In-situ observations on a moderate resolution scale for validation of the Global Change Observation Mission-Climate ecological products: The uncertainty quantification in ecological reference data

Abstract

We report the ground validation activity for the terrestrial ecology products (leaf area index, above-ground biomass, and fraction of absorbed photosynthetically active radiation) of the Second-generation Global Imager (SGLI) on JAXA’s satellite named “Global Change Observation Mission-Climate.” We gave special attention to quantifying the uncertainty propagating from errors in the ecological reference data (ERD) obtained by the field work. Specifically, for optimal design and practical implementation of the field work with small uncertainty and small cost, we proposed: 1) a practical target which defined the accuracy threshold of ERD as a quarter of the satellite accuracy threshold, and 2) a calculation method of the uncertainty quantification of ERD by accounting for the uncertainty propagating from the empirical regression equations (such as allometry equations) and the statistical distribution of the population. As a result, we obtained ERD for GCOM-C/SGLI in various plant functional types (a deciduous needle-leaved forest, a deciduous broad-leaved forest, an evergreen needle-leaved forest, and dry and wet grassland) with sufficient quality, especially with a coverage area of 500 m × 500 m which can include a footprint of the sensor (250 m × 250 m) in any situation. We demonstrated: 1) the accuracy target was the key decision to make the practical calibration/validation work, 2) the regression uncertainty had a large impact, although little literature provided sufficient ancillary data about the regression equations necessary for quantification of the uncertainty, and 3) the optimal protocols of ERD observation can change depending on situations (plant functional types, phenology stages, type of products, accuracy targets, resources, and development of observation instruments and techniques); hence the choice should be made on the basis of quantification of the uncertainties.