|
|
|
# Options for sensitivity tests
|
|
|
|
|
|
|
|
We have described a general [methodology](home/Methodology), but we have left out some details which open different options and choices. Depending on these choices, we can construct different forcings to compare to each other. First, there are different choices and options possible to calculate the average of a variable at the polygon level before calculating a bias.
|
|
|
|
We have described a general [methodology](home/Methodology), but we have left out some details which open different options and choices. Depending on these choices, we can construct different forcings to compare to each other. First, there are different choices and options to calculate a variable's average at the polygon level before calculating a bias.
|
|
|
|
|
|
|
|
## Choice of integration and bias calculation at the polygon level
|
|
|
|
|
|
|
|
SAFRAN variables are given by polygon and by altitude classes (corresponding to every 300m). For the model km-scale outputs, we have access to the orography and the altitude associated to each point. We can make different choices, favoring either the altitudinal distribution of SAFRAN or the one from the model outputs. This choice only impacts the part of the methodology associated to the bias calculation and its spatial disaggregation (steps 1, 2 and 3 in the Figure in [methodology](home/Methodology)). The rest of the methodology is not impacted.
|
|
|
|
|
|
|
|
1. _**Bias by altitude class**_
|
|
|
|
In this case, we calculate one bias by polygon and by altitude class. We first attribute each grid point of the model to a given altitude class. Then the model values are integrated at the polygon level with the average calculated for the point within the altitude class (Fig. 5.9). The bias between SAFRAN and the model (step 2 in the Figure in [methodology](home/Methodology)) is calculated for each class and for each polygon.
|
|
|
|
Finally, when we spatially disaggregate the bias for each polygon and altitude class, we apply a given bias to all the points within the polygon and the altitude class. With a correction by altitude class, this choice favors the altitudinal distribution in SAFRAN and accentuates a contrast between altitude classes (Fig. 5.10a).
|
|
|
|
Since the model orography may not be exactly similar to the one used to construct SAFRAN altitude classes, some points in the model grid are associated with an altitude not covered by the altitude classes of SAFRAN. Therefore, these points don’t have a calculated bias, and we attribute to them the average bias value of neighbouring points.
|
|
|
|
|
|
|
|
2. _**Average bias over the polygon**_
|
|
|
|
In this case, we calculate one bias per polygon, covering all altitude classes. The model values are integrated at the polygon level and averaged over all altitude classes, weighted by the relative ratio of each class within the polygon (Fig. 5.9). The bias between SAFRAN and the model (step 2, Fig. 5.5) is calculated once on this average for each polygon.
|
|
|
|
|
|
|
|
## Analogue days
|
|
|
|
|
|
|
|
Then for the moment, we have presented a methodology which works to construct a forcing over the period common to both initial datasets (SAFRAN and CPRCM run). However, here we introduce another method based on analogue days, which would allow to extend the methodology to the full period covered by SAFRAN (1979-2014). |
|
|
\ No newline at end of file |
|
|
|
Then for the moment, we have presented a methodology that works to construct a forcing over the period common to both initial datasets (SAFRAN and CPRCM run). However, here we introduce another method based on analogue days, which would allow to extend the methodology to the full period covered by SAFRAN (1979-2014). |
|
|
\ No newline at end of file |
