You are on page 1of 42




This publication was produced for review by the United States Agency for
International Development. It was prepared by Tetra Tech.

This publication was produced for review by the United States Agency for International Development by Tetra Tech ARD,
through a Task Order under the Prosperity, Livelihoods, and Conserving Ecosystems (PLACE) Indefinite Quantity
Contract Core Task Order (USAID Contract No. EPP-I-00-06-00008-00, Order Number AID-OAA-TO-11-00022).

Tetra Tech
159 Bank Street, Suite 300
Burlington, Vermont 05401 USA
Telephone: (802) 658-3890
Fax: (802) 658-4247

Tetra Tech Contacts:
Ian Deshmukh, Senior Technical Advisor/Manager

Forest Carbon, Markets and Communities (FCMC) Program
1611 North Kent Street
Suite 805
Arlington, Virginia 22209 USA
Telephone: (703) 592-6388
Fax: (866) 795-6462

Stephen Kelleher, Chief of Party

Olaf Zerbock, USAID Contracting Officer’s Representative

• Jennifer Hewson, Conservation International
• Marc Steininger, FCMC & Conservation International
• Stelios Pesmajoglou, Greenhouse Gas Management Institute

Contributing authors:
• Angel Parra, Consultant; GHG inventory & LULUCF sector expert
• Gordon Smith, Greenhouse Gas Management Institute
• David Shoch, TerraCarbon, LLC
• John Musinsky, National Ecological Observatory Network
• Fred Stolle, World Resources Institute
• Kemen Austin, World Resources Institute
• Irene Angeletti, Greenhouse Gas Management Institute

The US Agency for International Development (USAID) has launched the Forest Carbon, Markets and Communities
(FCMC) Program to provide its missions, partner governments, local and international stakeholders with assistance in
developing and implementing REDD+ initiatives. FCMC services include analysis, evaluation, tools and guidance for
program design support; training materials; and meeting and workshop development and facilitation that support US
Government contributions to international REDD+ architecture.

Please cite this report as:
Hewson, J., M.K. Steininger and S. Pesmajoglou, eds. 2014. REDD+ Measurement, Reporting and Verification (MRV) Manual,
Version 2.0. USAID-supported Forest Carbon, Markets and Communities Program. Washington, DC, USA.





The author’s views expressed in this publication do not necessarily reflect the
views of the United States Agency for International Development or the United
States Government.


TABLE OF CONTENTS....................................................................................... III
ACRONYMS AND ABBREVIATIONS ...................................................................... V
ACKNOWLEDGEMENTS ................................................................................... VIII
1.0 INTRODUCTION....................................................................................... 1
1.1 PURPOSE, SCOPE AND STRUCTURE ...................................................................................... 1
1.2 BACKGROUND ............................................................................................................................... 4
1.3 REFERENCES...................................................................................................................................... 8
2.0 INSTITUTIONAL ARRANGEMENTS.............................................................. 9
2.1 INTRODUCTION ............................................................................................................................ 9
2.2 ELEMENTS OF A MRV SYSTEM FOR REDD+ .......................................................................13
2.4 STEPS IN ESTABLISHING INSTITUTIONAL ARRANGEMENTS.....................................19
2.5 EXAMPLES ........................................................................................................................................22
2.6 EPA NATIONAL SYSTEM TEMPLATES ...................................................................................26
2.7 REFERENCES....................................................................................................................................30
3.1 INTRODUCTION ..........................................................................................................................31
3.2 IPCC GUIDANCE...........................................................................................................................34
3.3 INVENTORY AND REPORTING STEPS .................................................................................38
3.4 DEFINITIONS OF CARBON POOLS AND LAND USES ..................................................41
3.6 REFERENCES....................................................................................................................................49
4.0 FIELD-BASED INVENTORIES .................................................................... 50
4.1 INTRODUCTION ..........................................................................................................................50
4.2 CARBON POOLS AND THEIR MEASUREMENT ................................................................54
4.3 CONCEPTS AND CONSIDERATIONS IN INVENTORY DESIGN ................................ 57
4.4 THE FOREST CARBON INVENTORY TEAM .......................................................................66
4.5 FIELD WORK AND ANALYSIS .................................................................................................67
4.6 CALCULATING CARBON STOCKS FROM FIELD DATA ...............................................69


..2 COMMUNITY-BASED MONITORING .............9 THE GAIN-LOSS METHOD ..................................4 REFERENCES.................. 4.............7 COMMON SATELLITE DATA SOURCES FOR LAND-USE MONITORING ............ 151 7.2 REPORTING...................82 5...............................................................................................................................................................................................................................................................................76 4..................... 141 6.............................................81 5.........................84 5........................................6 REFERENCES........3 NEAR-REAL TIME MONITORING AND ALERT SYSTEMS .................1 HISTORY OF REDD+ UNDER THE UNFCCC ..........8 SELECTED RESOURCES..............................0 REPORTING AND VERIFICATION: ELEMENTS AND GUIDANCE ....... REPORTING AND VERIFICATION (MRV) MANUAL....79 4....................................10 REFERENCES ............................................0 iv .................................................................................... 152 7..................................................................................................... 121 6................................................................................................8 CONSOLIDATING INVENTORY DATASETS ...................................................................................................................................4 REMOTE SENSING OVERVIEW ................................... 125 6...................... 125 6..................... 165 7........................7 DATA CHECKING ...................... 127 6.............0 REMOTE SENSING OF LAND COVER CHANGE ....................................90 5..................................1 INTRODUCTION ........................... 109 5..................................3 OVERALL STEPS AND NEEDS...........................................................................75 4...... 152 7.............................................................1 INTRODUCTION .3 VERIFICATION ..................... 118 5................78 4..................... VERSION 2................................................................................ 82 5............... 114 5....................................................................93 5......................................................................................................11 SELECTED RESOURCES ................................................5 EMERGING AREAS OF RESEARCH............2 LAND USES AND CATEGORIES IN THE UNFCCC .................................................................................................................. 181 REDD+ MEASUREMENT.....................0 THEMATIC REVIEWS .......................................................................................

REPORTING AND VERIFICATION (MRV) MANUAL. Forestry and Other Land Use AGB Aboveground biomass BCEFs Biomass conversion and expansion factors BRDF Bi-directional reflectance distribution function BURs Biennial Update Reports CH4 Methane CI Conservation International CMP Conference of the Parties serving as the Meeting of the Parties to the Kyoto Protocol CO2 Carbon dioxide COP Conference of the Parties CV Coefficient of Variation DBH Diameter at Breast Height DEM Digital Elevation Model DTs Decision Trees EFDB Emissions Factor Database EFs Emissions Factors EM Electromagnetic EOS Earth Observation System EPA Environmental Protection Agency FAO Food and Agriculture Organization FAS Fire Alert System FCMC Forest Carbon. Markets and Communities Program FCPF Forest Carbon Partnership Facility REDD+ MEASUREMENT. VERSION 2. ACRONYMS AND ABBREVIATIONS ACR American Carbon Registry AD Activity Data AFOLU Agriculture.0 v .

Technology and Innovation MMU Minimum-mapping unit MRV Measurement. VERSION 2. Land-use Change and Forestry GPS Global Positioning System IDEAM Colombian Institute for Hydrology. Meteorology and Environmental Studies ILUA Integrated Land Use Assessment INPE Brazilian National Space Research Institute IPCC Intergovernmental Panel on Climate Change KCA Key Category Analysis LDCM Landsat Data Continuity Mission LEDS Low Emission Development Strategies LiDAR Light Detection and Ranging LUC Land-use Change MADS Colombian Ministry for Sustainable Development MCT Brazilian Ministry of Science. Reporting and Verification N20 Nitrogen oxide NAMA Nationally Appropriate Mitigation Strategies NASA National Aeronautics and Space Agency REDD+ MEASUREMENT.0 vi .FIRMS Fire Information and Resource Management System FREL Forest Reference Emission Level FRL Forest Reference Level FSI Forest Survey of India FUNCATE Foundation of Space Science. REPORTING AND VERIFICATION (MRV) MANUAL. Applications and Technology GEF Global Environmental Facility GFIMS Global Fire Information Management System GFOI MGD Global Forest Observation Initiative Methods and Guidance Documentation GFW Global Forest Watch GHG Greenhouse gas GHGMI Greenhouse Gas Management Institute GIS Geographic Information System GLAS Geoscience Laser Altimeter System GOFC-GOLD Global Observation of Forest and Land Cover Dynamics GPG-LULUCF Good Practice Guidance for Land Use.

NCs National Communications NFMS National Forest Monitoring System NGGIP National Greenhouse Gas Inventories Program NGO Non-governmental organization NNs Neural Networks NRT Near-real Time PCA Principal components analysis PRODES Projeto De Estimativa De Desflorestamento da Amazoni (Brazilian Amazon deforestation monitoring program) QA/QC Quality Assurance and Quality Control QUICC Quarterly Indicator of Cover Change RADAR Radio Detection and Ranging REDD+ Reducing emissions from deforestation and forest degradation. and consistency TOA Top-of-atmosphere UMD University of Maryland UNDP United Nations Development Programme UNEP United Nations Environment Programme UNFCCC United Nations Framework Convention on Climate Change USAID United States Agency for International Development USGS United States Geological Survey VCS Verified Carbon Standard WGs Working Groups WMO World Meteorological Organization WRI World Resources Institute REDD+ MEASUREMENT. completeness. comparability. SBSTA Subsidiary Body on Scientific and Technical Advice SES Social and Environmental Soundness SINA Colombian National Environmental System SLR Side Looking RADAR SRTM Shuttle Radar Topography Mission TACCC IPCC principles of transparency. sustainable forest management and enhancement of forest carbon stocks. plus the role of conservation. REPORTING AND VERIFICATION (MRV) MANUAL. VERSION 2.0 vii . accuracy.

including Colin Silver. REDD+ MEASUREMENT. Chris Potter of the US National Aeronautics and Space Administration’s Ames Research Center. Carly Green and Jim Penman of the Global Forest Observations Initiative. Christine Dragisic of the US State Department. and members of the USAID Climate Change office. The authors also thank those who have made graphics available for use in this Manual. SilvaCarbon Consultant. Deborah Lawrence of the University of Virginia. Karyn Tabor. REPORTING AND VERIFICATION (MRV) MANUAL. Maggie Roth. Megan McGroddy. Michael Gillenwater of the Greenhouse Gas Management Institute. Asim Banskota of the University of Minnesota. Brice Mora of the Global Observation of Forest Cover and Land Dynamics. Rishi Das.0 viii . VERSION 2. Ronald McRoberts. Frank Martin Seifert of the European Space Agency's Centre for Earth Observation ESRIN. and Leif Kindberg of FCMC. Mario Chacon and Johnson Cerda of Conservation International. Additional figures are from the websites of the United Nations Framework Convention on Climate Change. John Rogan of Boston University. the International Panel on Climate Change and the Landsat program. and Charles Scott of the US Forest Service. including Ned Horning of the American Museum of Natural History. Andrew Lister. and Jamie Eaton of TerraCarbon LLC. Axel Penndorf of BlackBridge.ACKNOWLEDGEMENTS The authors thank the various colleagues in our organizations who have commented on the text.

