Uncertainty [MGD Sections]

UNFCCC decisions and requirements
IPCC good practice guidance
Relationship to UNFCCC
GHGI coverage, approaches, methods and tiers
Design decisions relevant to national forest monitoring systems
Land cover, land use and stratification
Forest reference emission levels and forest reference levels
Quality assurance and quality control
Guiding principles – Requirements and design decisions
Estimation methods for REDD+ activities
Integration frameworks for estimating emission and removals
Selecting an integration framework
Activity data x emission/removal factor tools
Fully integrated tools
Practical considerations in choosing an integration tool
Guiding principles – Methods and approaches
Remote sensing observations
Coarse resolution optical data
Medium resolution optical data
High resolution optical data
L-band Synthetic aperture radar
C-band and X-band SAR
LIDAR
Global forest cover change datasets
Ground-based observations
National forest inventories
Auxiliary data
Guiding principles – Remote sensing and ground-based observations
Activity data
Methods for estimating activity data
Maps of forest/non-forest, land use, or forest stratification
Detecting areas of change
Additional map products from remote sensing
Estimating uncertainty of area and change in area
Estimating total emissions/removals and its uncertainty
REDD+ requirements and procedures
Reporting forest reference emission levels and forest reference levels
Technical assessment of forest reference emission levels and forest reference levels
Reporting results of REDD+ activities
Technical analysis of the REDD+ annex to the BUR
Additional advice on REDD+ reporting and verification
Guiding Principles – Reporting and verification of emissions and removals
Financial considerations
Country examples – Tier 3 integration
Use of global forest change map data
Relative efficiencies
Developing and using allometric models to estimate biomass

Record Keeping [MGD Sections]

Integration + Estimation [MGD Sections]

Ground Based Observations [MGD Sections]

4.1.2   Medium resolution optical data Previous topic Parent topic Child topic Next topic

Medium resolution lies in the range 10 to 80 metres. The most common imagery which may be used for monitoring REDD+ activities is 30 metre resolution, from the Landsat series of satellites (GOFC-GOLD Sourcebook, 2015 Opens in new window). Advantages associated with Landsat data include (a) a long history of use, (b) global acquisition, pre-processing and archiving of data, (c) free access to data in the US archive. Landsat will often be the only dataset available for estimating historical activity data. There is an ongoing effort to combine all collections of Landsat data into a single global archive, all processed to the same standards and distributed through the USGS (Wulder et al., 2015). Figure 9 shows the anticipated state of the archive when the Landsat Global Archive Consolidation (LGAC) process is complete, which should be in 2017.
Figure 9: Landsat images collected and expected to be in the Landsat global archive by 2017
Mapedit.jpg
Information is also available from the GFOI on the status of the Landsat archive for individual countries Opens in new window, like the example shown in Figure 10 for Bolivia. The number of images in the archive increased significantly following the launch of Landsat 7 in 1999. Similarly, the launch of Landsat 8 has increased in the number of images being collected and archived.
The Landsat data series goes back to the 1970s(1) and the successful launch of the Landsat 8 in February 2013 continues the time series for the foreseeable future. Construction of Landsat 9 has begun with an anticipated launch in 2021. The availability of an historical archive is particularly important for establishing reference levels. Similarly, consistent observations over time remains the key to automated methods to detect deforestation and forest degradation.
The use of optical sensors is a limitation in areas with persistent cloud cover, but the frequency of data collection has increased in recent years, which helps minimize this issue (see Box 14) and the accessibility and global coverage associated with Landsat generally make it the first data source to consider for a NFMS. For many purposes Landsat will serve to fulfil national remote sensing data requirements associated with REDD+ activity data collection. For example, the Landsat archive opens the possibility of conducting historical time series analysis within the suggested temporal ranges suggested for developing FREL/FRLs (10-15 years - Chapter 2, Section 2.3.3) and the confidence that this data source will be available for future monitoring requirements.
The CBERS-4 and Sentinel 2 satellites will increase availability of medium resolution data, including by making 10 m resolution data freely available and facilitating applications which have hitherto been regarded as only possible at high resolution.
Countries having national operational programs for forest cover monitoring using Landsat or Landsat-like data include Australia (Furby et al., 2008), Brazil (DMC and CBERS; Souza, 2006), India (IRS; Pandey, 2008) and the United States (Fry et al., 2009 Opens in new window).
Figure 10: Distribution of Landsat scenes with less than 20% cloud cover archived for Bolivia
Figure10.svg

Box 14: Removing clouds and cloud shadows in optical satellite imagery used for mapping activity data

As explained in Box 23, with the opening of the Landsat archive in 2008 (Woodcock et al., 2008) time series of Landsat data can be obtained for almost any location on Earth. Clouds can cause difficulty with optical sensors though techniques exist to address this: when classifying a single image or an image pair, it is straightforward to identify and classify any obvious clouds and cloud shadows (contaminations) present in the image. These pixels can then be removed from the analysis or replaced by pixels from cloud free images from the closest available point in time. .
When analysing a time series of observations for land surface activities using all available images, clouds and cloud shadows need to be accurately identified as anomalies in the time series which the classification algorithm could wrongly attribute to surface activities. Fortunately, use of a time series itself makes it easier to do this. For example, when using continuous change detection and classification (CCDC) for mapping activity data (Arevalo, 2016; Olofsson et al., 2016), the analyst first applies an algorithm that looks for clouds and cloud shadows screening each image individually but without use of previous or subsequent observations (Zhu & Woodcock, 2012b). A second algorithm, looking now at each pixel as part of the time series, then checks whether the omitted pixels were in fact anomalies or real changes at the surface time (Zhu & Woodcock, 2014a). The single image cloud screening algorithm referenced here, Fmask, is currently being implemented by the USGS to screen all Landsat images in the US archive, such that each image will be delivered with a Fmask-based cloud/cloud shadow mask. Fmask is not the only published algorithms that screen for clouds in Landsat images – many similar algorithms have been published in the last decade (Huang et al., 2010; Irish et al., 2006; Masek et al., 2006; Roy et al., 2010; Scaramuzza et al., 2012). These semi automated processes can still miss cloud and haze and should always be accompanied by manual checks and, where necessary, manual cloud, shadow and haze removal. As also mentioned in Box 23, an alternative use of a time series is to create composites by selecting certain observations in a time series according to some criteria. For example, if the median of the surface reflectance of annual time series of Landsat observations is computed, annual images are created that are free of cloud and cloud shadow provided that clouds are not present for most of the year. More advanced criteria can be developed that take phenology, spectral ratios, advanced statistics and/or results from a cloud screening algorithm into account (Griffiths et al., 2014; Hansen et al., 2013; Kennedy et al., 2010).
The discussion here focuses on Landsat because currently it is the only mission that provides free data in combination with a long enough record to allow for time series analysis. It also has a thermal band that helps in the identification of clouds. In the future, we will be able to construct time series of other data, including Sentinel-2 data, which lacks a thermal band but like Landsat-8 has a cirrus band that has been found to be helpful in the identification of clouds (Zhu et al., 2015).

 (1)
Consistent analysis ready Thematic Mapper (TM) images are available from the historic archive dating back to 1984, corresponding to the launch of Landsat 4.