Home page banner
Follow us on

Twitter Blogger Blogger


Google Earth and Forest Monitoring

Date: February 13, 2010*

During COP-15, Google announced a new Google Earth application that enables observation and measurement of changes in forest cover, and the UN-REDD Programme has agreed to work with Google to test their prototype in Africa.

Maurizio Teobaldelli, Senior Programme Officer for the UNEP’s World Conservation Monitoring Centre (WCMC) in Cambridge, UK, provides his analysis.

Maurizio Teobaldelli
During the recent Climate Change Summit (COP15) Google’s philanthropic arm, Google.org,  presented a new Earth observation application, developed together with the Carnegie Institution for Science, Imazon (Instituto do Homem e Meio Ambiente da Amazônia) and the Gordon and Betty Moore Foundation. The system will enable users to assess online, global-scale land cover change. Google’s system could therefore be suitable for supporting future forest monitoring as part of REDD, a key area for the UN-REDD Programme.

The strength of Google’s prototype is in combining innovative forestry science, including SAD (Sistema de Alerta de Deforestation) and CLASlite (Carnegie Landsat Analysis System–Lite), with technology resources, such as  raw satellite imagery datasets and a high-performance satellite imagery-processing engine running online in the Google Network System (Fig. 1). The application, which is not yet accessible to the general public, is expected to be more broadly available in the future as a not-for-profit service. So far, the prototype has only been released as a closed beta version to a select group of partners as a user test.

Google and the UN-REDD Programme agreed to test their prototype in Tanzania (one of the UN-REDD pilot countries) and also to include some applications from FAO Forest Resources Assessment.

The Imazon’s SAD, led by Carlos Souza, has been capable of generating deforestation maps and statistics of the entire Brazilian Amazon on a monthly basis since 2008, using MODIS sensor images. CLASlite, led by Greg Asner, provides a set of ecologically meaningful images identifying forest cover, forest cover changes and human-induced disturbances, such as extensive high-impact logging or repeated fires. In particular, CLASlite firstly harmonizes and processes remotely-sensed data (up to 10 images) acquired by different satellite platforms  and then automatically extracts three different land cover classes (live vegetation, non-live vegetation, and bare substrate) at sub-pixel resolution, providing at the same time any error in the estimation. 

Google’s prototype could be used to assess forest land cover changes which could then be integrated with active remotely-sensed data, such as LiDAR, Light Detection and Ranging, or RADAR, Radio Detection and Ranging and ground measurements to build biomass density maps of tropical forests.

Google’s prototype could be used to assess forest land cover changes which could then be integrated with active remotely-sensed data, such as LiDAR, Light Detection and Ranging, or RADAR, Radio Detection and Ranging and ground measurements to build biomass density maps of tropical forests.

It is not yet known which rules and definitions will be used to set up an MRV system in REDD and whether or not they will draw on past experience. However, the Conference of the Parties does request the use of “the most recent Intergovernmental Panel on Climate Change (IPCC) guidance and guidelines, as adopted or encouraged by the Conference of the Parties, as appropriate, as a basis for estimating anthropogenic forest-related greenhouse gas emissions by sources and removals by sinks, forest carbon stocks and forest area changes.”.

Fig.1: Selected results of the Google’s prototype showed during the COP15 in Copenhagen (December, 2009). CLASlite online (a) detects “deforestation” and “forest degradation” in Rondonia, Brazil from 1986 to 2008 whereas SAD online (b) shows a region of recent “deforestation” in Mato Grosso, Brazil. Credits: Google

According to the IPCC guidance, in order to estimate carbon stocks associated with human-induced activities it is necessary to collect activity data, that is, information representing existing land-use categories including classification, area data and sampling. This information, represented by maps or tabulations, varies in how and when it is collected by countries. They can have different reporting frequencies and different attributes, such as annual censuses, periodic surveys and remote sensing. In fact, characterizations of land use include considerations that go beyond only the observed biophysical cover on the earth’s surface; functions describing land use in an economic context, and activities that are defined as the combination of actions resulting in a certain type of product, are needed to define the parameters of land use and land use change.

Once countries have collected activity data, in terms of area of Land Use and Land Use Change (LULUC), it is possible to multiply the activity data by a carbon stock coefficient or “emission factor” to provide the source or sink estimates.

Google’s system will not give information regarding emission factors, so in situ measurements, such as sample ground plots and national forest inventories are still necessary in order to acquire information on forest parameters. These parameters should be used in combination with other technologies including LiDAR or RADAR to assess carbon stocks and carbon stock changes over large areas at Tier 2 or at Tier 3 levels.

Based on these considerations, simply knowing about land cover and land cover change is not enough to monitor, measure, report and verify carbon stocks and carbon stock changes in REDD. Google’s system still does not incorporate in situ measurements, nor does it allow for the measurement of changes in forest-related GHG emissions. But they are developing and using a system to collect data from the field in one of the UNREDD Programme pilot countries, which uses mobile phones equipped with GPS.

In summary, a user-friendly system package, integrated in the Google Earth Engine, which allows non-expert users to quickly assess the regional distribution of tropical forest and changes in forest land, would be useful in supporting REDD. Furthermore, a not-for-profit system available online for scientists and national experts from tropical countries will be very useful in helping to lower the costs of forest monitoring.

Maurizio Teobaldelli is a MRV specialist in the UN-REDD Programme working since August 2009 as senior programme officer at the UNEP World Conservation Monitoring Centre, Climate Change & Biodiversity Programme in Cambridge, UK. Prior to this, Maurizio headed an Italian consultancy company (Studio Tecnico Foreco), and has also worked as scientific programme officer at the European Commission (DG-Joint Research Centre, Institute for Environment and Sustainability, Climate Change Unit). Maurizio holds a MSc in Forestry from the University of Florence, and a PhD in Forest Ecology from the University of Padua in Italy.

An FAO/FRA/RSS image in Google Earth as part of the UN-REDD/MRV cooperation with Google
Credits: Google