Large-scale validation of primary BIOMASS forest products and analysis of their relationships with similar products in other SAR missions.

European Space Agency

Overview
The overall objective of the project is (1) to validate the two main BIOMASS Level-2 products above-ground biomass (AGBD) and upper canopy height (FH), and (2) contribute to test the utility of other SAR-based missions for relationships with the BIOMASS products AGBD, FH and areas of forest clearing (FD), and upscaling to the SOTR-radar areas where BIOMASS cannot deliver products.

The validation will be based on large-scale field sample surveys. Model-assisted estimators, which rely on the probabilistic nature of the sample data for the inference but take advantage of the information carried by the auxiliary data (e.g. AGBD and FH maps) for improved precision of the estimates, will be adopted in the analysis. Design-unbiased estimators for mean values (AGBD, FH) and variance exist for different statistical designs. Because of the unbiasedness of the estimators of the mean value, model-assisted estimators can be used to estimate systematic map error and inference for the systematic error in the form of e.g. a confidence interval. Such estimated systematic map errors can subsequently be corrected via a locally/regionally calibrated model, and uncertainty of the mean value over a larger territory can be re-estimated to demonstrate the improvement of a map by local or regional map calibration. We will perform such analyses across a range of biomes and geographical regions across the global south. Field data from probabilistic sample surveys conducted as part of national forest inventories in selected countries will form the basis for the analysis. Data from Brazil (1.8M km²) and other countries in Africa and South-East Asia will be analyzed, offering insight into the performance of the BIOMASS Level-2 products in different biomes in the global south. The major deliverables from the analysis will be estimates of systematic map error and precision of the AGBD and FH Level-2 products, both for uncalibrated maps and for locally/regionally calibrated maps.

In addition, we will explore the statistical relationships between the BIOMASS Level-2 AGBD, FH and FD data and data extracted from Sentinel-1 and NISAR. We will evaluate the feasibility of using models based on these SAR missions for redundancy to the BIOMASS mission. For NISAR, we will use Level-1 wrapped interferograms or Single Look Complex data. For Sentinel-1 we will set up a processor where we form repeat-pass data pairs and extract interferograms and coherence from them. We will carry out statistical analyses on the relationships between the BIOMASS data and the data extracted from these two other SAR missions. We will evaluate the model performance for possible model deployment. We foresee that model performance will vary between forest types and between seasons. Hence, we will identify under which conditions a given model is sufficiently accurate. Ultimately and in the best case we might find areas, forest types and seasons where our models could be used for filling BIOMASS voids in the SOTR conflict areas in North America, Europe and around Taiwan. The deliverables will comprise several reports on the statistical analysis results together with model performance evaluation. We will carry out these analyses three times, i.e. for each of the three BIOMASS output data sets that are expected in August 2026, May 2027 and February 2028.