1 INTRODUCTION Section 3. 4. of this Manual outlines the sequence of steps required for generating a national GHG inventory. REDD+ MRV MANUAL: CHAPTER 4. Inventory and Reporting Steps.0 – FIELD-BASED INVENTORIES 50 . This chapter is relevant to the activities highlighted on the following page. David Scoch 4.3. Irene Angeletti.0 FIELD-BASED INVENTORIES Authors: Gordon Smith.

appropriate to the tier level identified.0 – FIELD-BASED INVENTORIES 51 . through stratification and other methods. STEP 2: Conduct key category analysis (KCA) for the relevant categories. Within the categories designated as key. AD represents the extent over which a human activity occurs. Document and archive information used to produce the national emissions and removals estimates following specific instructions under each land- use category. pool or non-CO2 gas. STEP 3: Design a forest carbon inventory to generate Emissions Factors (EFs). if using the gain-loss method. assess significant non-CO2 gases and carbon pools and prioritize such pools in terms of methodological choice. EFs represent coefficients that quantify the emissions/removals per unit STEP 4: Generate Activity Data (AD). using the reporting tables. REDD+ MRV MANUAL: CHAPTER 4. STEP 5: Quantify emissions and removals. STEP 6: Report emissions and removals estimates. and worksheets where appropriate. carbon pool and non-CO2 source. for representing areas in the GPG-LULUCF.STEP 0: Establish Institutional Arrangements. estimating the uncertainty in each estimate. for the time period required. STEP 1: Estimate the land areas in each land-use category. STEP 7: Perform verification and implement quality control checks. ensuring that the requirements in terms of emission and removal factors are met. Emissions and removals estimates represent the product of the AD by the associated EFs. including expert peer review of the emission estimates following specific guidance under each land-use category.

accuracy.1 provides an illustrative inventory design and implementation work plan. comparability. Countries will need to perform these processes according to the Intergovernmental Panel on Climate Change (IPCC) principles of transparency.According to United Nations Framework Convention on Climate Change (UNFCCC) guidance for reducing emissions from deforestation and forest degradation. sustainable forest management and enhancement of forest carbon stocks (REDD+). and supporting the generation of GHG offset credits or national programs to mitigate emissions. anthropogenic forest-related greenhouse gas (GHG) emissions by sources and removals by sinks. A field-based forest carbon inventory has multiple purposes. completeness. as appropriate. REDD+ MRV MANUAL: CHAPTER 4.0 – FIELD-BASED INVENTORIES 52 . forest carbon stocks and forest-area changes. and consistency (TACCC). countries will have to establish National Forest Monitoring Systems (NFMS) that quantify changes in land cover and terrestrial carbon stocks using a combination of ground-based forest carbon inventory approaches for estimating. Box 4. plus the role of conservation. facilitating national communication of carbon emissions and removals from land use. including providing accurate input into a national GHG inventory. When a forest carbon inventory can serve multiple needs. it will likely be easier to obtain resources to prepare the inventory and maintain support for continued work over time.

Set up inventory database 5. and can serve other land management. Collect field measurement data and enter to database 2. Select and acquire measurement technologies/equipment 3. fixed area. pilot plot data from existing inventories. Submit results to external technical expert verification (as part of larger UNFCCC REDD+ verification process) Forest carbon inventory data have substantial overlap with timber inventory data. the Mexican National Forest and Soil Inventory carried out an effective consultation process to identify the information that various users of the inventory would require. wildlife.0 – FIELD-BASED INVENTORIES 53 . permanent or temporary) and sampling intensity. Data checking/quality control 2. Conduct periodic internal audits (via plot re-measurements) through implementation to confirm adherence to protocols and identify and correct any measurement errors • Analysis and reporting 1. This type of data sharing can make the inventory more cost effective and ensure financing from more sources.1: Illustrative inventory design and implementation work plan • Establish Institutional Conduct: a capacity and needs assessment through a consultative process engaging a wide range of potential users of data (beyond REDD+ MRV end uses) 1. Define population/forest area c. since multiple information users can advocate for its continued funding. systematic. Allocate samples and produce maps for inventory implementation f. Select desired precision /allowable uncertainty 3. variable radius. the latter with consideration to budget constraints. Define budget and personnel and capacity building needs • Design inventory 1.g. Select sampling approach (e.g. maps and relevant GIS coverages. Define forest carbon pools (informed by KCA analysis) and other environmental parameters to measure 2. It may be possible to extend the use of collected data. Select and validate allometric equations and define required measurements 3. two- stage) d. nested fixed area. Collect data to develop sample design (e. and land-use management needs. simple random. 2. Define inventory organization/administration 2. Define objective. Box 4. Document design and rationale • Prepare for inventory implementation 1. For example. aerial photographs and remote sensing data) b. Develop field measurement and data management protocols 4.g. or extend the geographic range of a forest carbon inventory and thereby jointly serve carbon inventory and other resource management needs. specifying end uses and desired outputs. Analyze data and calculate estimates and uncertainty 3. The national forest inventories supported by the Food and Agricultural Organization (FAO) National Forest Monitoring and Assessment REDD+ MRV MANUAL: CHAPTER 4. forest heterogeneity and plot size e. Produce reports and summary results 4. stratified. cluster. Organize and train field and administrative personnel 6. desired precision. Develop sample design a. Produce tactical plans to guide deployment of field teams • Implement inventory 1. Consider scale/resolution desired and whether focus is on estimation of stocks or stock change. Define sample unit (e. desired scale/resolution.

the stock change method does not produce emission factors and activity data. the information below highlights considerations specific to forest inventories. Besides the collection of information regarding timber species and volume. There has been no decision for non-Annex I Parties regarding use of the 2006 Guidelines.g. 2006).. non-timber forest products and socio-economic indicators 13. and implementation of sampling schemes. For perspective. In the latter case. and facilitate mapping. and best suited for landscapes where the area of forest undergoing transitions in structure is either difficult to delineate or relatively evenly distributed across large areas. REDD+ MRV MANUAL: CHAPTER 4. Other resource management goals may be addressed by having teams collect some additional types of data while doing their carbon inventory work.9 below. which focuses on identifying and measuring fluxes. 13 See http://www. mortality).1 of the Global Forest Observation Initiative (GFOI) Methods and Guidance Document (GFOI 2013).2 CARBON POOLS AND THEIR MEASUREMENT REDD+ forest carbon inventories should quantify stocks of carbon in pools that might change significantly under the REDD+ mechanism or under the REDD+ reference 14 The GPG-LULUCF (IPCC. and may eventually encompass an entire country. and thus in this Manual we use the former term throughout. In contrast. the stock change method is less targeted than the gain-loss method. A further discussion of gain-loss and stock change methods is included in section 2. While the different carbon pools are described in Chapter 3. 2003) uses the term stock change.g. Sampling and measurement guidance presented in this chapter is relevant to both the stock change and gain-loss methods. 4. This chapter also focuses on the derivation of country or region-specific data that could be used with a Tier 2 or Tier 3 method. In comparison. gain-loss. 2006). new model-based approaches in development. relating biomass to remote sensing data (see Chapter 5.program provide another example. derived by delineating the area of those transitions. these inventories collect data on carbon stock. Field-based measurements are critical elements in both traditional probability-based sampling and in model- based estimation approaches involving interpretation of remote sensing imagery. Geographic Information Systems (GIS) and remote sensing approaches (see Chapter 5) are part of this system.fao. while the Guidance from 2006 uses stock ddifference (IPCC. intact to degraded forest). As discussed in Chapter 3. Reporting and Verification (MRV) should be at a corresponding scale. meaning that inventories covering more area become less expensive when the cost is calculated on a per-hectare basis. The gain-loss method. and Measurement. the IPCC recognizes two methods to estimate carbon changes: the gain-loss method and the stock change method 14 and establishes hierarchical Tiers of data specificity (IPCC. Tier 1 methods rely on the use of default values. There are economies of scale in forest carbon inventories. measurements are used to produce emission factors associated with specified transitions in forest structure (e. diameter increment) and losses (e. In this context. forest inventory is most accurately conceived of as a collection of procedures and technologies that encompass a system. stratification. which is then expanded using activity data. or to estimate gains (e. but not replace. even though the 2006 Guidelines are more up-to-date and use the latter.g. Repeated measure from sample plots can be used to estimate stocks at two points in time. but instead produces overall estimates of stocks for the entire area under MRV (held constant) for two points in time and estimates emissions as the difference between the two estimates.0 – FIELD-BASED INVENTORIES 54 . “traditional” field-based measurement practices and probability-based approaches. REDD+ activities may encompass vast areas of land. This chapter focuses on implementation of the stock change method. hold promise to complement. is discussed in Section 4. emerging technologies).

REDD+ MRV MANUAL: CHAPTER 4. Pieces that are not large enough to be classified as coarse debris are classed as fine debris. dead woody stems where the long axis of the stem is within 45 degrees of vertical are classified as standing dead and stems where the long axis is more than 45 degrees off vertical are classified as lying dead wood. Forest inventories limited to commercial species or trees in commercial (e. To obtain the belowground biomass. tallying trees with diameters of at least 10 cm. smaller woody plants and non-woody plants are excluded or are measured separately from larger woody plants.g. and is neither double-counted nor excluded. Standing dead wood is typically measured along with living tree biomass (see aboveground biomass) and recorded as deadwood because its density often differs from live trees. and the shrub/small tree category includes woody plants at least 10 cm. a sub-component of dead organic matter. the inventory should include all tree species. Therefore. it may be important to measure dead wood carbon stocks (described in 4. 4. If there is substantial disturbance of forests via degradation. fine roots are often excluded.2 Belowground biomass Belowground biomass is an important carbon pool that may equal 25 percent or more of the aboveground biomass in many forests.g. there may be sub-categories such as small trees that are 10-40 cm DBH and large trees that are greater than 40 cm DBH. If litter is measured. To increase sampling efficiency. If forest land is converted to agricultural use or developed use. multiply the aboveground biomass by 1 + root-to-shoot ratio. By convention. with smaller pieces being classified as litter. so that any piece of material fits in exactly one category.0 – FIELD-BASED INVENTORIES 55 . The IPCC recommends 2 mm diameter.All inventories should measure live trees above a modest size because aboveground biomass (described in Section 4. To accurately estimate the forest carbon stock of live biomass..1) is typically the largest biotic carbon pool in a forest that would be lost via deforestation. An example of a default set of size categories is one where the tree category has a diameter at breast height (DBH) of at least 10 cm. substantial amounts of soil carbon (described in 4. The boundary between fine and coarse roots depends on the method used to estimate the belowground biomass. A common minimum size of fine debris pieces is 1 cm. over 30 cm DBH) diameter classes often exclude important components of total aboveground biomass and consequently offer limited reliability in estimation of carbon stocks and stock changes.2.2. in some systems.2. Typically. 4. or measured as part of the soil carbon pool due to the difficulties in Manually separating them from soil. includes standing and lying deadwood.5) may be lost. e. As discussed in Chapter 3.3 Dead wood Dead wood. the boundary definition must correspond to the smallest piece of woody debris.4 of the IPCC 2006 Guidelines.2.3). but measuring root biomass is time consuming and expensive. soil carbon loss resulting from conversion of forest to agricultural cropland can be greater than emissions from aboveground biomass. and thus monitoring of soil organic carbon stocks may be warranted. 50 cm or 1 m tall. peat swamp forests. 4.2. A typical minimum piece size for coarse woody debris is 10 cm in diameter and sometimes there is also a minimum length requirement that pieces be at least 1 m long.2.1 Aboveground biomass In a forest. aboveground biomass typically emits the most carbon upon conversion to non-forest. However. REDD+ inventory developers may choose to apply a Tier 1 approach for belowground biomass which uses the default root-to- shoot ratios provided in Table 4.

