Page tree
Skip to end of metadata
Go to start of metadata

Lightning

A new "lightning flash density" forecast product was introduced in Cycle 45r1, released in June 2018. This can be viewed within ecCharts. The units of lightning flash density as archived are "strikes/km2/day", but these are rescaled for ecCharts usage to "average strikes/km2/h". The product is derived from HRES and ENS, with ecCharts ENS representation being probabilistic (to exceed a certain density threshold). The threshold and period duration for the probabilities are under user control.

The diagnostics aim to represent cloud-to-ground plus intra-cloud lightning strikes. Note that many ground-based lightning sensing systems are much more adept at identifying cloud-to-ground strikes.

Within echCharts the lighting variables can be accessed via the Layers dropdown menu in the usual way.


Fig8.1.11.1A(left): ecCharts display of lightning flash density (average strikes/100km2/h average during the previous 6h) from HRES.

Fig8.1.11.1B(centre): ecCharts display of probability of lightning flash density (here >2 strikes/100km2/h average during the previous 6h).

Fig8.1.11.1C(right): ecCharts display of probability of lightning flash density (here >0.1 strikes/100km2/h average during the previous 6h).

In the example just above values for Bordeaux (location shown by the pin) are given in the probe display at the top where less than 1 strike per hour average during the last 6 hours is forecast within the surrounding 100 square km by HRES.  However, ENS gives probabilities for the same period of 45% for more than 2 strikes per hour average, and 80% for more than 0.1 strikes per hour average, during the same period.

Considerations when using Lightning charts

The parametrization of total lightning flash density uses the following quantities diagnosed from the ECMWF convection scheme:

  • convective available potential energy (CAPE),
  • convective flux of frozen precipitation,
  • cloud condensate amount within the convective updraught,
  • convective cloud-base height.

For the "flux of frozen precipitation" assumptions are used to infer the proportion of graupel, which thereby becomes a function of the underlying surface: land or sea.This affects the lightning density diagnostic. As a higher proportion of graupel is assumed over land, more lightning tends to be diagnosed over land (all other things being equal).

In the end the following factors will increase the prognosed lightning density:

  • Higher CAPE
  • More condensate in the convective updraught (between levels at 0C and -25C)
    • More snow in the said convective updraught
    • More graupel in the said convective updraught
  • A higher convective cloud base (if less than 1.8km)

Lightning flash density requires careful interpretation.  Lightning activity should not be considered as being precise in time and space but rather indicative of general activity within a region.  The values presented in ecCharts:

  • are displayed as units per 100km2 (i.e. per 10km x 10km square, or ~0.1ox0.1o) surrounding the relevant grid point.  So HRES does not indicate lightning strikes precisely at a given location. Recall also that HRES gridboxes measure about 9km by 9km, so greater detail than in raw model output is not possible. Note also that strikes more than 10km away may be visible (or audible) to an observer.
  • verify better when considered over larger areas and/or over longer time scales.
  • often show peak activity about an hour earlier than observed.
  • often show activity decaying away too early in the afternoon while in reality thunderstorms continue through the afternoon and often linger well into the night.
  • often show lightning activity to be too intense, particularly during periods of higher activity.
  • when considered over several years, tend to underestimate activity when compared with a 10 year climatology of satellite data, particularly over central Africa but also over parts of eastern Europe and central Asia.  Conversely, it is possible there is too much activity forecast over parts of the tropical Pacific.
  • can exhibit systematic errors in the European area during the autumn in particular; at times there will be much more activity over the Mediterranean (sea area) than predicted (this may relate to the assumptions about graupel distributions over land and sea).

The six figures below relate to the above bullets.

The probability of lightning activity is a good indicator of the risk of lightning strikes at a given location.  Selection of the threshold for lightning strike density allows the user to assess the risk at a location or within an area.  Selecting a low threshold (e.g. 0.1 strikes/100km2/h during the previous 3h) gives an indication of whether any lightning activity can be expected at all.  For many customers this is the aspect that is of primary interest.


Fig8.1.11.2: Mean correlation, for lightning flash density over Europe in summer 2015, between IFS short range forecasts and ground based observations, as a function of averaging spatial resolution (x-axis), and for different averaging periods (different lines).  Model forecasts of lightning activity should not be considered as being precise in time and space but rather as indicators of activity within a region, with there being greater confidence for larger areas over longer time scales.  Note lightning flash density is displayed on ecCharts as units per 100km2 (equates to spatial resolution ~ 0.1o).


Fig8.1.11.3: Mean normalised diurnal cycle of lightning activity from IFS short-range forecasts (black line) against three ground-based observation networks of lightning sensors (coloured) over Europe in summer 2015.  Model forecasts of lightning activity decay away too early in the afternoon while in reality thunderstorms continue through the afternoon and often linger well into the night.


Fig8.1.11.4: a) Observed mean lightning flash densities over 6hr and b) probability (>30%) of lightning flash density exceeding 0.5 flashes 100km-2 hr-1 derived from ENS.  The charts illustrate that on the broad scale the forecasts capture the areas at risk, but the location of the activity is rather imprecise.


Fig8.1.11.5: Time series of daily mean lightning flash densities from IFS short-range forecasts (blue) and from ground-based observation network of lightning sensors (red) over Europe during summer 2015.  The IFS forecast lightning activity tends to be too intense, particularly during periods of higher activity.


Fig8.1.11.6: Annual mean lightning flash densities from a) satellite climatology and b)10 year-long IFS model runs.  IFS  tends to underestimate activity, more particularly over central Africa but also evident over parts of eastern Europe and central Asia.  Conversely, it is possible there is too much activity forecast over parts of the tropical Pacific.



Fig8.1.11.7: Seasonal mean lightning flash density for months SON from 15 years of observations (left) and from 10 year-long 80km IFS simulations (right). Scale is: flashes/km2/year. Agreement for other seasons (not shown) is much better.


Additional Sources of Information

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

Read more on the promising results for lightning prediction.

Read more on a lightning parameterization scheme for the ECMWF Integrated Forecasting System.

  • No labels