- Calibration of the model. Novel statistical approaches may be necessary to
calibrate a model that includes many regional Human models. One possible
approach could be the use of modern data assimilation techniques such as
Ensemble Kalman Filter (EnKF) used in numerical weather prediction that have
the ability to estimate not only the model variables, but also unobserved
parameters. EnKF in principle could be used to perform an initial calibration of
the submodels of the system by estimating their optimal parameters. - Exploration of the dynamical behavior under different parameter ranges, such as
chaotic, oscillatory or stable behavior. The use of EnKF for calibration may
depend on whether the model is chaotic, in which case it is expected to work
well, or non-chaotic, in which case the behavior of EnKF is less well known. - Use of an ensemble of forecasts with different initial conditions and different
model parameters to explore and define the uncertainty of the projections of the
models. - Parameterization of the government policies (as sliding knobs in the model, or as
responses to the model evolution) so that trade-offs can be valued.
Figure 1: Schematic of a prototype of a fully coupled HES model. The left box represents an
example of a geographically distributed Earth System model, , and the right box is one of the
regional Human System models. The arrows represent one or two-way coupling, and the
horizontal red arrows represent trade and migration between different regions. Credit: Eugenia
Kalnay and Jorge Rivas.