Abstract
This discussion paper addresses (1) the challenge of concisely reporting uncertainties in forest remote sensing (RS) studies, primarily conducted at plot and stand level, and (2) the influence of reference data errors and how corrections for such errors can be made. Different common ways of reporting uncertainties are discussed, and a parametric error model is proposed as a core part of a comprehensive approach for reporting uncertainties (compared to, e.g., conventional reporting of root mean square error (RMSE)). The importance of handling reference data errors is currently increasing since estimates derived from RS data are becoming increasingly accurate; in extreme cases the accuracies of RS- and field-based estimates are of equal magnitude and there is a risk that reported RS accuracies are severely misjudged due to inclusion of errors from the field reference data. Novel methods for correcting for some types of reference data errors are proposed, both for the conventional RMSE uncertainty metric and for the case when a parametric error model is applied. The theoretical framework proposed in this paper is demonstrated using real data from a typical RS study where airborne laser scanning and synthetic aperture radar (SAR) data are applied for estimating biomass at the level of forest stands. With the proposed correction method, the RMSE for the RS-based estimates from laser scanning was reduced from 50.5 to 49.5 tons/ha when errors in the field references were properly accounted for. The RMSE for the estimates from SAR data was reduced from 28.5 to 26.1 tons/ha.
Highlights
Remote sensing (RS) is used in many forest applications, often in combination with field sample plots to obtain complete maps for some variable of interest (VOI)
A parametric error model is proposed as a core part of an approach to describe and quantify uncertainties. We argue that this approach has several advantages over the use of traditional metrics, such as the root mean square error (RMSE)
Note that the approaches we describe are all related to study set-ups where RS-based estimates at plot or stand level are available from any arbitrary RS-based methodology, and the objective is to assess the uncertainty of the estimates through comparing them with a set of field references
Summary
Remote sensing (RS) is used in many forest applications, often in combination with field sample plots to obtain complete maps for some variable of interest (VOI). A typical RS research paper in the forestry field applies the following suit of methods: (1) inventory a number of field sample plots for the VOI; (2) extract the corresponding plot level RS data; (3) estimate a model that relates the RS data to the field sample data; (4) apply the model to the entire RS dataset; and (5) quantify the uncertainty of estimates at plot or stand level by comparing the RS-based estimates with field reference data [1,2,3,4,5,6,7] This procedure can be repeated for different sensors, methods, regions or acquisition properties, in order to assess strengths and limitations of one or more RS-based methods of interest. The assumed conditions are, rarely relevant for studies related to remote sensing of forests
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
