is useful to note the relative differences in error with lead-time among ensemble control (CTRL) and a sample individual ensemble member. It is also useful to understand how the errors grow and saturate at different spatial scales.
The diagram below was derived from cycles before CY48r1 when the medium range ensemble had 18Km resolution.
Fig6.2.4-1: Plot shows the contributions to the forecast errors calculated for 500hPa heights over the Northern Hemisphere during the period Dec 2016-Feb 2017. These are plotted against forecast lead-time for ensemble control (red) and a single ensemble member(blue). Spatial scales greater than about 150km (wave numbers 0-63) are considered here.
- full lines: squared forecast error ((Forecast-Analysis)2.
- dotted lines: "activity" in the forecast (2x(Forecast-Climatology)2.
- dashed horizontal lines: "activity" in the analysis (2x(Analysis-Climatology)2.
- All values are normalised by the same amount so that the maximum value plotted is 1.
- The dots on the error curves indicate that the difference is statistically significant.
Fig6.2.4-1 shows the difference between ensemble control (red line) and a perturbed ensemble member (blue line). Comparing lead-times when each reach the same error show for this range of wave numbers:
- Up to about Day2.5:
- the ensemble control and a single ensemble are similar; neither is better (the thick and thin full lines are on top of each other).
- the error contributions to the total error are low and similar (red and blue dotted lines on top of each other).
- Beyond Day2.5:
- the ensemble control becomes just short of a day better (by Day10) than a single ensemble member for the same errors.
- the error contribution from the ensemble control (red dotted line) and an ensemble member (blue dotted line) remain very similar.
The ENS is designed to explore the less likely outcomes and, taken as a whole, generally contains more information than the ensemble control (Cy48 or earlier) or the HRES. On the other hand, at short lead-times, the ensemble control (or the HRES) can contain additional information at smaller spatial scales.
Note: these diagrams are based on earlier cycles of the IFS when the ensemble control and ensemble members had a resolution of 18km (and HRES a resolution of 9km).
Fig6.2.4-2: Plot shows the contributions to the forecast errors calculated for 500hPa heights over the Northern Hemisphere during the period Dec 2016-Feb 2017. These are plotted against forecast lead-time for ensemble control (red) and a single ensemble member(blue). Wave numbers 0-3 and 4-14 are considered.
- thick lines; planetary-wave scales (scales greater than about 2500km, zonal wave numbers 0-3).
- thin lines; synoptic scales (scales between 650-2500km, zonal wave numbers 4-14).
- full lines: squared forecast error ((Forecast-Analysis)2.
- dotted lines: "activity" in the forecast (2x(Forecast-Climatology)2.
- dashed horizontal lines: "activity" in the analysis (2x(Analysis-Climatology)2, normalised.
- values are normalised by the same amount so that the maximum value plotted is 1.
The dots on the error curves indicate that the difference is statistically significant.
Fig6.2.4-2 shows the difference between ensemble control (red lines) and a perturbed ensemble member (blue lines). Comparing lead-times when each reach the same error show for these ranges of wave numbers:
- Up to about Day7:
- the ensemble control (red):
- for wave numbers 4-14 (synoptic scales - thin line) and wave numbers 0-3 (planetary scales - thick line) are similar; neither is better (the thick and thin full lines are on top of each other).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and analysis are low (red dotted lines).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and climatology rise steadily (red thin dashed line).
- an ensemble member (blue):
- for wave numbers 4-14 (synoptic scales - thin line) and wave numbers 0-3 (planetary scales - thick line) are similar; neither is better (the thick and thin full lines are on top of each other).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and analysis are low (blue dotted lines). Slightly greater than for ensemble control.
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and climatology are (blue thin dashed line). Very similar to ensemble control.
- Beyond Day7:
- the ensemble control (red):
- by day10 wave numbers 4-14 (synoptic scales - thin line) becomes just short of a day better than for wave numbers 0-3 (planetary waves - thick line).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and analysis remain fairly low (thin red dotted line).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and climatology rises steadily (red thin dashed line). Approaches the atmospheric activity and implies ensemble control has little skill.
- an ensemble member (blue):
- for wave numbers 4-14 (synoptic scales - thin line) and wave numbers 0-3 (planetary scales - thick line) are similar; neither is better (the thick and thin full lines are on top of each other).
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and analysis are low (blue dotted lines). Slightly greater than ensemble control.
- for wave numbers 4-14 (synoptic scales - thin line) the error contributions to the total error between forecast and climatology rises (red thin dashed line). Steady, better than ensemble control with some skill retained.
The ENS is designed to explore the less likely outcomes and, taken as a whole, generally contains more information than the ensemble control (or HRES). On the other hand, at short lead-times, the ensemble control (or HRES) can contain additional information at smaller spatial scales. On the other hand, the planetary scales continue to include useful information beyond Day7. Their error continues to increase towards the (higher) level of planetary scale activity (thick dotted and dashed lines) but not as extreme as for synoptic scales.
Note: these diagrams are based on earlier cycles of the IFS when the ensemble control and ensemble members had a resolution of 18km (and HRES a resolution of 9km).
(FUG Associated with Cy50r1)