You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 5 Next »

Note: The Ensemble Control Forecast is identical to HRES output in Cycle49 and earlier cycles.  They have the same resolution and are scientifically, structurally and computationally identical.  Ensemble Control Forecast output and HRES output are fully equivalent where shown in diagrams.   At the time of some older diagrams, HRES had resolution of 9km and ensemble members had a resolution of 18km.

Types of precipitation - interpretation and effects

The method of  assessment of the type of surface precipitation by the atmospheric model depends primarily upon the temperature structure of the model atmosphere although humidity is relevant too.  Small differences in structure in the region where the precipitation is being generated and in the layers through which the droplets fall can both have a major impact.  A change from rain at the surface to snow can be brought about by cooling of the airmass by evaporation or melting as the precipitation falls.  A major weather hazard is freezing rain which requires a very specific and rare lower tropospheric temperature and humidity structure.  Modelling of all these processes can be difficult, especially if the structure of the model atmosphere is initially imprecisely defined.  Forecasters need to assess the structure of the lower layers of the atmosphere, particularly when warm air is advancing over a very much colder airmass.

A range of charts and diagrams of precipitation types are available on ecCharts. 

Errors in modelling the precipitation type

Users should be aware that errors in modelling the precipitation type can be due to:

  • errors in large scale flow (bringing cooler or warmer air).
  • surfaces too warm or cold.
  • precipitation rate incorrect (greater or less evaporative cooling or cooling through melting, which both change the melting level height).
  • incorrect height of the melting layer.
  • incorrect thickness of the melting layer.
  • incorrect ambient airmass moisture (in dry air the melting temperature of snow is not 0°C but instead is rather greater than this, perhaps by several deg C, due to evaporative cooling).

Deciding between rain and snow in marginal conditions is difficult for current atmospheric models; 1 or 2°C can make all the difference.


Fig9.7-1:  Snowfall in marginal conditions.  An example of the evolution of the atmospheric model temperature structure during an event where the forecast rain turned to snow.  Evaporation of precipitation during descent induced cooling of the model air temperature structure (between T+66 and T+72) and allowed downward penetration of snow and sleet.  Cooling of the air via latent heat absorption during the melting process probably also contributed.  It is likely that sleet reaching the ground has also been treated as an accumulation of snow giving a greater snow depth on the ground than was justified.  In wet snow or sleet conditions there is well-documented over-accumulation on the ground of snow in the IFS. 

Freezing Precipitation

Freezing Rain and Freezing Drizzle are super-cooled droplets that freeze upon impact with a surface that has a sub-zero temperature.  Accumulation of ice is also known as glaze or glazed ice and can be rapid and extensive.  Glazed ice on surfaces is a particular, and sometimes extreme, hazard to pedestrians and to road and rail transport, and to aircraft near and on the ground.  Similarly, accumulations of ice can occur on other sub-freezing or near freezing surfaces, such as foliage or power lines.  The weight of accumulated ice can be sufficient to fell trees, bring down power lines, cause pylons to collapse and in extreme cases possibly cause other structural failure through weight of ice alone.

A surge of warm air overrunning rather than replacing a pre-existing area of stagnated cold air (e.g. by differential advection in the vertical, or where warm air from the south overruns stationary very cold low-lying air) can deliver the special conditions needed for freezing rain generation.  Users should also consider the likely local temperature structure in mountainous areas where sub-zero layers may be trapped in valleys while not in evidence over adjacent more open areas.

A schematic cross-section on a transect through a precipitating warm front, based on a real case in Eastern Europe, is shown in Fig9.7-2 and Fig9.7-3.

Currently IFS output will only signify freezing rain when the vertical atmospheric structure is as shown in Fig 9.7-2.  Note, however, in conditions of shallow (generally non-frontal) cloud that is supercooled but not glaciated, drizzle or light rain can fall and also take on the characteristics of freezing rain or drizzle.  This is by virtue of the fact that the temperatures of the liquid droplets and the air through which they fall are both below zero.  The key difference relative to the depiction in Fig9.7-3 is that there is no elevated warm layer, and no snow higher up - the airmass temperature will be below zero from the surface up to cloud top at least.  This scenario is often know as 'freezing drizzle', though sometimes light (freezing) rain may arise in the same way.



