shown below. Also, you may explore data by clicking the EXPLORE DATA tab at the top of this window.
Summary of the number of sentinel site ground observation points collected as part of the SL initiative, by country (2013 - 2015).
The graphics below show an overview of selected biophysical site characteristics for each sentinel site in the SL network. Bubbles with labels or bars in the lower panels represent averages (means) for each site, dots represent medians, while whiskers show 25th and 75th percentiles. The middle panel shows an overview of the main vegetation structure classes in each sentinel site.
Click on the “Detailed summary for each SL sentinel site” tab above to see more details for each site.
Cultivated area (%)
Agroforestry (trees present in cultivated areas) (%)
Vegetation structure
Forest cover (%)
Average tree density (trees/ha)
Erosion prevalence
The data table below shows a summary of LDSF data from the sentinel landscapes, by sampling cluster. Clicking on a row in the table will open a map and show selected summary graphics below the table. Use the dropdown list above to select the Site you would like to explore.
Boxplots showing tree and shrub densities in each LDSF sampling cluster within the selected site.
The black box shows the cluster selected in the data table above.
Plot showing the relative proporion of erosion, compaction, trees and cultivated areas within the selected site.