11:00 〜 13:00
[ACG38-P02] Improvement of a diagnostic terrestrial model, BESS, based on multiple observation constraints and sequential optimization
キーワード：Carbon Cycle, Remote Sensing, Model Optimization, Diagnostic Model, Gross Primary Production, Flux
This presentation will introduce our current work on improvement of a diagnostic terrestrial model, BESS (Breathing Earth System Simulator) to develop a GCOM-C SGLI GPP/NPP product. BESS can simulate terrestrial energy, water, and carbon cycle in a mechanistic way with inputs by remote sensing data. Compared with empirical models, the mechanistic model can provide more steady results based on its physical and mechanistic structure but the model requires more inputs. Such a lot of data requests, we have to gather the different sources from flux sites or satellites sometime. So, optimization is necessary.
Our target is trying to figure out the feasibility of optimization methods for mechanistic models with multiple observation constraints. In this research, we apply an idea to divide the outputs in 3 tiers as Net Radiation(Rn), Evapotranspiration(ET) and Gross Primary Production (GPP).
In the site-level research, we will select some representative flux sites. It will optimize the parameters only related to one output in one tier and settle them before the next tier optimization begins. In the end, we can settle some parameters to reduce the need of input data and make sure the result is close to the observation constraints.
As for expanding into a map with the satellite data source, we would like to implement the original BESS model as a GCOM-C product and conduct evaluations using eddy-covariance site data and existing MODIS products with the optimization algorithms.
Our target is trying to figure out the feasibility of optimization methods for mechanistic models with multiple observation constraints. In this research, we apply an idea to divide the outputs in 3 tiers as Net Radiation(Rn), Evapotranspiration(ET) and Gross Primary Production (GPP).
In the site-level research, we will select some representative flux sites. It will optimize the parameters only related to one output in one tier and settle them before the next tier optimization begins. In the end, we can settle some parameters to reduce the need of input data and make sure the result is close to the observation constraints.
As for expanding into a map with the satellite data source, we would like to implement the original BESS model as a GCOM-C product and conduct evaluations using eddy-covariance site data and existing MODIS products with the optimization algorithms.