5 Preface Executive Summary

Remote Sensing and REDD

Remote sensing (RS) is of central importance to the development of REDD at both the project level and the national level. It allows regular and standardized monitoring of deforestation and to a lesser degree of forest degradation.

It should be kept in mind that RS technology is generally not capable of direct measurements of biomass or carbon, which is what is ultimately required for REDD. There are some emerging technologies that are beginning to address this; however, for the time being remotely sensed data needs to be supported by ground-based fieldwork in order to generate estimates of biomass and carbon.
The standing trees that are generally the target of remote sensing are not the only stock of carbon in a forest system; carbon accounting requires that other stocks also be incorporated. The other relevant carbon pools are below-ground biomass (i.e. root systems), dead organic matter on the forest floor (i.e., leaves and litter), and soil carbon. Below-ground biomass and surface organic matter can often be estimated due to close variation with above-ground biomass (e.g. root systems tend to have about 20% as much biomass as the standing vegetation, while surface organic matter tends to be between 10% and 20%⁹⁰). Soil carbon cannot be estimated as easily from the standing vegetation, and so must be surveyed on the ground.
From an accounting point of view, projects are able to ignore carbon pools if they can demonstrate that doing so will only reduce their estimates of carbon reductions.⁹¹ However, more complete carbon accounting will lead to more credits, meaning that greater accuracy may be worth the investment for projects. The “conservativeness” principle suggests that emissions reduction credits will be reduced relative to the degree of imprecision on project measurement. Projects with greater measurement precision will therefore be able to generate more credits (Figure 1).⁹²




Figure 4-1: Two project scenarios – high uncertainty and low uncertainty – show the same absolute reduction in carbon emissions; however the project on the right with lower measurement uncertainty is able to claim far more emissions reductions.

4.1 RS products available

There is a great variety of remotely sensed data available to assist the monitoring of above-ground carbon stocks. One of the most important distinctions between the many products is spatial resolution. Higher resolution enables the detection of more subtle changes in forest cover, such as small clearings or forest degradation. Products also vary in their spectral resolution; that is the number of frequency bands they are able to detect and record information on. Spatial and spectral resolution can be independent of each other. For example, the Landsat ETM+ sensor has coarser spatial resolution (30m x 30m pixels) than does SPOT-5 (10m x 10m); however, Landsat ETM+ records information from 7 different spectral bands, while SPOT-5 only records 4. This difference in spectral resolution can have important consequences for the ability to process the data further; for example, the suite of Landsat spectral bands can be used together to generate a useful indicator of forest degradation that will be discussed below.⁹³

All satellite images are produced by one of three broad categories of RS sensors: optical, radar, and laser. Optical sensors are referred to as ‘passive’ because they form their images based on light naturally reflecting from the earth’s surface in the same way a photograph might. Radar and laser sensors, on the other hand, are ‘active’ in that they send microwaves, radio signals, or laser light to the surface of the earth and measure the reflection of those emitted signals.

Type	Examples	Resolution	Benefits and limitations	Cost
Optical	MODIS	250m (coarse)	 Useful for large-scale frequent sampling (e.g. is used for Brazil’s near real-time deforestation monitoring program DETER).  Often used for monitoring fire frequency.  Resolution too coarse for monitoring small and medium-sized deforestation.	Low or free
	Landsat; SPOT; DMC	30m (medium)	 Widely-used. Landsat has continuous data since 1970s. – Capable of detecting degradation to a limited degree.  Landsat ETM+ developed problem in 2003 meaning all subsequent data requires image correction.	Archived Landsat data will be free from 2009. Otherwise <$0.001 / km^2 for older data, or $0.5 / km^2 for recent.
	Quickbird; IKONOS	<5m (high)	 Accurate picture, degradation can be assessed.  Can be used to train other images.  Expensive.  Demanding to process	$2-$30 per km^2
Radar	ALOS PALSAR	~30m	 Can generate images through cloud and haze.  Demanding to process.	Wide range.
Laser	GLAS LiDAR		 Potential to measure biomass and forest structure directly.  Difficult to implement in areas with steep slope.	Very expensive.

Table 4-1: Summary of selected sensors ⁹⁴

4.1.1 Optical sensors

To date, optical sensors have been used for almost all land cover analysis. The most common sensor used is Landsat, various generations of which have been in operation since 1972. There is a wide range of optical sensors available from high to low resolution (Table 2). They have the advantage over radar and laser sensors of having a longer historical record and of being more straightforward to process.⁹⁵

4.1.2 Radar and laser sensors

Because they rely on the transmission of radio waves, radar sensors have the great advantage of being able to penetrate smoke and cloud cover. This avoids the complicated image corrections and the need to collect multiple images that occurs when optical sensors encounter cloud cover. Radar is thus a very promising option for future forest monitoring. At this point, the main downside is that it is more complicated to use than optical imaging, and does not have the historical record that Landsat does.⁹⁶

Laser, or lidar, sensors are still very much an emerging technology as far as forest monitoring is concerned; however they may play a more significant role in the future. They are able to determine more about the structure of a forest than other sensors, and as such have great potential for being able to measure biomass directly. This would make them the ideal technology for REDD; however, at present lidar is too expensive to be practical in a general sense. To date, lidar has not been used for any large-scale deforestation monitoring.⁹⁷