particularly if there is no fine woody debris pool. except for sites that are so arid that few trees are likely to grow and sites with carbonaceous soils such as limestone. but if this pool occurs infrequently it may be included in the litter pool. or larger than. but typically the boundary is defined as a specific piece size or degree to which pieces are buried. By convention. foliage and twigs that are on the ground and not attached to a plant stem. may not be worth measuring since the pool is not typically large. By convention. The threshold size of roots and dead wood to be included in the soil carbon category must correspond to definitions used in the live belowground biomass and dead wood categories. in more arid woodlands. Typically. and wet forests.” includes fine woody debris. Although there are often measurable amounts of soil organic carbon down to depths of several meters. As a result. material less than 1 cm in diameter are defined as litter. Whatever boundary is chosen. Total soil carbon stocks are often as large as. soil carbon stocks may have large changes and should be measured.5 Soil organic matter As discussed in Chapter 3. many projects that maintain existing forest do not measure soil carbon stocks because the stocks are assumed to be constant. If measuring soil carbon. and as a result. However. but some projects have measured soil carbon to 1 m depth or more. better described as the “forest floor. while muck is black. However. Fine woody debris is small pieces of dead wood. in the case of forest clear-cutting and conversion to agriculture. an organic layer may form. at depths up to one meter or more. carbon is generally counted if it is in the top 20 or 30 cm of soil. Guidance provided by the IPCC specifies separate reporting of changes in organic soil carbon stocks and inorganic soil carbon stocks (but change in inorganic soil carbon stocks is assumed to be zero unless Tier 3 methods are used) (IPCC 2006). A humic layer of organic soil is the decomposed remnants of vegetative material and is typically not included in the litter pool. low pH. particularly if the agricultural activities include plowing. it should be measured separately from the litter and the mineral soil carbon pools. In closed. if inorganic soil carbon is likely to be present. At depths of more than one to three meters. . the litter pool may be defined as including fine woody debris up to 10 cm in diameter. or nutrient limitation.2. and changes in the stock are slow. In the years to decades following conversion to plowed fields. As a result. However. litter tends to decompose easily. The density of soil carbon decreases with depth. substantial carbon loss can occur in the deeper soil. woody biomass carbon stocks.4 Litter Litter. In undisturbed systems. if decomposition is slowed by factors such as cold temperatures. the density of soil carbon is low.2. If a significant decomposed organic layer is present between the litter and the mineral soil.4. but the total amounts can be significant because the mass of soil is so large. there is no inorganic carbon in soils. the same boundary must be used for the maximum size of pieces in the litter pool and the minimum size of pieces in the woody debris pool. decomposed organic material. When switching from plowing to trees REDD+ MRV MANUAL: CHAPTER 4. this category includes all organic carbon in mineral and organic soils to a specified depth. most of the change in soil carbon happens near the surface in the top 20-30cm. If there are small or modest degrees of disturbance of the forest. It is typically included in the soil pool. moist. However. soil carbon stocks are unlikely to change much.0 – FIELD-BASED INVENTORIES 56 . and live roots of 2 mm or greater are classified as belowground live biomass. 4. laboratory tests that do not differentiate organic from inorganic carbon should be avoided. and the amount of effort required to sample soil carbon increases with depth. there is more soil carbon per centimeter of depth at the surface than there is at 40 or 100 cm below. moisture saturation. live roots less than 2 mm in diameter are often classified as part of the soil carbon pool. There is less standardization of the definitional boundary between soil carbon and woody debris. a key decision is the depth to which soil will be measured. For some forest types. as well as live fine roots that are above the mineral or organic soil. Examples of this include peat and muck soils. peat is a buildup of minimally decomposed plant material. typically 40-50% of this carbon is lost.

When measuring soil carbon loss upon conversion of forest or grassland to cropland with plowing. This shallow sampling is especially common for inventories focused on detecting carbon stock increases.1 Needs assessment The first decision in designing an inventory is to assess what needs to be known as a result of the inventory.3. while limiting sampling effort. dry forest. the process of designing an inventory involves a sequence of steps: 1) Needs assessment 2) Sample design selection 3) Plot design considerations 4) Cost assessment and specification of sampling intensity 4. Producing accurate estimates of stocks and changes for all the various strata will significantly increase costs and may not prove the most efficient approach. NFMS “should build on existing systems. Initially. Consideration of inventory design should also take account of. In moist forest. A country should aim to understand trends for specific regions or forest types. such as 50 or 100 cm. even when measuring to significant depths. and grassland areas where there is conversion to agriculture. This requires choosing what is to be estimated. or forest subject to anthropogenic change. As a result. shrubland. In wet forests and coarse texture soils that have little carbon. per COP19 (Decision 11/CP. Therefore. Much of the gain in soil carbon in the first few years of conversion from crops to forest is in the top few centimeters of soil. The next decision is whether there are sub-divisions within the total area where extra information is needed.3 CONCEPTS AND CONSIDERATIONS IN INVENTORY DESIGN Many possible inventory designs can be used to estimate forest carbon stocks. or something in between.” Radically new approaches or REDD+ MRV MANUAL: CHAPTER 4. it can be efficient to only measure soil carbon to 20 or 30 cm depth.19 paragraph 4(a)). existing capacities and experience of personnel that will be involved in both field measurement and analysis. Larger changes in percentage terms are easier to detect with sampling. soil carbon losses are usually large enough to justify the cost and difficulty of measuring soil carbon loss to 1 m depth. REDD+ monitoring activities may also only focus on lands that are classified as managed forest. In general. avoided deforestation projects may sample soil much deeper than 30 cm. to be able to claim credit for avoiding emission of deeper soil carbon. In conversion of tilled cropland to grassland there can be significant carbon gains to more than 1 m of depth and it may be worth conducting deep sampling. such as to a depth of 1 m. 4. the gain in percentage terms is greatest when only the top few centimeters are measured. and over what geographic areas. To capture much of the carbon stock change that results from land management changes. The geographic scope might be a particular block of land amounting to only a few dozen hectares. many inventories sample only the top 20 or 30 cm of soil.0 – FIELD-BASED INVENTORIES 57 .or no-till cropping. an entire country. the percentage change in carbon stock may be large. and seek to build on. The goal is to choose an efficient design that achieves the desired level of precision at a minimum cost. it is possible that half the soil carbon gain in the first five to 10 years will be in the top 10 cm of soil.

0 – FIELD-BASED INVENTORIES 58 . systematic (right). and iv) two stage sampling. ii) systematic sampling. and stratified random (bottom) sampling.2 Sampling design selection Sampling must be unbiased to ensure that resulting inventories will be reliable.technologies should be introduced incrementally and only after trial applications can foster confidence of responsible personnel. Diagram of random (top). iii) stratified (random or systematic) sampling.1. Four common approaches. 4. illustrated and elaborated below in Figure 4. REDD+ MRV MANUAL: CHAPTER 4. Figure 4. There are many options available for developing a sampling design.3.1 are: i) random sampling..

Random Sampling A random sample approach locates plots within a study area at random. inventories must seriously consider perceptions of both specialists and non-specialists. Many people find systematic sampling attractive because it gives equal emphasis to all parts of the area being sampled. systematic sampling involves laying a regular grid over the geographic area to be inventoried.0 – FIELD-BASED INVENTORIES 59 . it is common to stratify by ecotype or forest type.S. The spacing of the grid lines is calculated so that the desired number of plots can be placed in the area. Training should include communication of the importance of measuring the various conditions that occur over the geographic extent of the inventory. Systematic Sampling Many national forest inventories instead use a systematic sampling design. about 20 percent of the points have been re-measured. Mexico provides another example. which. A variation on systematic sampling is to randomly locate one plot within each cell defined by the grid lines (as is done in the U. For example. such that all points are monitored once every five years (GOFC-GOLD. then the stocks of the strata are summed to estimate the stock (or stock change) of the entire area. and that quantifying this variability is a key part of quantifying how reliably the samples reflect the forest as a whole. Systematic sampling offers significant improvement over simple random sampling in terms of reduced uncertainty. From 2008 onward. Stratified Sampling Stratified sampling is accomplished by dividing the sampling area into relatively homogenous sub-areas. non- specialists may not fully understand bias and may have difficulty accepting random sampling and protocols that result in plots being located in sites that the non-specialist views as not representative of the forest. where regularly spaced plots are measured. Each point contains four sites of 400 m2. While many approaches to stratification exist. giving a more precise estimate of carbon stocks for a REDD+ MRV MANUAL: CHAPTER 4. A systematic sample ensures that all geographic areas are equally represented. It is important to understand that stratified sample designs do not produce estimates at the stratum level with equal precision as for the overall population – the goal of stratification is to distribute sampling effort more efficiently to produce a population-level estimate. where the National Forest and Soil inventory established a systematic sample grid of 25. Carbon stock (or stock change) is estimated for each stratum. Typically. The primary advantage of random sampling is that the calculation of means and uncertainty is simple. Within each stratum. 2013). An example of a recent national forest inventory performed using systematic sampling is the Integrated Land Use Assessment (ILUA) carried out by the Zambia Forestry Department (2005-2008). The ILUA set up 221 tracks (each track has 4 sampling plots) systematically across the country at 50 km distances. random allocation of plots often presents challenges to navigation in the field and produces plot distributions that invariably under- sample some areas and over-sample others. giving more precise estimates for the same or less effort. may appear to lack credible coverage. In practice. and is especially useful if little is known about forest conditions or dynamics. Its intuitive layout also facilitates navigation in the field. though the calculation of uncertainty for systematic samples is less straightforward than for random samples. and separately sampling each sub area.000 geo-referenced permanent points. and locating plot centers at the grid intersection points. This approach to stratification increases statistical power. a systematic sample or simple random sample is conducted. with the first intersection point located at random. while un-biased. Stratification increases efficiency of sampling. Forest Service Forest Inventory and Analysis program).

