Speaker
Description
Different disturbance events lead to a varied response of terrestrial biomass, which regulates the terrestrial ecosystems' short- and long-term carbon cycle dynamics. Quantifying the disturbance regimes is essential to understanding and reducing the uncertainty of vegetation mortality and its effects on biomass. Based on the synthetic exercise, we revealed a strong link between three disturbance regime parameters, $\mu$ (probability scale), $\alpha$ (clustering degree), $\beta$ (intensity slope), and the spatial pattern of emergent biomass. Relying on this connection, we are applying the real-world observation from a high-resolution biomass dataset, the GlobBiomass with a spatial resolution of 25 m, to infer the regional disturbance regime.
We first compared the remote sensing-based and model-simulated biomass statistics, the significant difference within several statistics that played an important role in predicting disturbance regimes was identified. This mismatch would lead to model extrapolation in predicting realistic disturbance regimes. To avoid this, we carried out a series of refinement exercises, including increasing the simulation scenarios for the disturbance regime, the process of photosynthesis, the recovery rates, the shapes of the disturbance shapes, and adjusting the spatial resolution of simulation and observation. In the end, we found that the difference comes from the mismatch in spatial representation, and the aggregation process could significantly reduce these differences. We then quantify the discrepancy between observation and simulation across gradient scales in different regions of the world to determine the best degree of aggregation. Finally, the aggregated statistics from the remote sensing observations were integrated into the trained machine learning parameter inversion model at the best degree to produce the spatially continuous distribution of the three disturbance regime parameters.
Given the novelty of assessing disturbance regimes with high-resolution biomass data, Our study provides insight into how to reduce the difference between model simulation and observation to avoid extrapolation, and offers opportunities to evaluate and improve the representation of disturbance dynamics in dynamic vegetation and Earth system models.
