1. Big Data & Ecosystem Accounting
Emil Ivanov, UNSD
10th Meeting of United Nations Committee of Experts on Environmental-Economic Accounting (UNCEEA)
24-26 June 2015, New York
Working on advancing natural capital accounting through SEEA-EEA

2. Outline: Big Data for official statistics and ecosystem accounts
Advancing Natural Capital Accounting through SEEA-EEA:
with Generic technical guidance and training materials
applications in pilot countries
Global and continental applications

3. Big Data for official statistics and ecosystem accounts
Pilot areas to explore the use of satellite imagery and geospatial data :
Agricultural statistics
Land cover and land use statistics
Geocoding statistical frame, such as business register and postcode address file
Ecosystem accounting
Integrated statistical production process
Area of environmental statistics, such as biomass, carbon, water, vegetation, soil, etc.
Census enumeration areas, business areas, rural/urban localities
Transportation statistics, spatial statistics, regional/local statistics
Disaster risk management
Each study area may include the following :
Methodological issues, covering quality concerns and data analytics
Legal and other issues of access to data sources
Privacy issues
Security, IT issues and management of data, including advanced ways of data dissemination, assessment of cloud computing and storage, and cost-benefit analysis
Sustainability of data production process
Capacity building, training and sharing of experience
Policy applications
Evaluated data sets based on the following criteria:
Dataset, type and format
Subject/purpose of the satellite
Spatial resolution
Time series availability
Sensor and Satellite
Coverage: Global, regional or national?
Cost: free or commercial?
There is a very close relation between Big Data and ecosystem accounting, in fact little can be done in ecosystem accounting without big data.
On the other hand there is a tremendous volume of work done with big data on ecosystems, but with results of varying quality, large errors and consequently unreliable information. Here the process of pursuing standardization, as required for official statistics plays the most important role for improved quality and reliability of that information.
The task team will undertake pilot projects to explore the potential use of satellite imagery and geospatial data in relevant areas of official statistics that may include, but not limited to, the following areas (see

4. 2. Advancing SEEA-EEA: generic guidelines for pilot countries
Generic technical guidance
Strong on concepts and structure, but yet weaker on methods and data inputs
Experimentation
Physical accounts in focus first
We are developing Generic technical guidance on defining, classifying and mapping ecosystem units, assessing their extent, condition and ecosystem services. The guidance includes strong background and broadly agreeable concepts and accounting structures, including recommendations for methods and data sources, but there is not (yet) definitive selection of most appropriate methods and data. Therefore these are particularly reliant on experimentation in pilot countries as well as other applications, continental and global.

5. 2. Advancing SEEA-EEA Associated countries: Global applications
The Netherlands
Australia
Canada
Brazil
Colombia
Costa Rica
Norway
Peru
Philippines
United Kingdom
USA
Global applications
Continental applications
Pilot countries:
South Africa
Mauritius
Chile
Mexico
Indonesia
Bhutan
Vietnam
The Technical guidance draws on experiences from the Pilot countries, associated countries but also work done in the EU and global applications.
Pilot applications are under way in each pilot country, also in several of the associated countries (Australia, Canada, Norway)
Examples from the outlined will be shown, but also global and European applications
In the following slides we will see examples of how elements of the generic guidelines correspond to concrete country and other applications.

6. 2. Defining, classifying and mapping ecosystem units
Vegetation/habitat/biotope classification methods
Mapping based on field work (existing inventories) aided by high-resolution remote sensing
Previously ecosystem units were approached through land cover mainly. Introducing ecological foundations into the definition and classification of ecosystem units. By looking at ecosystem structure, function and composition, as opposed to the previous approach based on land cover only. Still the reference to land cover is retained but at a more aggregate level.
Structure here is the existence of difference vegetation layers (trees of varying age, shrubs, herbs etc), composition is the existence of different species in each of these layers, and both of these ecosystem properties define corresponding functions, like younger growing trees underpin higher rates of carbon sequestration, wetland vegetation regulates water flows etc.
The classification is aided by botanic methods such as Braun-Blanquet complemented by habitat considerations, much EU experience is published.
The mapping is particularly reliant on experimentation, no strict methods can be prescribed, but generally existing vegetation inventories, aided by additional fieldwork and high-resolution remote sensing are recommended.
Now lets see how this is put in practice … in Chile
Ecosystem Unit = EU
Land cover = LC
LC4: Water bodies
LC5: Cropland
LC1: Tree-covered area
EU5
EU10
EC7
EU13
EU1, LC1
EC2, LC1
LC7: Urban land
EC8
EC11
EU3
EU4
EU9
EU12
EU6
LC3: Wetland
LC6: Bared land
LC2: Grassland

7. 3. Mapping ecosystem units (biotopes) in Chile
Early work, Luebert & Pliscoff (2006)
Pilot area: central, Mediterranean Chile
Mapping done at two levels:.
formations (coarse 1:100000, MMU 25 ha) equivalent ERA
and biotopes (fine 1:25000, MMU 1.6 ha) equivalent to an EU
Non-vegetated areas (use of NDVI, SAVI)
Cultural vegetation (plantations, parks, crops)
Natural vegetation (object-based classification of imagery, sampling and fieldwork)
Mapping done at two levels, fully consistent with the generic guidelines.
formations (coarse 1:100000, MMU 25 ha) equivalent to ecosystem reporting area (ERA)
and biotopes (fine 1:25000, MMU 1.6 ha) equivalent to an Ecosystem Unit (EU)
Groundwork – sampling; satellite object-based analysis of vegetation lead the distribution of sampling releves
Ecosystem units with their present extent mapped in this way for a pilot area, that can be used in support of ecosystem management, conservation and restoration policies.
Mapping done using ubiquitously available environmental variables
Meso-climate (Med. Xeric, -oceanic)
Altitude (hills, pre-Andes, Andes, High Andes)
Geo-form (peaks, passes, plains, slopes shaded and illuminated)
Floristic composition and coverage (24 dominant species)
Dominant life forms (7 trees, 12 shrubs, 4 herbs, 1 succulent)
Results: 125 biotopes of natural veg. mapped ( ha)