2011) is maintained within each stratum. using the same calculation as a pre-stratified sample introduces some additional uncertainty due to the random and changing sample sizes within strata. the selected sampling intensities in these areas may be less than in accessible areas. the stratification can be done initially using global ecological datasets. sampling efforts can be focused towards more variable ecotypes or forest types.5 (GFOI 2013). Calculation of variance for a post-stratified sample is presented in Cochran (1977). The goal is to "block" differences into different strata.fao. as is often the case with degradation. and where sufficient sample size (of 10 or more plots. and requires that the new strata areas are known exactly and that the existing plots can be unambiguously re-assigned to the new strata (e. To stratify within each land-cover type. areas with different sampling intensities constitute different strata. World Wildlife Fund Ecoregions 16.. one can use various GIS data on elevation.g. Westfall et al. Stratifying by expected change in carbon stock can be desirable. 4.given number of plots of a given design and also increases the likelihood that plots within a particular stratum will be similar to other plots in that stratum. through an original systematic sample).4.g.When choosing strata. The added uncertainty produced by post-stratification tends to be small.g. and such areas should be identified and delineated as separate priority strata. allocation of plots should be weighted toward areas where carbon stocks are susceptible to decrease from degradation or deforestation. These include indices based on 15 http://geodata. In this having lower variance between plots gives a higher probability that the total carbon stock will be close to the carbon stock estimated from the sampling.. Various remote-sensing-based data can also be used.cfm 17 http://www. through monitoring of forest cover and change classes with remotely sensing). therefore.1). section 1. particularly where proportional allocation can be roughly maintained (e. Thus. If the goal of an inventory is to precisely quantify changes in forest carbon stocks. For a given number of plots. Post-stratification involves re-allocating sample plots among strata. after the original plot allocation has taken place. soils or other parameters. but for this forest to be different from other strata. the goal is to have relatively homogeneous forest (in terms of structure).worldwildlife. Chapter 3.unep.. Stratification approaches are further discussed in the GFOI Methods and Guidance. Within each stratum. While this post-stratification approach is straightforward in terms of estimation of the mean. ecological zone. or by deploying supplemental samples. combining forest types with small areas and sample sizes).org/geonetwork/srv/en/main. Stratification is generally performed ex ante. or forest with the same carbon stock dynamics. but strata can also be redefined ex post via post-stratification and plots re-allocated in a changing landscape. developers should consider what is known about the forest and the dynamics of carbon stock change. and may be essential if net changes in stocks are small compared to total stocks. To stratify a country’s forest.home REDD+ MRV MANUAL: CHAPTER 4. This would mean stratifying by the expected future change in carbon stock. it is first necessary to have a current map of the dimension being used to stratify. Of equal or more importance. and allocating more sampling effort to strata that are expected to have greater change over time (decrease or increase). 16 http://www.2. such as maps of Holdridge life zones 15. Homogeneous strata need few plots to precisely estimate their carbon stocks and. This may be a national forest benchmark map or some other valid map source. the cost of placing large numbers of plots in remote inaccessible forest areas may be prohibitively expensive and/or logistically impossible.0 – FIELD-BASED INVENTORIES 60 . If no previous information on forest types exists in the country. and FAO ecological zones 17.grid. The latter can be ensured by consideration in the definition of strata (e. The IPCC recommends stratifying by climate. or increase from forest growth and regeneration. and management practices (Vol.3.

yet require substantially different levels of field effort to accomplish them: 67 sample plots for the stratified design (Design 3). like strata 3 and 4.000 10. Discussions of these can be found in Chapter 5. which ignores sources of variability within the population.000 31.0 83.1.0 – FIELD-BASED INVENTORIES 61 .0 37.3 100.6 43.1: Comparison of required sampling intensities for stratified and unstratified sample designs Design 1 allows for reporting with the targeted precision at the individual stratum level by implementing an independent sampling effort in each forest stratum. The advantage of stratification is illustrated in comparing Designs 2 and 3. They also include classifications or derived products. as travel between sample units represents the largest time and labor investment in forest inventory. as shown in Design 3.000 12. not at the stratum level.6 27. The value of stratification is illustrated with the following example (Table 4. achieves the targeted precision at the population level.500 Pilot observations 20 90 65 50 30 110 25 85 70 180 95 120 120 140 40 75 25 75 90 170 45 110 65 100 Mean 51. in that stratification.3 41. Stratum 1 Stratum 2 Stratum 3 Stratum 4 Total Area (ha) 2. This can be achieved using REDD+ MRV MANUAL: CHAPTER 4. highly variable strata covering small areas. This is an important consideration. and 110 plots for the unstratified design (Design 2). will have less influence on the total number of plots needed than those strata that cover larger areas.1). in which the sample sizes required to target 95 percent confidence intervals equal to ± 10 % of the mean are calculated for a dataset both with and without stratification. like stratum 1 in Table 4.1 Standard deviation 38. such as global biomass maps produced by research programs. Two stage sampling Sampling vast hard-to-access landscapes necessitates efficient deployment of field effort.6 CV 74% 32% 43% 42% 52% Design 1 Sample size 217 41 74 69 402 for 4 independent forest inventories Design 2 Sample size 110 (not stratified) Design 3 Sample size 6 16 20 25 67 (stratified with Neyman allocation among strata) Table 4. For example.7 117.spectral and/or temporal information. such as greenness indices and seasonality indices.5 63. both of which are designed to generate estimates at the population level.500 7.

Inventory developers will also need to consider whether plots will be permanent or temporary. A transect plot design.0 – FIELD-BASED INVENTORIES 62 . and the second step involve allocating sample points within selected polygons. it concentrates field effort and results in minimizing overall effort. the first step might involve selecting polygons to sample from a population of polygons. • Areas (two dimensional): all the trees found on a determined area are measured. Typically. Often these plots are called "fixed area" plots because the size is fixed.a two-stage sample. • Lines (one dimensional): on a sample line 18.2: Sampling design including nested plots and line samples to measure the different types of carbon pools. (2007). While two stage sampling introduces an added source of sampling error. even if long and narrow. For example. involving two steps in sampling. From Smith et al. fixed area plots are circular or rectangular (including transects). This method can be used to calculate the volume of coarse woody debris. it can be observed how many features intersects the line.3 Plot design considerations Plot design considerations include the type of plot to be used.3. typical options for a plot design are: • Points (dimensionless): variable radius plots can be implemented using a relascope or wedge prism. Figure 4. and the variables that need to be measured. Each is discussed below. as well as the size of the plot. 18 It is important to note that a sample line plot design isn’t the same as a transect plot design. Plot types In a forest inventory. 4. where a tree is determined to be in or out as a function of the ratio of its diameter to distance from plot center. REDD+ MRV MANUAL: CHAPTER 4. is an area (two-dimensional).

2). not the values observed on individual plots.0 – FIELD-BASED INVENTORIES 63 . Therefore. Schreuder et al. An example of a cluster pattern would be five plots per cluster where one plot is centered on the central point of the cluster. they are difficult to use in dense stands of small trees (e. and the variability used in calculations of plots needed for the inventory will imply an approximate plot design. Fixed area and variable radius plots are equally valid choices for a forest inventory (Grosenbaugh and Stover. early successional forest). While variable radius plots are more effective in directing sampling effort to the most influential elements in a population (large trees). the cluster is simply a larger. targeting different diameter classes. the sample unit is the cluster. effectively. the key issue is length. It is often advisable to keep with tradition to avoid extended learning periods and consequent measurement errors. Forest inventories report measurements over horizontal areas.g. in forests where visibility and penetrability is limited. Also.. transects tend to cover more site conditions. These issues determine the “size” of the plot. 1957. such as combining lines with fixed area or point plots within which to sample lying dead wood and aboveground biomass.. For area plots. respectively (Figure 4. can account for the fact that distances measured along a slope are smaller when projected into a horizontal map plane. Data is analyzed using cluster means. where plot size is increased. and surpassed. Plot types and sizes should be kept constant within strata. the key issue is the prism factor. In contrast. which sample trees with probability proportional to size. using variable radius plots. and how many trees are included. Where plots overlap the REDD+ MRV MANUAL: CHAPTER 4. allows for increased efficiency in plot measurements as they distribute measurement effort more evenly across diameter distributions. it is preferable to have circular plots because they have the smallest perimeter for the same area.e. reducing the amount of border trees. increasing the variability within plots. Plot size affects the variability of carbon stocks observed on different plots. Fixed area plots can have different shapes: circular. and consequently have lower inter-plot variability. These typically need be applied only if the slope is greater than 10 percent. thus. Normally. plot size is simply the area encompassed by the plot. in the cardinal directions. and the remaining four plots are located with plot centers 200m away from the cluster center. Nesting designs that incorporate fixed area subplots of different sizes.. dis-aggregated plot. However. In the field. For prism plots. not the plot. required sampling intensities (see below) can be reduced by increasing plot size. the choice of fixed area versus variable radius plots depends on the level of comfort and familiarity of personnel. The typical diameter distribution in a mature forest has a negative J shape (i. transects tend to be preferable to facilitate accessibility to the entire plot and ensure that trees do not remain uncounted. different plot types may be used in different strata to accommodate different forest structures and logistics. Plot size Plot “size” is assessed differently for different plot types. Larger plots capture more variation in forest structure. For line plots. Slope corrections. The same efficiencies are achieved.Plots of different types may also be combined to achieve efficiencies in measuring different forms of forest biomass. Effectively. Defining the sample unit as a cluster of plots may also be used as a means of reducing inter-sample variability. it is recommended to analyze actual plot data to estimate the variability that will result if different plot sizes are selected for different sizes or types of trees. plots may need to be adjusted to account for slope and boundary considerations. a very high number of small trees and fewer larger trees). When designing an inventory to achieve a target level of precision. square. and achieves the same reduction in inter-plot variability as larger plot sizes. Most importantly. which is the ratio of diameter to distance from plot center that determines whether or not a tree is measured. or rectangular. 1987).

for the baseline number of plots needed. corrections are applied to produce equivalent measurements for a complete sample (e. when visibly marked. a metal stake may be placed completely in the ground for re-identification with a metal detector (Smith et al. not by measurements at the plots. forest variability and plot size employed. when forest carbon is being measured. including budget constraints. Thus. 2004). Diaz.g. An advantage of temporary plots is that both stratum boundaries and the intensity of sampling can be easily changed over time. the number of samples measured) predicates the precision and resolution of estimates that can be achieved by a forest inventory.0 – FIELD-BASED INVENTORIES 64 . First. 4..3. The goal of sampling is to reach a desired precision of the estimate of carbon stocks for an acceptable cost. A range of inventory designs should be compared with representative cost data to find a design that meets the specified needs at an acceptable cost.4 Cost considerations and specification of sampling intensity The sampling intensity (i. The majority of field effort is represented by travel between plots. by 5 to 20 percent. providing a cushion in case some permanent plots cannot be relocated or land cover changes. can improve efficiency in field effort. Consequently. It is strongly recommended to work in reverse. Selecting the sampling intensity depends on a range of factors. getting extremely precise estimates may become expensive. Typically. every five years). approaches that minimize travel time. Temporary plots are often used in timber inventories. it is good practice to increase the minimum number of plots. Second. with variable radius plots. Inventory costs are driven by variable costs that are a function of field effort. 2011). Permanent plots versus temporary plots Permanent plots.g. When establishing permanent plots. allow for estimating the stand growth and disturbances with more precision (for a given number of plots) and can therefore quantify small increases or decreases in stocks. more plots yield lower sampling errors. Some general principles dictating required sampling intensity must be considered. and thus determine what data should be collected. The greater the variability of the forest. There is a risk that plots. desired precision. To reduce uncertainty by half can require four times as many plots.boundary of an inventory area. it is necessary to detect the magnitude of change in carbon stocks over a short period of time such as five years or less. including two stage samples and cluster sample designs. 2007. may be treated differently by forest users or plantation managers and. desired scale/resolution of estimates. Parameters to measure The parameters that will be measured in the sample plots depend on the carbon pools of interest and the allometric equations that will be used to convert tree measurements into biomass. first considering what needs to be known as a result of the inventory. REDD+ MRV MANUAL: CHAPTER 4. the more plots will be needed to obtain a given level of precision. the statistical precision of a biomass estimate depends on the variability of the forest. Ducey et al. therefore.e... and attempting to quantify annual changes in forest carbon stocks resulting from human activities can be confounded by weather and wildfire. Note that weather and disturbance events can cause annual changes in forest carbon stock that are larger than anthropogenic changes. then considering the analysis steps. it may be desirable to mark plot centers with monuments that are not visible to the human eye.. a walk through method (Avery and Burkhart 1994. For example. using a mirage method) or. which are re-measured periodically (e. back to the plot data..