Fig9.7-2:  Chart of the north Adriatic and adjacent countries showing assignment of precipitation type represented by colours: Green-Rain, Blue-Snow, Yellow-Ice Pellets, Red-Freezing Rain, Pink-Sleet, Turquoise-Wet snow.  Shading of each colour denotes intensity - darker for more intense.


Fig9.7-3: Schematic cross-section north to south along the black line in the chart Fig 9.7-2 (in many cases the ice pellet zone will be much narrower in the horizontal direction than shown here).  The section intersects a warm front zone with an elevated layer where temperatures are above 0ºC.  Precipitation is assigned to each precipitation type according to the structure of the model atmosphere.

 

Accretion of glaze or glazed ice on surfaces

It is important to appreciate that diagnosis of freezing rain in IFS takes no account of the temperature of the surface, only on the temperature structure of the boundary layers.  An IFS forecast of freezing precipitation indicates the likely presence of super-cooled droplets in the lower atmosphere - not the effect of their impact on a surface.  It should be remembered that an indication of freezing precipitation gives no information of likely accumulation of glazed ice although there must be a serious risk, and heavier precipitation rates suggest a potential  for greater accumulation.  The user should consider the kind of surface in question and its probable temperature and any heat fluxes (e.g.ground heat) before assessing the result of any indicated freezing precipitation.  

Physically, accretion of ice on power lines and road ice accretion are different. 

Power and telephone lines

Overhead wires and cables are likely to fairly readily acquire the sub-zero temperature of the surrounding air and thus are likely to be prone to glaze or glazed ice accretion.  IFS products for freezing precipitation should provide good guidance.

Road surfaces etc

Despite sub-zero air temperature, road surface temperatures may or may not have sub-zero temperatures due to warming, albeit slight, from traffic or upward heat flux from the ground.  Heat from buildings and underground constructions in cities may also raise road surface temperatures.   In these cases, ice accretion on road surfaces may be patchy (arguably an even greater hazard).  IFS products for freezing precipitation should mostly provide good guidance, but the user should consider the likely temperature of road surfaces etc before deciding whether glazed ice is likely to be extensive or not.

Conversely, even when the air temperature is above zero and IFS indicates rain (not freezing precipitation), roads can be sufficiently cold, for rain to freeze on contact. Usually such conditions do not last long, because the surface temperatures quickly rises to the ambient air temperature, but whilst it lasts this situation can be very dangerous.


Model processes for precipitation types falling onto different surfaces

Precipitation reaching the surface accumulates or is removed by various modelled processes.  It is important to note that ECMWF's precipitation type categorisation is in part a diagnostic quantity that does not fully reflect how precipitation types are stored in the model.  In particular forecast precipitation is only considered to be liquid (rain), solid (snow, frozen rain, wet snow), or a mixture of these (rain and snow mixed/sleet/ice pellets).  Hail is not forecast.  Freezing precipitation is diagnosed as described above.  The accumulation of precipitation at the ground is governed by the characteristics of the surface onto which the precipitation falls.  Land surface tiles have varying characteristics, and include vegetation which intercepts both snow and rain.  The related behaviours, both in reality and in the IFS, are complex, and the IFS representation is imperfect.  The points below summarise most of the main features of precipitation - surface interaction, although the list is by no means exhaustive:

Rain falling to the ground surface:

  • is considered to runoff or infiltrate the ground. after falling it can also evaporate from the surface. Where there is vegetation small amounts can also be intercepted by and stored on leaves.
  • is generally considered to accumulate within the soil, but that process will exclude any proportions that are intercepted, that run off, or that evaporate.

Rain falling onto snow:

  • is assumed to be at 0°C as it reaches the snowpack.
    • the rain interception is associated with an increase of snow density due to the refreezing process, and also a warming of the snowpack due to the latent heat release by the refreezing. This may or may not increase the snow depth, depending on the energy balance.
      However:
      • there is a maximum amount of liquid water that can be intercepted by the snowpack (about 3-6% of the total snow mass depending on the snow density).
      • when the maximum amount of liquid water is reached the remaining water goes into the ground or runs off.
      • currently too much rain re-freezes

Snow falling to the ground surface:

  • is considered to accumulate where the surface temperature is below 0°C and is added the previous snow lying on that surface. the full snowpack is represented with one (evolving) density, as currently there is only surface snow layer.
    • predicted snow depth will be in error whenever snow of one density falls onto snow that has a different density; if the density difference is large errors in forecast depth can be substantial.
  • is considered to immediately melt where the surface is above 0°C, although associated latent heat release will generally cool the surface leading soon to accumulation.
    • in practice, melting is often underdone, which commonly results in too much snow on the ground in 'wet snow' situations.
  • when convective component dominates, the model tends to produce high precipitation rates with a significant snow fraction event when surface temperatures are well above freezing;
  • the IFS tends to produce rather too active convection with too much wintry precipitation or snow (e.g. a mixture of snow and rain (sleet) at the surface while the 2-metre temperature is 8.2°C).  These showers sometimes leave snow on the ground (mainly <1 cm) which is then slow to thaw. This can lead to some false alarms, especially if users are not aware of this tendency.

  • dry snow is often too compact in IFS output.  Forecast depths can be underdone as a result even if the liquid water equivalent is right.

Freezing rain falling onto snow on the ground:

  • will in general immediately freeze.
  • will not necessarily increase snow depth.
  • may increase snow density.

Freezing rain falling onto bare ground:

  • will add to soil moisture (unless the soil water is frozen).
  • will not add to snow accumulation (ice accumulation is not modelled).

Ice Pellets and Sleet (rain and snow mixed) falling onto snow on the ground:

  • will increase snow depth according to the proportion of snow in the mix - acting as snow falling onto snow as given above.
  • will increase snow density according to the rain in the mix - acting as rain falling onto snow as given above.
  • commonly snow depth/mass on the ground increases too much when ice pellets and/or sleet are falling (i.e. melting of the evolving snowpack is underdone)

Snow forecast information

Users should be aware of possible limitations of model snow forecasts.  

  • High impact considerations

    • Snow cover analyses rely on snow reports, background forecasts and satellite-derived snow-cover products.  Any or all could be in error.
    • Incorrect analyses and forecasts of snow are quite possible at altitudes above 1500m, or in data sparse areas, or after a prolonged period without observations.
    • Incorrect reports of shallow snow (say by thick deposition of frost) may be assimilated.  This can be persistent in the model and give misleading forecasts.

    • IFS tends to melt snow away too slowly.  Snow cover and associated colder surface temperatures may persist for longer than they should.  This could influence other parameters.  

    • The direction and strength of the low level winds can have a strong effect on snowfall:

      • Surface wind from land - temperatures can be lower and snowfall deeper.

      • Surface wind from sea – temperatures slightly higher and snow more sleety, at least at lower levels.

    • IFS snowfall can be insufficient (by factor of ~2) when IFS dewpoint depression is >~4C.  This is due to under-evaporation of falling particles in IFS.
    • IFS snowfall often accumulates incorrectly on the ground when IFS develops rain and snow.  
    • Snow may accumulate then melt (e.g. with rain, or as as a warm front advances over a cold area).
    • The extent and thickness of cloud or freezing fog has a strong influence energy fluxes into and from the snowpack.  Consider possible cloud formation, persistence or clearance and assess possible changes in energy transfer between cloud and snowpack.  Thick cloud at any level will reduce solar radiation.  But low cloud could be warmer than the underlying snow surface resulting in a net increase in downwards long wave radiation. 
    • The characteristics of each grid box and areal extent of each tile type are updated through the forecast period and can vary in a rapid and interactive way.  Tiles in HTESSEL may be incorrectly assigned.
      • Model forecast snowfall might increase the area or depth of snow cover incorrectly.  Partial cover of snow may incorrectly become full cover as the accumulated model snow depth becomes >10cm.
      • The statistical information on the slope and aspect of orography within each grid box is not detailed enough for forecasts at an individual location.  This can be important in mountainous areas and HTESSEL may under- or over-estimate solar heating and runoff.
      • A valley in a mountainous area may allow low-level moisture into the sheltered side.  This can enhance model precipitation rates there instead of evaporating snow in dry air (in a fohn area).  
    • Differing snowfall among the ENS members can cause increasing differences in evolution during the remainder of the model forecast period.  Nevertheless each member remains equally probable.