8. 4. Mapping ecosystem units in South Africa
Slightly different but still very similar approach in South Africa. The distribution of Potential vegetation was mapped as predicted by the same environmental variables. Yet the current extent and condition is addressed at a second step.
Ecosystems can be defined at a range of spatial scales. In the terrestrial environment, vegetation types provide an excellent way of delineating ecosystems at a relatively fine scale. Vegetation types are based on a range of factors, such as geology, soil types, rainfall, temperature and altitude, which determine the composition and structure of plant communities. They provide a good indication of terrestrial biodiversity other than plant species, because many animals, birds, insects and other organisms are associated with particular vegetation types or groups of vegetation types.
Vegetation map of South Africa 2006 ~440 vegetation types
Distribution of potential vegetation predicted by environmental variables

9. 5. Mapping Ecological Vegetation Classes in Victoria (Australia)
Developed by Department of Environment and Primary Industries, Victoria
Mapped by ground work of species inventories and Landsat imagery
Similar approach applied in Victoria (Australia) for mapping Ecological Vegetation classes, equivalent to Ecosystem Units, that can also be referenced to the SEEA 14 land cover classes.
Fully operational approach, applied for management and restoration of natural vegetation in Victoria.
Statistics produced on ecosystem extent, condition and rates of major ecosystem services, such as reduction of soil erosion and water infiltration.
For reporting and international comparability the EVCs are grouped into the 14 SEEA land cover classes.

10. 6. Global mapping of Ecological land units
Bioclimate
6. Global mapping of Ecological land units
ESRI/USGS work (Roger Sayre 2015)
Landform
Lithology
Ecological units mapped globally at 250 m and higher detail for the African continent (90m)
These examples show considerable consolidation of methods producing very comparable work in terms of definition, classification and mapping (assessing extent) based on Big-Data only (at continental and global level) and complemented by field work and pilot areas.
Now a different sort of example for the European Union, but reliant even more on Big Data
Land Cover
3,923 ELUs Mapped
250 m Spatial Resolution

11. 7. Mapping ecosystem carbon accounts in EU
Carbon Accounting items Data sources
Opening Stocks
1. Soil organic carbon (SOC)
JRC map of SOC (Hiederer and Köchy, 2012), global at 1km, 30 cm and 1m depth; EEA estimate of SOC, 30cm
2. Biomass (TCB)
Downscaled forest biomass by EEA
Upscaled biomass for non-forest biomass by EEA
Fluxes and transfers
3. Primary production (GPP)
Downscaled NASA-CASA NPP (from 8km to 1km), converted to GPP by adding autotrophic respiration from MODIS (Running et al.)
4. Carbon release / respiration (TER)
Downscaled NASA-CASA Soil respiration (from 8km to 1km), converted to TER by adding autotrophic respiration from MODIS (Running et al.)
5. Human use of primary production (TPPU)
Downscaled regional statistics on crops (EUROSTAT), timber (EFISCEN, National FI and EFIMED) and grazing livestock, using land-cover and vegetation indices
6. Carbon imports (TCR)
Downscaled deposition of dry sludge and manure (from livestock distribution)
Balances
7. NEP, Ecosystem carbon balance (NECB)
NEP estimated from GPP and TER, NECB estimated by aggregating all flows
Work done for the European Environment Agency
Mapping ecosystem carbon flows and stocks. No structural unit definitions applied, but most big functions are addressed here. E.g. carbon storage in main pools (soil and biomass) and carbon flows between the ecosystems, atmosphere and anthroposphere (human activities)

12. Balancing estimates
The two basic balancing items are designed to summarize ‘vertical’ and ‘horizontal’ carbon transfers
Net ecosystem production = Gross Primary Production – Terrestrial Ecosystem Respiration
Balance of lateral imports and exports
= Carbon returns – carbon ‘uses’
61 parameters were mapped to produce carbon accounts in the EU countries, following CBD-ENCA ‘Quick-start package’
The main parameters were evaluated against independent reference data to assess their quality and test them for estimating consistent carbon accounts.
The two basic balancing items are designed to summarize ‘vertical’ and ‘horizontal’ carbon transfers

13. Estimation of NECB – Sinks and sources!
The maps shows a decade average, with areas in green indicating prevailing sink (most of Europe) and in red – prevailing source functions (e.g. parts of North West Europe, Po valley in Italy, and spots of forest-burned areas of Portugal).
Statistics can be extracted on the carbon storage and sequestration at various scales and units, for example countries, NUTS, protected areas etc
Statistics can be extracted on the carbon storage and sequestration at various scales and units, for example countries, NUTS, protected areas etc

14. 7. Concluding remarks
Strong movement to expand from land-cover to stronger ecological foundations
Scale
Work is being done at many scales
Challenge is linking classifications across scales and geographical domains
Common methods for mapping ecosystem units are emerging
Accounting criteria useful for improving consistency, comparability and methodological standardization
Challenges related to estimating error in both spatial and tabulated data
Sinks and sources
Spatial link to supply and use accounts

15. Thank You!