The CV is a measure of how different plots are from each other. In such a case. the CV can be less than 30 percent. one must choose an estimate of variability between plots. greater statistical precision might be obtained by using fewer and larger plots than the theoretical optimum calculated without considering costs. In this case. with 100 percent of plots measured every five years. In fully stocked plantations. larger plots may average out some of the fine-scale variations in forests. the number of plots is relatively independent from the size of the area. the cost of increasing the size of existing plots no longer yields a significant reduction of variance when compared to that which could be achieved by increasing the number of plots to the sample. As explained previously. the inventory developers will have to consider the density of large trees in the forest and the range of sizes of gaps. the CV is the standard deviation divided by the mean. calculated as the coefficient of variation (CV). For small plots in forest with gaps. The chosen variability implies a plot size. +/. These statistics are discussed in Section 4. Many inventories aim to keep crews continuously employed but only re-measure plots once every five years. If no prior surveys exist. and panel sampling. the number of plots required to meet an increasing level of precision increases by four to reduce the uncertainty by half. For example.6. two-stage sampling.2 Coefficient of +/-20 Acceptable +/. However. a biometrician or statistician should be consulted to ensure that sampling intensity calculations and data analysis procedures are correct. most plots will have the required number of trees. travel costs can have more effect on total cost than the number of plots. when sampling large areas. and choose a plot size that is large enough that with the clumped spacing of trees in the forest. a pilot study should be undertaken to estimate the CV. On the other hand. the CV can be well over 100 percent. 20 percent of plots would be measured each year. When calculating the number of plots needed. but are not dependent on the spatial extent of the project. ranked set sampling. There are other more complex sampling systems that may or may not give more power for a given level of effort. The CV can be estimated from prior surveys that use a similar plot design in similar forests. This is an example of a panel sample. These include stratified random cluster sampling.10 Acceptable Acceptable Variation Error Acceptable Error Error Error 100% 98 392 1568 9801 50% 25 98 392 2450 20% 4 16 63 392 15% 2 9 35 221 Table 4. giving less plot-to-plot variability than smaller plots. There is a theoretical optimal balance between plot size and number of plots that can be achieved through some combination of field experiments or prior knowledge. Plot numbers in stratified sampling are dependent on the variability of the carbon stock in each stratum and the level of precision required. a level of variability might assume that almost all plots contain at least four large trees and that very few plots will contain gaps with few or no medium or large trees. The significant level is 95 percent for a large area REDD+ MRV MANUAL: CHAPTER 4.2 shows the final results of a hypothetical example of estimating sampling sizes needed to reach specified sampling errors. when choosing plot size. and for a given amount of money.The key input to estimating the number of plots needed to obtain a given level of precision is the variation between plots. Table 4. If any of these more complex sampling systems are considered.0 – FIELD-BASED INVENTORIES 65 . Technically.5 on calculating uncertainties. At some point.5 +/. Thus.2: Example of the number of sample plots needed to achieve specified sampling errors with simple random sampling.

as discussed in Section 7. o Entering data (including translating local species names into scientific names). and o Transmitting data to the central national office. and REDD+ MRV MANUAL: CHAPTER 4. should be one goal for national inventories. • Field crews that are responsible for data collection. o Setting up protocols for data collection.4 THE FOREST CARBON INVENTORY TEAM A national forest inventory team should be comprised of: • An entity with overall responsibility for the entire inventory and the ability to make decisions that are binding to regions (if regions are used). The inclusion of members of the local community is crucial for the following reasons: • Allows access to the plots. including the training and incorporation of local community members into the inventory. • Provides information on the local names of species measured. If the inventory covers a very large area. o Coordinating with the land-cover mapping team. A key issue is how field crews will be staffed. The entity responsible for the entire inventory may be governmental or may be part of a university or some other non-governmental organization with appropriate expertise and ability to continue operation. o Performing quality checks on data collection performed by field teams. o Data processing and analysis. o Organizing the procurement of data collecting equipment. field crews should be a combination of technicians with measurement skills accompanied by local community members. This entity must also coordinate closely with the single national entity designated with the overall responsibility for the GHG inventory.0 – FIELD-BASED INVENTORIES 66 . which includes: o Selecting the sampling and plot design. and o Coordinating with regions and users of inventory results. The entity is responsible for developing the inventory. Ideally. • Regional offices that are responsible for: o Organizing and training field teams. it may be efficient to have different staff in different regions. o Providing backstopping support to field teams.4. Community-based monitoring.2. A well-established national inventory where measurements are repeated regularly should have its own staff.

especially if data collection is not performed frequently.0 – FIELD-BASED INVENTORIES 67 . • Quality objectives for each measurement. with well-trained field crews. and mis-implementation of boundary correction methods. Instead. where a supervisor or another person spot checks data by re-taking measurements (while the team is still on the plot). REDD+ MRV MANUAL: CHAPTER 4. but also species misidentification. in part because it cannot be readily calculated like sampling error. food. misinterpretation of border trees and strata boundaries. After training. Supervision should include visiting crews as they are doing plot work. ideally with rewards for good quality work and additional training if deficiencies are found. as the data is being measured and recorded. and check cruising. There must be a process for giving crews feedback on the quality of their work. including not only mis- measurement of DBH and height (especially the latter in closed canopy broadleaf forests). Careful adherence to meticulous field measurement protocols serves to reduce measurement error. Field crews will need training to apply the selected protocols of the inventory. It is highly desirable to have local community members included in the field crews. the training of local community members to collect forest inventory measurements may not be cost effective. measurement error of around 4 percent is achievable. Quality assurance procedures should include both immediate checks.1 Field work Preparation for field work requires more than writing a field protocol and choosing plot locations. lodging. species identification ability. and local knowledge. This might mean having technicians with measurement skills who travel around large areas and temporary crew members who know local terrain and assist in measurements. • Field Manual specifying how field work is done. and an independent person compares the two sets of measurements to make sure the measurements are within the required accuracy and precision limits. • Provides information on the uses of important species found in the plot. an often over-looked source of error in forest inventories. Key components of successful field work are: • Logistics planning and implementation to ensure that training. misinterpretation of live versus dead. supplies. an experienced supervisor should keep in close contact with crews during their first month of work. 4. One solution is to have teams composed of individuals who. and how to address unusual cases.5. capture the necessary measurement skills. On the other hand. and checking the accuracy of their measurements (quality assurance and quality control is addressed below). or procedures where team members check each other. and • Field data check procedures. measurement error is typically assessed through remeasurement of a sample of plots via the spot checks described above. such as new logging or clearing. and communications are all provided as needed. measurement error may reach 8 percent. where a different team independently re-visits and re-measures a subset of plots. which may limit acceptable values of data entries if electronic data recorders are used. particularly in closed canopy forest. At the plot level. Community monitoring may be more practical for detecting and specifying locations of infrequent events.5 FIELD WORK AND ANALYSIS 4. equipment. because they tend to know access routes and other locally unique information. transport. together. Measurement error results from a host of issues. Where field measurements include tree heights.

Electronic data recorders also require substantial skill to set up. For example. taking into account the published accuracy of the receiver type. however. Particular attention should be given to errors that would have a large effect on final carbon stock or stock change estimates. and may be more durable than electronic measuring instruments. this depends more on the amount of time it takes to get from one plot to the next than the amount of time spent at each plot. Ecuador. the greatest challenge in estimating costs is the number of plots per day that a field crew can measure. there must be a method for relocating plots that does not only rely on consumer-level GPS readings. Many inventories use monuments to mark plot centers. Many projects record field data on paper data sheets. It is recommended that countries carefully review multiple Manuals when developing their own field Manual. data recorders can save considerable costs in terms of reducing the need to print data sheets and copy data from paper sheets to electronic form. tape measures. using laser hypsometers is much faster than tape measures and clinometers for measuring tree heights. Often. and field test procedures before adopting them. However. easy for field technicians to use. The monument must be something that is unlikely to be removed over time. REDD+ MRV MANUAL: CHAPTER 4. Maximum allowable errors should be developed in consultation with experts in both field work and data analysis. Paper data sheets are both low cost and familiar. and re-find the rebar with a metal detector. best practices when collecting Global Positioning System (GPS) data should be used. based on a desktop GIS analysis.0 – FIELD-BASED INVENTORIES 68 . recording the location of. Typically. it may not be adaptable to record the data that an inventory needs to record. as long as the tags are not commonly removed by people or animals. Russia and others have detailed field Manuals that provide useful examples when designing inventories. and tagging. Locating plots To ensure the integrity of the sample design and avoid bias. individual trees within a plot helps check for trees missing from the measurement. Electronic forms can be designed to prompt users to fill in missing values and can question or reject implausible values. Tree species can also be specified using a menu. Electronic data recorders can be hard to keep charged through multiday periods in the field. plots must be located prior to fieldwork. Canada. avoiding considerable time spent sorting out spelling errors in species names. and relocating plot centers. Further. Aligning tree tags or painted markings on trees toward the plot center can assist in relocating plot centers. Manual instruments such as diameter tapes. The United States. and as long as the markings do not cause the trees in the plots to be treated differently from the trees outside the plots.All inventories should have written standards specifying the maximum inaccuracy allowed for each piece of data. And since errors can occur with GPS locations. Recording the distance and direction from the plot center to each tree is very useful in later relocating plot centers and facilitating check cruises. and clinometers may be easier for field technicians to learn to use. Over time. Laser rangefinders may be needed to estimate the heights of tall trees in dense forests. For example. Mexico. and do not fail due to dead batteries or mechanical problems. and data should be removed from field recorders daily which can be difficult if teams go into the forest for a week or two at a time. check cruising the accuracy of measurements. Including specific location details or complementary types of data can assist in checking and correcting errors and other problems. There are a variety of textbooks and Manuals available that describe how to perform field work. many inventories drive a section of steel rebar completely into the ground at the plot center. While commercial timber cruising software is readily available.