  • Snow temperature considerations 
    • Surface reflectance (slow albedo) can influence surface heat flux and skin temperatures (by1°C-4°C).
      • Fresh (white) snow has high albedo reflecting much of the incoming radiation.
      • Dirty or older (greyer) snow absorbs more radiation with greater heat flux into the snowpack.
    • Solar elevation and radiation is limited at high latitudes (and in winter non-existent).  
    • Thermal properties of the snow can cause heat and moisture fluxes to be effectively decoupled.  Snow, especially new dry snow, is a good thermal insulator.
    • Over deep snow, forecast T2m temperatures compared with observations:
      • between −15°C < 0°C forecast T2m has good agreement with observations, but with a relatively constant cold bias during the day of ~1.5°C.
      • below -15°C forecast T2m tends to be too warm.
    • Large night-time errors of forecast T2m are possible, even by as much as 10°C too warm.  These are more likely under observed clear skies, even when clear skies correctly simulated by the model.
    • Water phase changes can cause a delay in warming or sublimation of snow.  
    • In IFS, falling (airborne) snow tends to sublimate much more readily than the undisturbed snow on the ground.  This can reduce snowfall.
    • In IFS, shallow surface snow often takes too long to melt, even if ground surface temperatures are above 0°C.  This can have an adverse impact on albedo and radiation fluxes.

    • In IFS, snow surfaces are likely to melt a little more readily in forests.  The heat flux at the snow/atmosphere interface is rather larger than with exposed snow.

    • In IFS, snow depths may reduce gradually (and also in reality) because the density of the snow has increased through compaction as the days progress.

    • In IFS, forest snow night time model temperatures fall too low.  T2m is evaluated as if exposed snow "tile", even if the forest is dominant.  This is because verifying SYNOP stations are always in a clearing.  In reality, forest generated turbulence maintains turbulent exchange over the clearing and prevents extreme cooling.

    • IFS forecast 2m temperatures over deep snow:
      • have good agreement with observations between −15°C and 0°C.
      • tend to be too warm by around 3-5°C compared to observations when T2m <-15°C.   Large night time errors of forecast temperatures, even by as much as 10°C too warm, are more likely under clear skies.  This can occur even when clear skies have been correctly simulated by the model. 
      • have a relatively constant cold bias during the day of ~1.5°C compared to observations.
      • the amplitude of the forecast T2m diurnal cycle underestimates the amplitude of the observed diurnal cycle by between ~10% to 30%:  Forecasts of:
        • night-time T2m minima tend to be warmer.
        • daytime maxima colder than observations. 
    • T2m verification can be difficult due to variations in the height that temperature observations are made.  Some countries and locations:
      • maintain the sensors 2m above the snow surface, adjusted after every fall of snow.
      • have sensors higher than 2m above the ground to ensure measurement of air temperature throughout the year even after large accumulations of snow.   Deep late winter snow reduces the distance between snow surface and sensor.  In warmer periods of the year the sensor will be further from the ground surface than normal observations.  See Fig2.1.4.4-6. 


  • Snow depth and coverage considerations

    • Smooth snow surfaces can cause momentum fluxes to be decoupled and winds increase in the absence of friction.  
    • Strong winds can alter snow depth and snow compaction.  Strong winds, especially if prolonged:
      • can bring areas of drifting with snow compaction and associated increase in density.  This can be particularly effective with polar snow with extremely low temperatures throughout the winter and compaction due to other processes is limited.
      • can carry away dry surface snow and reduce snow depth in exposed areas. 
    • Bias in snow depths:
      • Short-range snow depth forecasts on average show high quality compared with independent observations.
      • There is a slight overestimation of snow depth in the background and analysis fields.
      • There is a tendency towards underestimation of snow depth in central Eurasia.  This implies over-estimation of either melting or compaction for these forested areas.

    • Snow depths are often underestimated due to the very conservative snow density relation between model precipitation and snow accumulation, even though the precipitation is predicted pretty well.  The density of snow settling on the ground increases very rapidly as wind speed increases.  Very roughly:

        • Little or no wind - 2 cm snow for 1 mm rainfall equivalent.

        • Strong wind – 0.6 cm for 1 mm rainfall equivalent.

  • Ice

    • Thin sea-ice or lake-ice covered by thin snow grows or melts much faster than does thick ice with deep snow.

Additional sources of information

(Note: In older material there may be references to issues that have subsequently been addressed)


(FUG associated with Cy50r1)



  • No labels