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
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]

Appendix G   Brief review of the potential for estimation of biomass by remote sensing Previous topic Parent topic Child topic Next topic

There is active research on methods to estimate biomass in tropical forests using remote sensing techniques, including analysis of spectral indices, and use of SAR and LIDAR. In general these methods require calibration using ground-based data. Saturation may be a problem, especially in tropical countries because the correlation between biomass and the remote sensing data may not be effective at high biomass densities.
A key issue when using tree height (estimated using LIDAR or SAR) to estimate biomass, is that the relationship between height and biomass is likely to differ markedly with forest type, tree age, speciation, and following forest disturbance (e.g. between primary and secondary forest). Such differences need to be understood and taken into account in order to improve estimates of forest biomass and change.
This review suggests that existing large-scale biomass maps derived from remote sensing data need extensive in-country testing to confirm that they are reliable for application in specific forest types and at the spatial scale of interest. Biomass estimation error using remote sensing is high at the plot scale (< 1 ha) and up to 1 sq km (100 ha) (Saatchi et al., 2011) and therefore robust field estimates of biomass based on adequate plot size, sufficient spatial sampling, and use of appropriate allometrics are needed (e.g. Chave, et al., 2004; Avitabile et al., 2011). This means that currently the method is unlikely to be cost efficient.
A brief review of recent work to produce biomass estimates for tropical forests follows.
Use of LIDAR for biomass estimation
Biomass estimates are usually obtained by combining LIDAR data with field observations and sometimes optical data (e.g. MODIS surface reflectance) for obtaining wall-to-wall maps of biomass from point-based estimates as in Baccini et al. (2011).
Baccini et al., (2008) produced a spatial biomass map for Africa by combining remote sensing and field estimates of biomass derived from a range of sources. Mitchard et al., (2011) criticised this map, claiming that the ground data used for calibrating the remote sensing were inadequate, and resulted in significant underestimation of biomass, especially for areas with high biomass densities. Avitabile et.al. (2011) reported poor correspondence between 7 biomass maps (derived either by extrapolation of field estimates of biomass, or derived using remote sensing) for Uganda, both in terms of average biomass densities and spatial patterns. They concluded that the next critical step to increasing reliability of biomass maps was the collection of more reliable field biomass data for key forest types.
Saatchi et.al. (2011) used remote sensing to derive a biomass map for tropical forests at 1 km resolution, and to estimate the errors of biomass estimates made at different spatial scales. They established a relationship between forest stand height and biomass at 493 locations across the tropics. This relationship was able to predict ground estimates of biomass for many other locations with an uncertainty of about 24% on average. Estimates of forest height derived from space-borne LIDAR were then used to estimate biomass at many more locations. The biomass estimates derived from ground measurements and those estimated using LIDAR were then extrapolated across the entire tropical forest using a data-fusion model and satellite imagery from a range of sources. No validation of these new biomass estimates appears to have been undertaken. The authors assumed that their initial field estimates of biomass were error free, but acknowledged that there may have been significant and systematic non-random errors in the estimates used. Analysis by Chave et al. (2004) of the sources of error involved in biomass estimation at both plot and landscape scale in tropical forests, suggests that such errors were very likely. Chave et al., 2004 provide advice on how to minimize biomass estimation errors, and identified the critical importance of appropriate selection of allometric models which they concluded were a high contributor to uncertainty.
Baccini et al., (2012) used a combination of field data and remote sensing to generate a biomass map for tropical forests at 500m resolution. They used continental allometric models (of moist, dry, and wet forests) to convert field data measurements to forest biomass at a range of locations across several countries, and then correlated biomass with vegetation structure metrics derived from space-borne LIDAR.
Tyukavina et al., (2015) generated a pantropical vegetation stratification based on remotely sensed data and used the space-borne LIDAR biomass estimates from Baccini et al., 2012 to assign a biomass density value to each stratum, in essence using the LIDAR estimates as a surrogate for forest inventory.
Baccini and Asner 2013 reports on the improvements resulting from the substitution of space borne GLAS LIDAR with high resolution airborne LIDAR to calibrate remotely sensed data. They first show the agreement between Baccini et al., 2012 biomass map estimates and independent high resolution LIDAR measurements in the Peruvian and Colombian Amazon and report that pantropical datasets explain about 70% of the variance in biomass, second they show that by fine calibration with airborne LIDAR samples the root mean squared error decreases of about 38-44% and suggest great potential of the integration of airborne LIDAR sampling with space remote sensing to generate wall-to-wall biomass estimates.
