Section 4.2 Analysis Increments

Charts of analysis increments (e.g. real time example charts ) allow the user to see where the IFS model analysis has adjusted to observations. 

Large "increments" in geopotential, wind, temperature and humidity fields show where observations have caused the analysis to depart significantly from the background.  Often during manual analysis of model behaviour one will focus on upper level increments.  Alteration to the initial upper flow may well induce a corresponding modification in the evolution downstream as the forecast progresses.  Particularly large upper level increments, which can occasionally be seen, are often associated with areas of substantial, organised convection.  Examples are:

• across North America, notably the centre of the USA (e.g. Fig4.2-3 & Fig4.2-5),suggesting underestimation of the jet stream winds at short time ranges. Often the outflow from large, deep, energetic convective outbreaks distorts the upper flow.  The subsequent upper flow can differ significantly from earlier forecast runs. A downstream upper ridge may amplify and/or a perturbation may propagate downstream through the jet. Convective outbreaks over the central United States have thus been known to be "responsible" for lower forecast skill across Europe.
• over western Africa, again related to an area of convection (e.g. Fig4.2-7).  The westward propagation of this type of feature over the Atlantic can be an ingredient in tropical cyclone development (African Easterly Waves).
• over South America, notably Argentina, where extreme convection is quite commonplace during the summer half of the year

The user must decide whether

• the background field was possibly incorrect and the new observations have been used to improve the analysis. Or
• the observations were possibly incorrect and may have incorrectly influenced the analysis.  ECMWF "blocklists" observations or observation types if this happens repeatedly.  Blocklisted observations are excluded from the analysis until the quality improves.

In either case, the subsequent evolution should be carefully monitored, or even treated with suspicion, as the instability in the structure of the IFS atmosphere transfers downstream.  In some cases there can be jumpiness in the forecast conditions for several days later at locations well away from the initial differences.  For this reason it can help to inspect the analysis increment data before committing to a forecast.

Large increments in many variables are also sometimes seen where observations become available near a vigorous pressure system (e.g. dropsondes near a hurricane).  These indicate there were shortcomings in the background forecast of the feature that the analysis system is trying to reduce.

There are some model uncertainties in representing tropical convection - not deep/strong enough, with resulting weak outflow. 

Large increments are often associated with MCAs or areas of deep active convection and/or intense cyclogenesis.   These often occur over southern USA in association with:

• a sharp upper trough.
• occasionally with tropical cyclones and hurricanes.

The user should inspect the forecast output for changes in the evolution extending downstream through the forecast period.  This can lead to differences, and hence uncertainty, in the forecast weather as a consequence. 

In all cases it is important to assess whether the observed values are reliable or credible and also the changes the increments have made to the location and sharpness of the upper trough.   The user should study the effect of these changes in the analysis of upper troughs and other features through the subsequent forecast period. 

Users should inspect upper level wind and height increment charts to identify potential sources of significant changes in the downstream evolution (e.g. Fig4.2-2, Fig4.2-3, Fig4.2-6).

Fig4.2-1: To view Analysis Increments:

1. On Charts page, enter Analysis Increments.
2. Click on Analysis Increments diagram.. Display of product appears - increments (differences from the IFS background) for 1000hPa height. 
   
3. As desired select other base times or level -  increments (differences from the IFS background) for 200hPa winds. 

Fig4.2-2:  Rapid growth of uncertainty (in the background forecasts of the Ensemble of Data Assimilations (EDA)) for PV on the surface where potential temperature=315K (shaded as scale).   Also shown are the CTRL forecast PV=2 on 315K (red contour) and 850hPa wind vectors, and ensemble mean precipitation (dots; size indicates rate).   Rapid growth of uncertainty can be associated with cyclogenesis and warm conveyor-belts.  Mesoscale convective systems (e.g. over USA) can also distort the upper flow significantly.  The ENS perturbations may not capture such rapid growth adequately and the upper flow may well become modified more than modelled.  This can cause significant downstream differences at a later time in consequence.  Energetic, fairly large convective systems or strong dynamic upslope motions in warm front conveyors can have an impact on IFS performance.

