healthy/unhealthy plants are challenging, particularly using messy real-world
data. Thus, understanding how the health of indicator species can give early
warning of problems in an ecosystem will require new mathematical and
statistical tools.
Finally, we need better methods to figure out just how good our
mathematical models are. For example, we have models that describe how
substances move between the atmosphere, land and oceans over time, using a
series of non-linear ordinary differential equations. This cycle is critical for marine
life, which relies on the balance between nutrients, phytoplankton, and
zooplankton, which are in turn influenced by temperature, light, and resource
availability. The equations of the model are, of course, approximations, but
standard analyses commonly ignore this, and the result is that the model doesn’t
properly predict the size distribution of phytoplankton and zooplankton. Similar
difficulties arise in the highly complex problem of modeling air quality. If air
quality is substantially worse than our large reaction-diffusion models predict, we
need to know whether that is because the underlying dynamics are different from
what we thought, because of inherent randomness in the system, or because our
models captured those dynamics badly. Models have become vastly more
complex over time, making existing methods of evaluation inadequate.
Mathematical sciences challenges in the area of measuring and
monitoring progress toward sustainability include development of the
following:
- Tools for uncertainty quantification via probabilistic modeling
approaches, including tools to deal with the following challenges in this
area alone:
9 Characterizing the bias or discrepancy between models and reality
(data);
9 Recognizing that cost constraints often mean that models can only
be run for certain combinations of input parameters, requiring
extrapolation of model output to other input parameters;
9 Accounting for uncertainties in the initial conditions;
9 Estimating unknown parameters in the process models;
9 Accommodating stochastic features of the process models;
9 Producing predictions that arise by combining models and
observational data, as might occur via data assimilation methods.