Sources of LIDAR
LIDAR systems emit laser pulses and by measuring the timing and intensity of the returns, three-dimensional information on vegetation structure is inferred which in turn allows for prediction of forest structure attributes related to aboveground biomass. The most feasible approach for obtaining biomass estimates from remote sensing data is to make use of LIDAR-based measurements of vegetation structure. There are two main sources of LIDAR data: (1) small footprint, airborne LIDAR data and (2) full waveform, space-borne LIDAR data. At the time of writing there is no operational LIDAR satellite; data availability is limited to what is available from the GLAS instrument on the now defunct ICESat satellite between 2003 and 2009.
Airborne LIDAR data
Airborne LIDAR data, if available for a sample of the study area, can be used to estimate biomass. The LIDAR data provides three-dimensional information on the vegetation structure that can be regressed against plot-level aboveground measurements of biomass to provide biomass estimates for each LIDAR observation. While allometric models exist for a range of conditions which allows for biomass estimation without in situ collection of biomass, biomass measurements within the area covered by the LIDAR flight tracks will ensure that regional and local variation in the LIDAR-biomass relationship is included (Asner, 2009). Examples of how to use airborne LIDAR data together with field plots to estimate biomass are provided by: Asner et al., (2010) (IPCC-compliant estimates of carbon stocks and emissions in the Peruvian Amazon); Nelson et al., (2004) (biomass estimation in Delaware, United States); Næsset et al., (2013) (biomass change estimates in boreal forests, Norway); and Lefsky et al., (1999) (biomass estimation in deciduous forests in Maryland, United States).
Satellite LIDAR data
LIDAR observations from space are currently limited to data from the GLAS sensor on board the ICESat. The sensor collected LIDAR data from 2003 to 2009 which is available for free download at NASA Reverb. Opens in new window ICESat-2, which will carry LIDAR instruments, is planned for launch in early 2016. No other missions are planned at the time of writing. Therefore there is a data gap in space borne LIDAR observations between 2009 and 2015.
Research indicates that, while it is possible to estimate tree height from ICESat/GLAS data which in turn can be regressed to obtain biomass estimates (Sun et al., 2007), estimating tree height from GLAS data is less straightforward compared to using airborne, small footprint LIDAR data. On sloped areas, topographic information is required to estimate tree height because of the elliptical shape of the GLAS footprint (Lefksy et al., 2005). Sources that provide descriptions of using GLAS data for estimating tree height and biomass include: Baccini et al., 2012; Saatchi et al., 2011; Nelson et al., 2008; Boudreau et al., 2008; Lefksy et al., 2005.
Existing large-scale biomass products include:
  • The National Level Carbon Stock Dataset (Tropics) Woods Hole Research Center (WHRC) provides maps of above-ground live woody biomass for the tropics. Using a combination of field measurements and space-borne LIDAR observations at 70 m spatial resolution from the Geoscience Laser Altimeter System (GLAS) instrument on board the Ice, Cloud and land Elevation Satellite (ICESat), and optical MODIS imagery at 500 m spatial resolution, the WHRC National Level Carbon Stock Dataset provides above-ground live woody biomass at 500m resolution for the tropics 2007-2008 (Baccini et al., 2012). The data are available online. Opens in new window
  • The National Biomass and Carbon Dataset (NBCD2000) WHRC provides a 30 m biomass product for the conterminous United States. This map does not cover tropical areas, but it provides a model for how NFI plot data can be combined with remote sensing data to make maps of biomass. NBCD2000 is based on a combination of data from the USDA Forest Service Forest Inventory and Analysis (FIA), the 2000 Shuttle Radar Topography Mission (SRTM) and Landsat-7/ ETM+. It provides basal area-weighted canopy height, aboveground live dry biomass, and standing carbon stock for the year 2000 (Kellndorfer, et al., 2010).
  • The JPL Carbon Maps. The Jet Propulsion Laboratory of NASA and the California Institute of Technology provide a biomass product similar to that of the WHRC National Level Carbon Stock Dataset. The maps provide forest above-ground carbon and biomass for sub-Saharan Africa, the Americas south of latitude 30° N, and South-East Asia and Australia between the latitudes of 40° N and 30° S at 1 km resolution. Point-based estimates of biomass generated from a combination of field data and space-borne LIDAR data from ICESat/GLAS were extrapolated using optical data from MODIS and SAR data from SRTM and QuickSCAT (Saatchi et al., 2011). Opens in new window
Use of synthetic aperture radar (SAR) for biomass estimation
Although synthetic aperture radar (SAR) has demonstrated potential in the estimation of aboveground biomass, there are limitations arising from:
  • for some bands, rapid saturation of the signal at low aboveground biomass stock
  • terrain
  • rainfall and soil moisture effects
  • localised algorithm development focussing on a single biome or mono-species stands
  • lack of consistency in estimates as a function of sensor parameters.
