Remote sensing is based on the principle of measuring reflected electromagnetic energy from a target under investigation. The source of the energy can either be energy (photons) from the sun or an artificial source, like a laser pulse in the case of a LiDAR instrument. However, the energy measured by the sensor interacted with the target, as well as having travelled through layers of the atmosphere and sensor components. Pre-processing, or the process of transforming the measured energy to levels as it were just after being emitted or reflected from/off the target, therefore is required before data are used. This process ensures compatibility between different sensors and even different times, and includes efforts related to atmospheric compensation, e. g., conversion from digital numbers (DN) to radiance (W m“2) to reflectance (unitless). Feature extraction then measures the object under investigation (target), which could be a pixel, object, single tree crown, stand of trees, etc. The process is validated by comparing the extracted result with measured data from a field survey.

Inventory based on remote sensing data nearly always involves a multi-phase approach with a terrestrial component, used for calibration. In regional assessments, assumptions about the terrestrial component are made and the remote sensing

information is interpreted in accordance with measured point samples. Two phase inventories seem to be the most common option, in which detail terrestrial infor­mation is collected at a few sample locations, which is then used as training and verification of the image processing process (Fig. 2.4). A decision should be made during the planning phase on the appropriate sensor and time of acquisition, as well as the sampling strategy for field survey or ground truthing. In short, all remote sensing approaches or remote sensing inventory models should be calibrated and validated using a sample of field-measured values, especially when a model is used across different regions, site conditions, or species.