Geoscience Reference
In-Depth Information
1. Assessing the value of the forecast variables. The main goal of the evaluation
exercise is to evaluate quantitatively and qualitatively whether the modelling
system is successfully predicting the temporal and spatial evolution of a
particular process.
2. Determining the suitability of a specific application and configuration. Explore
the adequacy and correctness of the science represented in the model for the
purposes for which the model is applied. Comparison with other models in
addition to the observations can be helpful in identifying the strength and
weakness of the system.
3. Guiding improvement. Evaluation results should lead to new directions in model
development and improvement
A forecast system is judged by its ability to simulate the spatial and temporal
evolution of chosen forecast variables. In this regard, the metric is what drives
model development and optimisation. Hence, the evaluation metrics must be chosen
carefully.
The first evaluation is done right after the forecast period and depends on
observations that are made available shortly after they were taken. This type of
evaluation, sometimes referred to as verification, is generally part of the operational
forecasting process and is therefore done on a regular basis in near real-time.
The end result is the quantification of confidence and predictive accuracy of
the model products. An additional and different type of evaluation is where the
model's performance to simulate a given event or an annual cycle is examined in
depth. This case study evaluation can be made any time after the forecast period,
and observations that were not available for the near real-time evaluation can be
included. The purpose is to identify potential sources of error in order to improve
the model. In both cases, the evaluation process will depend on the intended use of
the forecast product.
10.5.2
Observational Data for Evaluation
The first problem in identifying appropriate routine measurements for evaluation
of dust models is the scarcity of observations of dust events. The location of the
main dust sources in unpopulated areas complicates the establishment of observing
networks.
Thus, the first option to address the evaluation of dust models has been the use of
satellite products. They have the advantage of a large spatial coverage (regional
to global), they are made regularly, and their observations are made available
to weather centres and other institutions in near real-time. Shortcomings include
satellite measurements' highly integrated nature, not only over the atmospheric
column but also over all aerosol components. Therefore, applications involving a
particular aerosol type (like dust) might be limited in some cases to seasons and
regions, when or where that type dominates the aerosol composition (Basart et al.
Search WWH ::




Custom Search