Fig4.2-3A(left):IFS background 1000hPa heights have been adjusted (purple) in response to observations. The arrows show the vector differences of the analysed 200hPa winds from the background 200hPa winds.  In this case a major distortion of the upper flow was not evident in the background field.

Fig4.2-3A(right): IFS background 200hPa heights have been raised (red) or lowered (blue) in response to observations.  The lines show anomalies of the analysed 200hPa height field from the background 200hPa height field.  In this case a low level background field has been adjusted into a less deep and more relaxed trough.

The increments are defined as analysis minus background.  They show where observations have caused the analysis to depart significantly from the background.  The purple arrows are the vector differences between background and the analysis.  Both charts show large increments associated with a major convective system over the southeastern United States.  These show where observations departed significantly from the background field and the IFS adjusted its analysis to a significant degree in response.

Fig4.2-3B(right): IFS background 200hPa velocity has been adjusted (purple) in response to observations. The arrows show the vector differences of the analysed 200hPa winds from the background 200hPa winds.  In this case a major distortion of the upper flow was not evident in the background field.

Fig4.2-3B(left): IFS background 200hPa heights have been raised (red) or lowered (blue) in response to observations.  The lines show anomalies of the analysed 200hPa height field from the background 200hPa height field.  In this case a low level background field has been adjusted into a less deep and more relaxed trough.

Taken together, Figs4.2-3 show a pattern typical of spring and early summer over the USA, when MCS activity is significant.  Often the IFS model under-represents the associated net upward mass flux (in convective updraughts).  This in turn shows itself as a lack of divergence at upper levels where the updraughts spread out.  The upper level increments then look divergent as a result.  At the same time the upper level height field may not be high enough (due to latent heat released in the updraughts) .  This is commonly indicated as positive (red) upper level height increments.

 Fig4.2-4: Analysis increments show the 200hPa vector differences in (purple) and height (red) between the IFS analysis and the IFS background.  The red areas show where the IFS background height was too low compared with observations.  Consequently 200hPa heights (black lines) have been raised in the region and the trough near and just to the west of the mass of active convective cloud is sharpened.

Figs4.2-5 & Fig4.2-6 show an example where large increments over the mid-West of the USA.  This has induced differences in the forecast upper flow over East Canada two days later and over Europe five days later.

Fig4.2-5: 200hPa wind increments at 00UTC 28 Aug 2019.  Large differences near 90W-95W indicate observations depart significantly from the IFS background.

Fig4.2-6: Forecast 500hPa heights based on 00UTC 27 Aug 2019 (red) and 00UTC 28 Aug 2019 (black).  These compare the evolution of 500hPa heights before and after incorporation of observations over the USA at 00UTC 28 Aug which departed significantly from the IFS background at that time.  The analysis at 00UTC 28 Aug has been adjusted significantly in order to better agree with the observations. The difference in 500hPa height between the analysis at T+0 and that from a previous run at T+24 (both verifying at 00UTC 28 Aug) is highlighted by the yellow/blue "dipole" over the eastern USA.  The subsequent evolution differs from that of the earlier forecast run, first in the handling of the upper ridge over eastern Canada and then in the downstream trough moving over Europe at day5. This is an example in which differences moved and developed downstream, but did not grow substantially. Very occasionally, sequences of this type following large increments, can show substantial non-linear downstream growth of the differences between the previous and current forecasts.

Fig4.2-7: 200hPa wind increments at 00UTC 26 Aug 2019.  The large differences over West Africa indicate that observations depart significantly from the IFS background. The structure of these increments implies that divergence is being "added" to the upper level flow. This is a relatively common occurrence in convective regions.  It can be caused by insufficient upward net mass flux in the convecting area.  This in turn may be because the model's convection is insufficiently vigorous and/or organised.  MCS development commonly relates to this and is known to be problematic for the IFS.

Fig4.2-8: Large 100hPa increments assigned to 12000hPa are incorrect and will not be used.  The observation at 3200M above mean sea level is near 700hPa but the terrain following height levels s of the model will suffer some modification.