Calibration of the retrieval algorithm depends on reliable ground data, which need to be collected under a representative range of environmental conditions. As such, there is limited transferability of algorithms within and between different forest structural types and, so far, no reliable means of estimating aboveground biomass (Lucas et al., 2010).
Recent work in Australia (Biomass Plot Library Opens in new window) as focused on collating, at a national level, forest inventory data that has been collected by various agencies and individuals since the 1990s. Measurements of the size (e.g., diameter) of individual trees have been used as input to standardized allometric models to generate tree and subsequently stand-level estimates of above and also below ground biomass. Estimates of uncertainty have also been provided, with these considering the errors associated with the measurement of individual trees, the application of different allometric models and the scaling from the tree to the plot level. The intention of the biomass library is to facilitate continual upload of forest inventory data in return for tree and stand level biomass estimates which, in turn, will be released publicly. As well as providing data to support, for example, carbon cycle science, the data are also intended to support the development and validation of algorithms for the retrieval of biomass at national to global levels from past, current and future sensors including ESA's BIOMASS Mission, JAXA's ALOS-2 PALSAR-2 and NASA's NiSAR and GEDI.
SAR based estimation of aboveground biomass has been more successful in temperate forests compared to tropical forests, due largely to fewer species and lower biomass (Castro et al., 2003). Increased sensitivity has been achieved using ratios or correlations between multi-frequency, multi-polarisation backscatter and biomass components (Castro et al., 2003). Alternative approaches, including SAR interferometry, polarimetric interferometry, tomography and integration with LIDAR and other data are the focus of current investigations.
SAR has a demonstrated capacity to quantify biomass up to a certain level, depending on the frequency used. Once saturation of the signal is reached, the data are no longer useful for biomass estimation (Bottcher et al., 2007 and 2009; Gibbs et al., 2007). Cross-polarised backscatter demonstrates greater sensitivity to forest biomass than co-polarised backscatter, however, the use of multiple polarisations is recommended for use in retrieval algorithms (Castro et al., 2003). L-band SAR is useful for discriminating regrowth stage and estimating biomass in low biomass (40-150 t/ha) forests. Dual polarisation and dual-season coverage is required. C-band SAR is only useful in very low biomass forests (30-50 t/ha). The shorter wavelength does not penetrate further than the leafy canopy (Castro et al., 2003). Texture analysis of multi-temporal, high resolution C-band data may provide some useful input (Castro et al., 2003).
ESA has recently approved the BIOMASS mission, a P-band interferometer which will provide global scale estimation of aboveground biomass in the 2020 timeframe. P-band SAR can facilitate biomass estimation in high biomass (100-300 t/ha) forest.
Sub-national demonstrations
Biomass estimating using SAR requires sophisticated processing and extensive ground calibration, and while the research is progressing, there are few demonstrations at sub-national scale. Successful demonstrations have largely relied on GFOI non-core data streams, including airborne (GeoSAR) and satellite SAR (ALOS PALSAR, ENVISAT ASAR).
  • Eastern Australia: Relationships established between ALOS PALSAR L-HH and HV backscatter and field measured AGB led to the production of an interim AGB map (Lucas et. al., 2010). Validation underway. Improvements likely through the integration of Landsat and ICESat data products
  • Mexico: Wall-to-wall AGB map produced using ALOS PALSAR data acquired in 2008 at 15 m spatial resolution (GEO, 2011)
  • North-eastern USA: Inversion of semi-empirical model calibrated for ALOS PALSAR FBD images to estimate biomass (Cartus et al., 2012). Retrieval accuracy for HV intensity data was consistently better than for HH. Weighted combinations of single-date biomass estimates in a multi-temporal stack significantly improved performance. RMSE of 12.9 t/ha (R2 = 0.86) compared to forest inventory
  • Boreal forest: Model based estimation of growing stock volume (GSV) up to 300 m3/ha using hyper-temporal ENVISAT ASAR ScanSAR images (Santoro, et. al., 2011). RMSE of 34.2 – 48.1 % at 1 km pixel size. GSV was improved by averaging over neighbouring pixels. Transferability of method to tropical forest requires investigation
  • Woody vegetation in Queensland, Australia: the study aimed to establish whether the relationship between PALSAR HH and HV backscattering coefficients and AGB was consistent within and between structural formations (forests, woodlands and open woodlands, including scrub). The study examines saturation and concludes that PALSAR data acquired when surface moisture and rainfall are minimal allow better estimation of the AGB of woody vegetation and that retrieval algorithms ideally need to consider differences in surface moisture conditions and vegetation structure (Lucas et al., 2010).