laboratory analysis of woody biomass samples is not needed. Carbon content is determined by laboratory analysis. a statistician should be consulted to give proper methods of calculating the total uncertainty for the land-cover REDD+ MRV MANUAL: CHAPTER 4. and tracking any changes to data. 4.6 CALCULATING CARBON STOCKS FROM FIELD DATA 4. Soil carbon does require laboratory analysis. if separate reporting of each pool is not required. there is the option of calculating the carbon stock of each carbon pool on a per hectare basis.4. Key components of soil carbon quantification are: • Soil depth to be measured (in cm. If only some pools are measured. For UNFCCC reporting. usually 30 cm). This approach involves oxidizing a small sample at very high temperatures and using infrared gas absorption or gas chromatography to measure the amount of CO2 emitted. However. Bulk density can be measured on samples from which a subsample is later removed for carbon measurement. limiting who can make changes to data. and • Organic carbon content (percent). However. and why the particular values are used in particular situations.2 Laboratory analysis of samples Generally. The depth of sampling is specified in the inventory design. and then summing the pools to get the per hectare carbon stock represented by each plot. The most common techniques for analyzing the carbon proportion of soil are based on measuring emissions from dry combustion of the samples.0 – FIELD-BASED INVENTORIES 69 . it may be necessary to separately calculate the stock (or stock change) of each reported carbon pool. to ensure that all needed data will be collected. If carbon stocks (or stock changes) are calculated separately for different pools occurring at a particular site.6. version tracking. Calculation procedures should be tested on pilot data prior to committing to a particular inventory design.5. Factors used in calculations should be well documented as to their values. (2007) provide a discussion of wood density determination. • Soil bulk density (in g/cm3). Combining all pools is statistically appropriate and tends to give somewhat lower plot-to-plot variability than separately calculating the stock of each carbon pool. and default Tier 1 factors are used for other pools. Smith et al. or from a separate sample taken at the same location as sampling for carbon. Having all plots on a per-hectare basis allows calculation of statistical confidence of measurements based on the variability across plots and the numbers of plots. Exceptions are identification of unknown tree species and determination of wood densities. sources. there is often interest in knowing the change in stocks of a particular pool – especially the live tree pool – and it is often desirable to separately calculate stocks for different pools or groups of pools. the non-measured pools should not be combined with the measured pools before the calculation of uncertainty. Bulk density is calculated for each sample from the measured mass and measured volume of samples.1 Data management for calculations Calculating carbon stocks from field data must be done in an organized manner or errors will occur. Procedures should include specifying the sequence of calculations.

herbaceous and litter pools. Problems should be checked against plot sheets or earlier forms of the data. and the same plots are not measured at the two different time periods. unless the biomass estimates for these larger trees are compared to measured biomass of other large trees and REDD+ MRV MANUAL: CHAPTER 4. biomass equations should not be used for trees larger than the largest tree used to develop the equation in question. for example.2 Allometric equations Selection of allometric equations Carbon equations usually take two forms: allometric equations or biomass conversion and expansion factors (BCEFs). Without robust records. If temporary plots are used. Methods for calculating uncertainty in simple situations are described below. the sequence of calculations. As a result. drop the data from the data set. For example. factors and equations used. The carbon stock for a particular stratum is obtained by calculating the average carbon stock per hectare of all the plots within a stratum and multiplying by the area of the stratum to get the stratum carbon stock. 4. which would be the case if each observation of each pool counted as a different sample. this approach usually gives greater statistical confidence (as long as plots with significant disturbance – such as logging or fire – are not mixed with plots without disturbance) and statistical uncertainty is calculated from the set of changes observed on the different plots. and this information is key in the subsequent verification phase of MRV. or weighing of harvested trees and relating one or more structural variables — typically DBH and tree height — to a variable of interest. the difference of the means must be used. It is important to record the details of data manipulations performed. and there will be no way to determine the size of the errors. including the sources of those factors and equations. pan-tropical equations are applied to derive estimates for smaller geographic areas. The latter is a particularly important issue where. from this. and ideally should be drawn from a comparable geographic scale. Alternatively. Before calculations commence. recording the reason why the data was dropped. should be representative of the population of trees to which it will be applied. shrub. otherwise scale dependent variability in allometric relationships will be lost. and the reason for each calculation. the change in stock is often calculated as the difference of means between the two time periods. Data should be examined for missing and implausible values. coarse woody debris. such as tree volume or biomass (Diaz and Delaney. The sample dataset from which the allometric equation is derived. For a given number of plots. 2011). the change can be calculated for each plot. including: corrections of errors in the data. Data should not be removed from the analysis only because values are outliers.type. Allometric model selection is often one of the largest sources of error in forest inventories. which can usually be identified from source documents. and corrected where possible. all data should be compiled into a single file for each carbon pool. Allometric equations are regressions derived from detailed measurements of volume of trees. not n = 600.6. If the carbon stock is calculated separately for each inventory date. Total carbon stock is then calculated by summing the stocks of the different strata. dead tree. if 100 plots are measured and there are live tree.0 – FIELD-BASED INVENTORIES 70 . Significant errors are also likely to occur if equations are applied to trees larger than the range from which the equation was developed. The sample size is n = 100. the different pools do not count as different samples when calculating uncertainty. deletions of uncorrectable data. it is impossible to check the quality and accuracy of calculations and resulting carbon stock estimates. population level estimates of the total amount of change can be calculated. and. if permanent plots are measured. If correction is not possible. Calculation of carbon stocks from field data requires good organization to ensure that the data are efficiently sorted and the resulting calculations are correctly generated.

the wood-density adjustment factor is calculated by dividing the specific gravity of the species to which the equation will be applied by the specific gravity of the species used to develop the equation. In this case. equations that use diameter only should be considered useful for rough estimates of biomass. BCEFs are dimensionless factors that convert the merchantable volume of trees into their total aboveground biomass.. REDD+ MRV MANUAL: CHAPTER 4. but a larger sample is desirable. Alternatively. Most allometric equations give unrealistic results when applied to trees larger than the trees from which the equation was developed. Unless the allometric equation was developed using measurements of trees in the area where the equation will be used. and from stands with similar trajectories of development as the stands to which the equation will be applied. or equations that use height. If an equation will be applied to a wide variety of species.2 of the IPCC Guidelines (2006). When using BCEFs. When considering taxa-specific 20 The most useful groupings may be by morphology class (e. In particular. REDD+ projects may have to develop new allometric equations. Equations that use height and diameter but not wood density can be adapted to estimate the biomass of species different from the species from which the equation was developed. BCEFs are used for rough estimates of biomass when a timber inventory is available but resources are not available to measure carbon stocks in forests. where the rate of increase in predicted biomass declines as the diameter increases. Dietz and Kuyah (2011) prepared guidelines for establishing regional allometric equations through destructive sampling21. New allometric equations can be developed with a relatively small sampling of approximately 30 trees for a particular species or group of species. and if estimates are adjusted for differences in wood densities. or wood density should be incorporated into the biomass estimation. 21 More guidance on developing biomass tables is found in MacDicken (1997) Annex 4. and 19 A database developed by FAO. such as local forestry institutions. Under the Global Environmental Facility’s (GEF) Carbon Benefit Project. it is preferable to use allometric equations that are developed from trees similar to those being studied. 1997). General equations are provided in Annex 4A. Wood density may need measuring. GlobAllomeTree19 and published literature. Various sources can be useful when seeking allometric equations. Therefore. Section C. and adjusted for the difference in wood density. multiple-stemmed trees. or group of species 20. the French Research Centre CIRAD and Tuscia University.globallometree.0 – FIELD-BASED INVENTORIES 71 . shrubs) (MacDicken. BCEFs are unreliable when applied to forests of different structure from the forest where the BCEF was developed. Logistic equations. single-stemmed trees.g. if the growth forms of the species are all similar. This unreliability is particularly great with equations that are simple exponential models. tend to have less error when applied to trees larger than the trees from which the equation was developed. the species (or combination of species) or growth form and potential biomass should be similar. one should consider both the availability of analysts who can identify the species and the additional cost and uncertainty of tree-height measurements.the estimates are documented to be reasonable. equations for a similar species may be used. it is critical to apply the same definition of merchantable volume (or growing stock) that was used in the derivation of the BCEF. http://www. either the species should be grouped so that each group has a similar wood density. Developing new and testing existing equations If no information is found regarding a certain species. Another guide for developing allometric equations is Aldred and Alemdag (1988).

because roots of different trees and shrubs intertwine. or growing conditions from where the equations were developed.g.because forest measurements are made on live green trees. because these measurements may give significantly greater biomass than default ratios for sites of any productivity. where more energy is allocated to root growth. dn = diameter (m) of each of the n pieces intersecting the line. Volume is converted to mass using the appropriate density factor. undertaking supplemental harvest (e. stem wood. the ratio of belowground to aboveground biomass is higher for dry and nutrient-deficient sites. Belowground biomass can be measured by digging and weighing root balls. The field measurements are transformed to dry weight and scaled to per hectare basis. Care should be taken to achieve a representative sample of tree species across a range of diameter classes (> 15cm dbh. REDD+ MRV MANUAL: CHAPTER 4. or dead herbaceous vegetation). In general. destructive sampling may be worthwhile to measure the biomass of roots in the project area. As a result. Belowground biomass Belowground biomass is extremely difficult to measure for an individual tree. Harmon and Sexton. significantly reducing cost and effort required for this labor intensive activity.. and calculating a weighted average carbon fraction. foliage versus branch wood. belowground biomass is often estimated using general equations that estimate it as a function of aboveground biomass. and L = the length of the line (100 m.0 – FIELD-BASED INVENTORIES 72 . Checking the fit of published allometric equations is good practice when equations are applied to sites with different productivities. d2. and for young successionals forests. 4. Volume is measured by dividing the tree trunk into segments and measuring the two end diameters and the length of each segment. non-commercial species). as necessary. Destructive sampling involves cutting down and weighing a sample of trees. Dry wood volume cannot be used. Processing of litter samples and calculations of litter biomass are similar to the methods used for herbaceous vegetation. or by examining samples to see what plant parts compose the litter (e..g. Samples are dried and weighed and the dry to field weight ratio is calculated. The carbon proportion of dry biomass weight can be estimated either by laboratory analysis. and cutting a subsample of tree parts. (1999). climate conditions.6.3 Non-tree pools Scaling up from samples to a per-hectare mass is straight-forward. below which there are rarely substantial inaccuracies in equations). weighing them in the field and drying them to develop a dry-to-field weight ratio for the weights of the whole trees. The volume of coarse woody debris per hectare is calculated for each density class for each stratum: Volume of coarse wood debris (m3/m2) = π2 * [(d12 + d22 + … + dn2)/8L] where d1. Results are plotted against equation predictions to assess bias and either validate the equation or re-parameterize the equation to produce a better fit. and also taking measurements on a sample of branches. finding the carbon contents of each component in the literature. Limited destructive sampling for this purpose could be conducted in coordination with an active harvest operation in the area of interest. Methods are described in Bledsoe et al. 1996). This can be checked by destructive sampling or measuring the volumes of a few trees of different sizes. For large projects on dry sites. green wood volumes must be used to calculate wood density. and coring a sample of locations between the stems. to achieve this.

