3. Remote sensing material

3.1. Background

The current and some coming remote sensing sensors and their specifications are described in Appendix 2. The image prices are given when they are available. The remote sensing data can be divided into different categories using many criteria. The most important criteria are listed below.

The wavelength range: In principle, the remote sensing data can be based on any physical mechanism transferring energy between the target on the ground and the sensor on the remote platform (satellite, airplane, helicopter, balloon, etc.). In practice, the choices are limited by sensor technology, target properties, source of energy, the atmosphere between the target and the sensor, and the interaction of the target and the transfer mechanism. The only feasible alternative is electromagnetic radiation of certain wavelengths (optical/infrared, thermal, microwave region).

Depending on the wavelength, the electromagnetic radiation may be available from a natural source or it must be transmitted by the instrument. If the radiation is available from natural sources, the instruments are called passive. These instruments are found in the optical and thermal spectral regions and short wavelength microwave regions. Some instruments transmit radiation and measure the amount of radiation returning from the target. These instruments are called active and they can be found in the microwave region and optical/infrared region.

The usefulness of different wavelengths in forest analysis varies. The reason for this is in the interaction between the radiation and the different constituents of the forest. The most useful wavelengths are in the visible (VIS, wavelength 400-700 nm) and infrared region (near infrared NIR 700-1000 nm and, to a lesser extent, shortwave infrared SWIR 1000-2500 nm). The interaction in the microwave region is weaker.

The pixel size: The size on ground of the picture elements (pixels) produced by different sensors varies very much. The smallest pixels in current optical instruments are about 60 cm whereas the largest pixel sizes in passive microwave instruments are in the order of tens of kilometres. In some cases (e.g., passive microwave instruments) the pixel size is limited by the physics whereas in other cases (e.g., the weather satellites) the pixel size is limited by the application properties.

The ground resolution of an instrument is more difficult to define than the pixel size. For any reasonable imager the resolution is close to the pixel size and it is sufficient for this discussion to use the pixel size for characterisation of the instruments.

The most useful pixel size for forest applications depends on the information needs. The spatial units being investigated should usually include at least a small number of pixels. The number of pixels falling on the borders between the units should not be too large.

The image size grows rapidly when the pixel size is made smaller. This limits the useful resolution because the processing load must be within the economical limits set by the application. However, one should not that the processing capability of the computers and the sizes of storage media increase rapidly. This allows use of better resolution in future if the human operating work does not grow.

The data transfer capabilities between satellites and ground stations are limited. This sets a limit to the amount of information that can be collected by a sensor in a unit of time. This means that with a smaller pixel size the area covered by the images gets smaller and the number of images needed to cover a specified are gets larger.

The radiometric characteristics: The power of the electromagnetic signal available at the sensor element of a passive instrument depends on the pixel size and the wavelength range. Active instruments add the transmitted power to this list. The radiometric (brightness) resolution in a data pixel depends also on the sensitivity of the sensor system.

In practice, the radiometric accuracy of the existing sensors can be considered adequate but, in some cases, it limits the usability of the data in some applications. For instance, the reflectivity of forest targets is low in the visible range and the radiometric resolution of a data source may limit the obtainable accuracy when using channels from the visible range. This is significant especially at high latitudes where the sun elevation is low.

Availability of the images: The availability of images from different sensors is limited by several factors. Some of these limits come from the nature and some of the factors from commercial and organisational sources.

The natural limitations are different for different wavelengths. Acquisition of the optical images is limited by the clouds. The severity of this limitation depends on the area but it is a considerable limitation for all areas with forest cover. Another limitation of the passive optical sensors is that they can operate only in daylight conditions. The elevation of the sun has effect on the usability of the images and this further limits the availability of the images.

The microwave instruments are nearly insensitive to weather conditions within the atmosphere. However, the moisture of the target affects the images strongly.

The time one satellite can collect images during one orbit may be limited by data storage capacity within the satellite, the downlink capacity, or by power conditions. Together with the pixel size, this limits the area the satellite can image within a time unit. Some satellites have a limited pointing capability and this increases the possibilities to obtain an image from a certain site if the pixel size is small. However, this off-nadir viewing affects the quality of the images.

The satellites usually acquire images based on programming (a customer orders an image covering a certain area and the satellite is tasked to do this) and as background activity (the extra instrument operating time available after specific orders is used to collect images from areas that may be used in future). The images from both acquisition types are usually archived for possible later use. The availability and the condition of these archives must be checked if a system is to use existing imagery.

In practice, the availability of images for an application may be limited by the price of images. This applies mostly to the sensors developed specifically for commercial applications (e.g., sensors with small pixel size).

Availability of images on a long-term basis is an important consideration when building a monitoring system. Continuity requirements may be fulfilled either by a series of similar instrument or by different instruments providing similar capabilities. The design life of one remote sensing satellite is typically about five years. However, a satellite may operate much longer or it may fail sooner. This is not known in advance.

In addition to radiometrically and/or geometrically corrected images, some satellite systems provide so called Level 3 image products. One example of these is the MODIS land cover product which classifies the land cover into 17 classes.

Pricing: The price of image data varies according to the instrument but also according to many other factors. These include the receiving station, geographic area, customer type, archive vs. tasking, acquisition priority, and number of images ordered. The prices given below indicate the order of magnitude of the prices of the different products when writing this document.

3.2. Sensor Classification

Classification based on pixel size is shown in Table 2. Another classification done on the basis of the information mediator is in Table 3.

Table 2: The sensor classes

Table 3: The information mediator based instrument classification

Classifications for applications can be derived based on these classifications and algorithmic and system consideration. If an application requires a certain set of channels within some pixel size and swath width limits, these constraints define a class that contains one or more instruments (or none if the constraints are too tight).

A practical classification useful for predicting the availability and price level of the data comes from the purpose of the satellites. This classification divides the satellites into four classes: experimental, research, non-commercial operational, and commercial operational.

The amount of data available from experimental satellites is usually small and there is no guaranteed continuity. This data may be useful locally in some special cases.

The research satellites may be specialised or general. Some produce only a small amount of data and are useful in special cases. Some others (like NASA’s MODIS and ESA’s Envisat) produce large amounts of data worldwide. The price of data from the research satellites is usually low (handling costs) but the continuity of similar data after the satellite has failed is not guaranteed.

The non-commercial operational satellites (like the weather satellites and Landsat) produce large amounts of data and the continuity is good. This usually means also that there exist large archives of images from previous years. The price of data from these satellites is low because the governments are paying most of the costs.

The commercial operational satellites (e.g., Spot and the very-high resolution satellites) include the costs of the satellites and operation into the image prices. The continuity is good if the commercial demand for the image type continues.

Note that inclusion of a satellite into one of the two operational categories may depend on the nationality of the data buyer. In many cases, the buyers from countries financing the satellite are charged non-commercial price whereas others pay commercial prices.

3.3. Covering the globe with image samples

Taking into account the objective of the global survey, the price of the images and needed workload, sampling may be the only feasible way to utilise remote sensing data in the case of high and very high resolution remote sensing data. Total cover is feasible with medium resolution data, such as MODIS. Examples of required image numbers and prices are given in Table 4 with some sampling options. MODIS, Landsat ETM and Ikonos are used as examples. The price vary by region and dealer, a price of 600 US$ for Landsat ETM and 2700 US$ for Ikonos have been applied in the example. MODIS images are supposed to be charge free. The total price in the example is dominated by the Ikonos images.

Table 4: The number of the needed images estimated costs with 0.1 % and 1 % Ikonos sampling options