Asner, G.P. (2009). Tropical forest carbon assessment: integration satellite and airborne mapping approaches. Environmental Research Letters, 4(3), 034009.
Asner, G.P., Powell, G.V.N., Mascaro, J., Knapp, D.E., Clark, J.K., Jacobson, J., Kennedy-Bowdoin, T., Balaji, A., Paez-Acosta, G., Victoria, E., Secada, L., Valqui, M. & Hughes, R.F. (2010). High-resolution forest carbon stocks and emissions in the Amazon. Proceedings of the National Academy of Sciences, 107: 16738-16742.
Avitabile, V., Herold, M., Henry, M., & Schmullius, C. (2011). Mapping biomass with remote sensing: a comparison of methods for the case study of Uganda. Carbon Balance and Management. 6:7.
Baccini, A., Goetz, S.J., Walker, W.S., Laporte, N.T., Sun, M., Sulla-Menashe, D., Hackler, J., Beck, P.S.A., Dubayah, R., Friedl, M.A., Samanta, S., Houghton, R.A. (2012). Estimated carbon dioxide emissions from tropical deforestation improved by carbon-density maps. Nature Climate Change. 2:182-185.
Baccini A. & Asner, G.P. (2013). Improving pantropical forest carbon maps with airborne LIDAR sampling. Carbon Management. 4(6):591–600.
Cartus, O., Santoro, M., & Kellndorfer, J. (2012). Mapping forest aboveground biomass in the Northeastern United States with ALOS PALSAR dual-polarisation L-band. Remote Sensing of Environment. 124:466-478.
Castro, K.L., Sanchez-Azofeifa, G.A., & Rivard, B. (2003). Monitoring secondary tropical forests using space-borne data: implications for Central America. International Journal of Remote Sensing. 24:1853-1894.
Chave, J., Condit, R., Aguilar, S., Herandez, A., Lao, S. & Perez, R. (2004). Error propagation and scaling of tropical forest biomass estimates. Philosophical Transactions B. Royal Society London. 359:409-420.
GEO, (2011). Observations, Group on Earth.GEO-FCT Product Development Team Technical Status Report v2.0. s.l. : GEO, 2011.
Gibbs, H.K., Brown, S., Niles, J.O., & Foley, J.A. (2007). Monitoring and estimating tropical forest carbon stocks: making REDD a reality. Environmental Research Letters. 2:045023.
Kellndorfer, J.M., Walker,W.S, LaPoint, E., Kirsch, K., Bishop, J., & Fiske, G. (2010). Statistical fusion of lidar, InSAR, and optical remote sensing data for forest stand height characterization: A regional-scale method based on LVIS, SRTM, Landsat ETM+, and ancillary data sets. Journal of Geophysical Research. 115 G00E08. 
Lucas, R., Armston, J., Fairfax, R., Fensham, R., Accad, A., Carreiras, J., Kelley, J., Bunting, P., Clewley, D., Bray, S., Metcalfe, D., Dwyer, J., Bowen, M., Eyre, T., Laidlaw, M., & Shimada, M. (2010). An Evaluation of the ALOS PALSAR L-Band Backscatter—Above Ground Biomass Relationship Queensland, Australia: Impacts of Surface Moisture Condition and Vegetation Structure. IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing. 3:576-593;
Mitchard, E. T. A., Saatchi, S. S., Lewis, S. L., Feldpausch, F. F., Woodhouse, I. H. and P. Meir (2011). Comment on a First map of tropical forest biomass derived from satellite imagery. Environmental Research Letters. 6:049001
Næsset, E., Bollandsås, O.M., Gobakken, T., Gregoire, T.G. & Ståhl. G. (2013). Model-assisted estimation of change in forest biomass over an 11 year period in a sample survey supported by airborne LIDAR: A case study with post-stratification to provide activity data. Remote Sensing of Environment. 128:299-314.
Saatchi, S.S., Houghton, R.A., dos Santos Alvalá, R.C., Soares, J.V., Yu, Y. (2007). Distribution of aboveground live biomass in the Amazon basin. Global Change Biology. 13:816-837.
Santoro, M, Beer, C., Cartus, O., Schmullius, C., Shvidenko, A., McCallum, I., Wegmüller, U., Wiesmann, A. (2011). Retrieval of growing stock volume in boreal forest using hyper-temporal series of Envisat ASAR ScanSAR backscatter measurements. Remote Sensing of Environment. 115:490-507.
Tyukavina A., Baccini, A., Hansen, M.C., Potapov, P.V., Stehman, S.V., Houghton, R.A., Krylov, A.M., Turubanova, S., & Goetz, S.J. (2015). Aboveground carbon loss in natural and managed tropical forests from 2000 to 2012. Environmental Research Letters. 10(7): 74002.