A new Approach to derive Productivity of Tropical Forests using Radar Remote

<p>Supplemental material to the publication " A new Approach to derive<br>Productivity of Tropical Forests using Radar Remote Sensing Measurements" in Royal Society Open Science</p><p>Abstract:</p><p>Deriving gross &amp; net primary productivity (GPP &amp; NPP) andcarbon turnover time of forests from remote sensing remainschallenging. This study presents a novel approach to estimateforest productivity by combining radar remote sensingmeasurements, machine learning and an individual-basedforest model. In this study, we analyse the role of differentspatial resolutions on predictions in the context of the RadarBIOMASS mission (by ESA). In our analysis, we use the forestgap model FORMIND in combination with a boostedregression tree (BRT) to explore how spatial biomassdistributions can be used to predict GPP, NPP and carbonturnover time (τ) at different resolutions. We simulatedifferent spatial biomass resolutions (4 ha, 1 ha and 0.04 ha) incombination with different vertical resolutions (20, 10 and 2m). Additionally, we analysed the robustness of this approachand applied it to disturbed and mature forests. Disturbedforests have a strong influence on the predictions which leadsto high correlations (R2&gt; 0.8) at the spatial scale of 4 ha and 1ha. Increased vertical resolution leads generally to betterpredictions for productivity (GPP &amp; NPP). Increasing spatialresolution leads to better predictions for mature forests andlower correlations for disturbed forests. Our results emphasizethe value of the forthcoming BIOMASS satellite missionand highlight the potential of deriving estimates for forestproductivity from information on forest structure. If appliedto more and larger areas, the approach might ultimatelycontribute to a better understanding of forest ecosystems.</p>