a 95 percent confidence interval of +/. and independent checking of measurements. it is better to assign a date that the measurements represent. Technically. The technical name for this chance difference is sampling error. If inventories are taken over multiple years. measurements taken during a dry season of November 2011 through February 2012 could be taken to represent the carbon stock present as of February 1 of 2012. The standard deviation is a measurement of how different individual samples are from each other. A common index of uncertainty associated with an estimate from an inventory is the confidence interval. for example. most assume that these non-sampling errors are random and not biased. In reality. Nonetheless. Some inventories only specify the year that measurements represent. and calculations to detect and fix human errors. or models should be used to normalize the data to a single year. measurements are taken to represent a particular date. typically the mean—that is. most likely.4 Combining carbon pools The per-hectare carbon stock of each pool in each plot is summed with the other pools in that plot to give the per hectare carbon stock for each plot.6. made by different people. For example. There are several mechanisms that can be used to limit errors other than sampling errors. The confidence interval represents a range of values surrounding an estimate. Confidence intervals can be calculated for different “confidence levels”. The standard error is a property of the sample and can be reduced by REDD+ MRV MANUAL: CHAPTER 4. The standard error of the estimate is a measurement of uncertainty of the estimate of the mean value. measurements are taken over a period of weeks to months. The standard deviation is a property of the population. between 90 and 110 tons per hectare. the population value would be the carbon stock measured if every tree was measured. The true population value is the value that would be found if every individual in the population was measured. People often interpret that to mean that one can be 95 percent confident that the true value lies within the confidence interval.4. one can say that if a similar inventory was conducted many times in the exact same way but choosing a different set of plots. Consult an appropriate textbook for guidance on how to use these methods. For example. To interpret. these uncertainties are reporting the chance that the sample is different from the actual total population. because some carbon stocks vary between seasons. In this example. 4. However. might differ by a few millimeters. data. These include quality standards. for the sake of reporting and change over time. There are many other kinds of errors that could lead to false numbers. The width of the confidence interval conveys to the data consumer a sense of confidence in the accuracy of the estimate. The total carbon stock is calculated by multiplying the average value per hectare times the number of hectares. then approximately 2/3 of the plots will have carbon stocks within 50 tons per hectare of the average carbon stock. in this example. 95 percent of the confidence intervals generated would contain the true population value. All these potential errors mean that two independent measurements of the same tree. and thus that they increase the confidence interval and do not bias the stock estimates. Typical confidence levels are 90 percent and 95 percent. To calculate a confidence interval. either a panel design should be used to calculate the average and changes.5 Quantifying uncertainty The reliability of carbon stock estimates is reported in the form of statistical confidence intervals that quantify the chance that the sample plots used to calculate carbon stocks might be different from the actual conditions that exist throughout the entire forest.0 – FIELD-BASED INVENTORIES 73 .6. first the standard deviation and standard error of the estimate must be calculated. However. and are based on statistical theory.10 percent surrounding an estimate of 100 tons per hectare of carbon. These methods are beyond the scope of this Manual. if the standard deviation of a set of plots is 50 tons per hectare of carbon.

The methods differ depending on the degree of difference of type between the pools or strata. depending on whether temporary or permanent plots are used. Methods differ for calculating the difference of means confidence. The confidence interval can be expressed as a percentage of the mean: Un = (CI/ � X) where: Un is the uncertainty in percent for pool and stratum n. should be combined to estimate the biomass carbon stock. a 95 percent confidence would leave 0. S is the standard deviation of the estimated mean carbon stock per hectare of the particular carbon pool and stratum.025 of the probability in each tail of the distribution. the uncertainty is reduced relative to the same number of plots in a simple random sample. Note that typical uncertainties in forest inventories are generally weighted by the number of sample units observed in each stratum or by area.measuring a larger sample. measuring more plots. To calculate the uncertainty of a stratified inventory. on-site biomass and carbon stored in wood products are separate pools. t is the tcritical point from a table of student t test values. rather than by the number of tons in each stratum. Also. consult a forest measurement textbook such as Avery and Burkhart (1994) or a statistics textbook. The details of calculating the uncertainty of a stratified inventory are beyond the scope of this Manual. the covariance term is subtracted in the estimation of standard error. and the degrees of freedom is typically the number of plots minus one. and the degree of congruence of sampling methods used in the different pools or strata. Permanent plots achieve greater precision in the estimation of change than temporary plots – for permanent plots. Separate classes of biomass. The confidence interval is: CI = ±t × SE where CI is the confidence interval. i. If calculating the change in carbon stock from one simple random sample to another measured at a later date. For example. The confidence interval for each carbon pool within each stratum is then calculated. in tons per stratum. This is for a two-tailed test. such as live trees and dead trees. There are multiple acceptable methods for combining uncertainties across multiple pools or strata. and are presented in many statistics textbooks. i. The standard error of the estimated mean for each carbon pool within each stratum is: SE = √(S/n) where SE is the standard error of the estimated mean carbon stock per hectare for the particular carbon pool and stratum. for the appropriate confidence level and degrees of freedom. REDD+ MRV MANUAL: CHAPTER 4. consult a statistics textbook for guidance.. If the inventory is stratified. and n is the number of plots in the stratum. SE is the standard error for the particular stratum and pool. Pools should be independent and pools should technically be spatially separated.e.. CI is the confidence interval for that pool and � is the average estimated carbon stock of that pool and stratum. if doing paired sampling.0 – FIELD-BASED INVENTORIES 74 .e. and X hectare. in tons per hectare. the uncertainty is calculated for each stratum and the uncertainties are weighted and combined. For guidance. the change is calculated as the difference of means.

it is important to develop a plan for how data and metadata will be stored and protected from unauthorized changes or loss. and necessary. Methods for measurement. and ensure that later measurements and calculations are comparable to earlier data. REDD+ MRV MANUAL: CHAPTER 4. At a minimum. what is included. Data should be archived in at least two locations. if no reliable correction of a data error can be achieved.7. data cleaning. Inventory design. and Tier 3 measurements of live trees) the combined uncertainty of the estimated change can be calculated using Equation 5. This is also important for satisfying the IPCC principles of TACCC.6. and En is the emission or removal for the stratum. However. in tons. particularly if different pools are measured with different UNFCCC methodology tiers (such as Tier 1 factors for shrubs and forest floor.g. for the pool n. Regardless of the strength of the foundation. as new equipment and software are adopted. Having multiple teams involved in data analysis maintains awareness of the data and appropriate uses of these data.7 DATA CHECKING 4. data must be thoroughly checked before carbon stocks are calculated. step in an inventory as it improves the efficiency of future inventories. data from an inventory must be stored so it can be retrieved at a later date. professional data managers will be consulted in the design of data storage forms and use of data storage equipment.6 Propagation of error If the uncertainty for a combined estimate of sinks or emissions from multiple. limited time is often invested in these activities. in tons. If field data have substantial errors. Correctly and consistently archiving data and associated metadata represents an important.2. checking and accuracy standards Data quality is essential. 4. Land-Use Change and Forestry (GPG-LULUCF) (2003): (U1 + E1 )2 + (U2 + E2 )2 + ⋯ + (Un + En )2 UE = � |E1 + E2 + ⋯ +En | where UE is the combined uncertainty in percent for the sum of changes in all pools 1 to n. as well as the underlying sampling methodologies applied (e. the entire inventory could be worthless. Metadata describe how data are collected and what they represent. and any adjustments or calculations must be clearly specified to build confidence in later users. it increases the likelihood of institutional memory transfer. and information about where data is stored. Relatively frequent use of the data ensures that data will be transferred to new storage and retrieval media formats and processing systems. sample design. Further. 4. This includes checking for missing data and implausible data values.7.2 Archiving data and metadata To recalculate changes in carbon stocks over time. and who controls access should be readily available. the faulty plot should be excluded from carbon stock calculations.0 – FIELD-BASED INVENTORIES 75 .2 of the IPCC Good Practice Guidance for Land Use. Key aspects of forest inventory metadata are the protocols used to direct field measurements.1 Data cleaning. This is especially important given the long time periods involved in REDD+ MRV. field technician training and management of field crews are the foundations of data quality. delineation of sample population area). Un is the uncertainty in percent for pool n. independent pools is being calculated. As discussed above. Ideally.4.

g.4. A second critical consideration is that the candidate data must be screened to ensure a minimum acceptable level of data consistency and quality.0 – FIELD-BASED INVENTORIES 76 . New needs and questions arise. This is just as important as the plot data itself. and are a way of evaluating changes over time without having to wait years or decades for a new set of measurements. or at least estimates of wood volume in live trees. Foremost among these considerations is that the population that a given inventory represents be clearly delineated. they may contribute to the study of additional dynamics as well. or to start estimating changes in carbon stocks. from commercial forest inventories. 4. • Calculations of stock changes over time. that the area within which samples were drawn (with a known probability) is mapped and documented. it may be necessary in the early stages of national REDD+ accounting. • All data must have been collected with minimum quality assurance and quality control (QA/QC) procedures. and • Reporting uncertainties across pools and strata. there are a number of critical considerations to ensure that data are consolidated to avoid or minimize bias in the derivation of an overall stock estimate. prior to the implementation of a national forest inventory..3 Data analysis and reports Typical reports from forest carbon inventories include: • Calculations of biomass and carbon stocks. • Timber inventories. to develop estimates from existing datasets (e.. and a well-documented historic data set can provide a window into past conditions or changes. as repeated measurements form an archive. often with reports by pool and stratum. Important assessments are: • Raw data (plot measurements with individual tree data) should be preferentially sourced. REDD+ projects. REDD+ MRV MANUAL: CHAPTER 4. to ensure consistency in estimation procedures (e. forest inventories become irreplaceable windows to the past. Over time. and experiences show that a well-maintained inventory will likely have many valuable uses. The initial use of an inventory might be only to measure timber volume or carbon stock.7.8 CONSOLIDATING INVENTORY DATASETS Given resource constraints. environmental research). Depending on parameters measured.e. i. allometric equations used) across inventories. the value of these data will increase. While it is not necessary to collect different datasets using the same sampling approaches. One cannot foresee what issues will become important in the future. and permit aligning inventories to common minimum diameter thresholds. otherwise it is not clear what population (forest area) that data represents.g. However. ensuring minimal measurement error and sample bias (see discussion of QA/QC in this same chapter).

accepting a certain level of unknown bias.0 – FIELD-BASED INVENTORIES 77 . a shorter allowable timeframe is recommended. • All data must cover the same pools and use the same pool definitions. with pieces represented by sampled areas. a 2014 estimate might be derived from data collected up to 5 or 10 years prior to the reporting date).g. estimated from stand tables per Gillespie et al. (1992). • Minimum sample size of 10 to ensure stable.g.000 150 10 B 5. REDD+ MRV MANUAL: CHAPTER 4. the individual inventory estimates are weighted by their respective (sampled) population areas. satellite imagery). spatially-explicit geo-physical data. if they are an insignificant area).g. that can either be: • Delineated and excluded from MRV. with each sampled area treated as an individual stratum (Table 4. • Ignored (e. The result is an area-weighted average. each with their own estimate. Inventory Sample population Estimate (mean t Standard error area (ha) CO2/ha aboveground (mean tCO2/ha live biomass) aboveground live biomass) A 10.. thereby focusing MRV on managed lands. In this way.. the entire population of interest (e.000 139 81 Table 4. but strictly commercial inventories cannot be re-interpreted as all species total biomass).g.. McRoberts et al. 2011). for example.000 200 20 X (un-sampled) 10.3). To derive a total estimate.. or • Assigned estimates based on predictions derived from relationships of known estimates with ancillary data (e. 2007). un-biased estimates of mean and variance (Westfall et al. There will inevitably be gaps remaining that have not been sampled. including diameter thresholds for tree measurements (in some cases it may be possible to reconstruct diameter distributions 22. in the case of remote..000 125 50 C 1.000 130 (prediction) 60 (model prediction error) TOTAL 26.g.. • All data must be collected from within a maximum time range (e. inaccessible areas where anthropogenic changes are unlikely to occur. • Delineated and sampled with a new targeted field effort to fill the gaps.3: Example of an area-weighted average with each sampled area treated as an individual stratum 22 E. e. e.g.. using regression or nearest- neighbor approaches (see. national forested area) is constructed as a sort of jigsaw puzzle. Where substantial areas of forest are in an early successional stage.g.

requires targeted field-based sampling of areas subject to extraction activities. it acknowledges the value of data collected via project-level MRV. national/jurisdictional MRV. Remote sensing-based approaches have not yet achieved sufficient resolution of stock change to be viable for direct estimation of emissions from degradation. Where an un-sampled area is assigned an estimate based on a prediction. Importantly. Some recent advances have applied this approach to estimation of logging impacts by Griscom et al. and • Emissions from wood removals due to logging and fuel wood collection. reliable information on wood removals from which harvest-related emissions can be estimated. 2005) and could be used.e. the IPCC recognizes gain-loss as an alternate method to estimate carbon changes (IPCC. though they have achieved success in identifying areas subject to degradation (Souza et al. diseases). Probability-based sampling approaches.g.. to target sampling and scale emission estimates. • Emissions from natural mortality (competition-related and senescence) and disturbance (mortality due to wind. (2014). (2014) and Pearson et al. this approach of building a national inventory from smaller scale datasets (e.. Many challenges remain. from anthropogenic wood removals.. inventories of REDD projects) also facilitates participation from multiple stakeholders. similar to a stratified sample. A more efficient approach to estimate small changes in comparison to the stock size is to focus directly on measurement of fluxes to and from the stock (i. which understandably are difficult to locate and access (for targeted field sampling) or for which reliable and complete estimates of wood removals are unavailable. and projects would become contributors to. or alternatively.g. and participants in. referencing against a validation dataset). root mean squared error of predicted – observed.2). in tandem with field-based approaches. Typical fluxes (i.0 – FIELD-BASED INVENTORIES 78 . Emerging technologies in remote sensing are discussed in Chapter 5.. the error of the prediction should be included (e. particularly where the objective is to resolve those changes with even modest levels of statistical significance. 2006).e. is challenging. sinks and sources) tracked by a gain-loss type approach would include: • Sequestration from forest growth. By utilizing information from multiple scale datasets.9 THE GAIN-LOSS METHOD Estimating comparatively “rare” and small changes in biomass distributed over a large landscape. fire. REDD+ MRV MANUAL: CHAPTER 4. the gain-loss method). as well as any error around the estimate of the independent variable(s) (Table 4. would typically require prohibitively high sample intensities using the stock change method. The latter source of emissions.Uncertainty in the overall estimate is derived by propagating inventory level errors. 4. As discussed in Chapter 3. particularly with regard to subsistence and illegal logging. in addition to the stock change method. even with permanent plots (which permit greater resolution of change. like those produced from selective cut timber extraction activities. and are particularly useful for monitoring forest growth).

Estimating areal means and variances of forest attributes using the k-nearest neighbors technique and satellite imagery. E.E. Tomppo. Sampling Techniques for Forest Resource Inventory. 4th edition. 134 pp Cochran. T. Statistical techniques for sampling and monitoring natural resources. Aldred.pdf. Kim Iles and L. John Wiley & Sons. H. 1977. Rep. Forest Science 50: 427-435 Grosenbaugh.0 – FIELD-BASED INVENTORIES 79 . Ducey. World Agroforestry Center (ICRAF). Delaney. Comparison of three sampling methods in estimating stand parameters for a tropical forest.T. 2011. 3rd ed. Alemdag. H.S. A.. eds. Forest Service. and B. Sampling Methods for Multiresource Forest Inventory. John Wiley & Sons. Forest Inventory. 1993. Kohl. Tropical forestry handbook. NY. Canadian Forestry Service. Dietz. McRoberts R. Gregoire and G.G.. Inc. 1993. Valentine. 111 p. 2011. NY. Springer-Verlag. USA. Borders. 2007. Tech. Guidelines for Forest Biomass Inventory. L. Beers and J. 2002. Johannes. Carbon Stock Assessment Guidance: Inventory and Monitoring Procedures. Gove. 356 pp. and H. Forest Trends. Point-sampling compared with plot-sampling in southeast Texas. ed. Inc. Finley and J. Forest mensuration.4. CO. 2002. Inc. REDD+ MRV MANUAL: CHAPTER 4.. Global Environment Facility.E. USA. B. Nairobi. Olander. Iles. Kuyah. 1957. RMRS-GTR-126. Husch. Schreuder. E. and H. Brink. A Sampler of Inventory Topics. Ltd. Wood. J. Inc. Available at: http://www. 1996. 2004. Diaz. T.. In: Building Forest Carbon Projects. Forest Science 3: 2-14. S.T.H. 2003.msu.10 REFERENCES Forest inventory. Berlin. and M. U. Ramirez-Maldonado. Ernst and H. Kenya. 2004. Burkhart. A. and W. New York. Petawawa National Forestry Institute. Department of Agriculture. New York. DC. and S. Fifth edition. K. and I. John Wiley and Sons. R.O. sampling design and statistics Avery.J.. Forest Ecology and Management 21:119-127. Stover.. Gen. Chapter 5 in Pancel. T. J. Rocky Mountain Research Station. Guidelines for establishing regional allometric equations for biomass estimation through destructive sampling. A Walkthrough Solution to the Boundary Overlap Problem. Sampling techniques. E.S.. W. and J. 464 pp. Remote Sensing Environment 111(4): 466–480 Schreuder.. 1987. Heikkinen. Germany. Washington. Shiver.. M. pp. Kershaw. H. Fort Collins. M. McGraw-Hill. B.H. Carbon Benefits Project: Modelling.O.goes. Information Report PI-X-77. USA. John Wiley & Sons. Schreuder. 456 pp. D. Measurement and Monitoring. Banyard and G. 243-332. 1988. Forest Measurements.

The Netherlands). USA. Integrating remote-sensing and ground-based observations for estimation of emissions and removals of greenhouse gases in forests: Methods and Guidance from the Global Forest Observations Initiative: Pub: Group on Earth Observations. Warren. Intermountain Forest and Range Experiment Station. Rome. Ogle. Forest Resources Assessment Programme Working Paper 94/E. Switzerland. The line intersect method in forest fuel sampling. 1964. Remote Sensing Environment 98: 329-343 Westfall. GFOI. Tropical forest biomass estimation from truncated stand tables. Forest Ecology and Management 48:69-88. G.G. S. L. A Guide to Monitoring Carbon Storage in Forestry and Agroforestry Projects.J. R. WA. L. S. A line intersect technique for assessing logging waste. Nicholas Institute for Environmental Policy Solutions. 1992. GOFC-GOLD Report version COP19-2.E. USA. Souza. 2011. J. H. Intergovernmental Panel on Climate Change. NC.F. Carlos M. Sampling and measurement of downed dead wood Brown. Raleigh. 229p. G. Duke University Press. McCarl. Barbour. A. Japan. K. Brown and A. 1974. Hammerschlag. Harmon.0 – FIELD-BASED INVENTORIES 80 . Ogden. Holtkamp. K. University of Washington. Land-Use Change and Forestry. MacDicken. Wageningen University. Good Practice Guidance for Land Use. Forest Science 14: 20-26. A sourcebook of methods and procedures for monitoring and reporting anthropogenic greenhouse gas emissions and removals associated with deforestation. K. Cochrane. and forestation. Reynolds. REDD+ MRV MANUAL: CHAPTER 4. 2007.. Forest Science 10: 267-276. Guidelines for measurements of wood detritus in forest ecosystems. 20. M. 1968. Lugo. IPCC Guidelines for National Greenhouse Gas Inventories. Jr. 2003. US LTER Publication No. Edited by Zach Willey and Bill Chameides. General technical report INT-16. Combining spectral and spatial information to map canopy damage from selective logging and forest fires.E. Roberts and M. M. J.. Patterson and J. Harnessing farms and forests in the low-carbon economy: how to create. and W. P.A. R. Winrock. J. IPCC. Seattle.. W. National Forest Inventory Field Manual Template. J. 2004. 2013. 2005. C. 2013. gains and losses of carbon stocks in forests remaining forests. Sexton. 1997. FAO. Olsen.A. Forest Service. Hayama. UT: US Department of Agriculture. Handbook for Inventorying Downed Woody Material. Parton. Changsheng. Intergovernmental Panel on Climate Change.R. GOFC-GOLD. measure.. J. and P. W. Paustian. Synthesis documents on forest carbon inventory and REDD+ field-based sampling MRV Gillespie A.W.Smith.L. (GOFC-GOLD Land Cover Project Office. D. and J. Sass. Geneva. 1996. R.A. Coulston. Hayama. Japan. IPCC. Post-stratified estimation: within-strata and total sample size recommendations.E. 2006. Bruce. Van Wagner. and verify greenhouse gas offsets. Canadian Journal of Forest Research 41: 1130-1139. US LTER Network Office.

Land-Use Change and Forestry Projects. S.E. R. R. 57pp. 2014. 20: 923–937. Nicholas Institute for Environmental Policy Solutions.. R. Edited by Zach Willey and Bill Chameides. Ellis and F.Pearson. Brown and R.winrock. Brown. Available at: www.H. Barbour.fed. A.M. Gain-Loss approaches to estimation of logging impacts Paustian. G.nrs. Casarim. T. Pearson. J. 47 pp. 2005. 2014. 4. T.pdf. Smith. McCarl. Environmental Research Letters 9: 034017-034028..leafasia. Ogle. Carbon emissions performance of commercial logging in East Kalimantan. Global Change REDD+ MRV MANUAL: CHAPTER 4. B. NC. Parton. L. H.. Holtkamp. S. Winrock International and the World Bank Biocarbon Fund. Carbon emissions from tropical forest degradation caused by logging. Hammerschlag. Pearson.11 SELECTED RESOURCES Carbon Measurement and Monitoring (CMM) Module of the USAID-supported Lowering Emissions in Asia’s Forests (LEAF) program. and verify greenhouse gas offsets. and W.. Sass. P. S. L.pdf. Duke University Press. Raleigh. S. Walker and S. General Technical Report NRS-18. J. Sourcebook for Land Use.R. J. Bruce. United States Department of Agriculture. Available at: http://www. Available at: http://www. Birdsey.. K. 2007. T. Putz. 229 p. M. Harnessing farms and forests in the low-carbon economy: how to create. Reynolds. measure.fs. Measurement Guidelines for the Sequestration of Forest Carbon. Brown and F. Indonesia.0 – FIELD-BASED INVENTORIES 81 . W. Changsheng. . DC 20523 Tel: (202) 712-0000 Fax: (202) 216-3524 www.usaid. NW Washington. Agency for International Development 1300 Pennsylvania Avenue.