Explanation

The following explains the heatforecast meteogram in more detail

Uncertainty

To measure uncertainty, the heat forecast is not based on a single forecast, but on ECMWF’s ensemble prediction system, which runs 50 different forecasts covering the range of possible weather over the next 10 days. If 10 ensemble members simulate a storm but 40 do not, then there’s a 20% chance of this storm to occur. For every variable like temperature or wind, we calculate the minimum, 10th, 25th, 50th (median), 75th, 90th percentile and the maximum across the distribution of the 50 ensemble members. Wider distributions with a larger difference between minimum and maximum or between th 10th and 90th percentile (the decile range) are less certain. For narrow distributions, more typical in the first days of a forecast, all 50 simulations basically agree what is going to happen, providing high certainty.

Spatial resolution

The heat forecast meteogram is based on a global resolution of about 9km. We chose the location 48.82°N and 2.29°E, just south of the périphérique close to the Vanves-Malakoff train station, of the global grid to be representative for Paris. But local differences remain especially in urban areas, which we cannot resolve.

Temporal resolution

The meteogram shows hourly data for the first three days then decreasing to 3-hourly data before further decreasing to 6-hourly data on day 7 of the forecast. For visualisation purposes we use a spline interpolation to avoid a counter-intuitive and physically unreasonable linear interpolation between long time steps. As a result, the feels-like temperatures are very sinusoidal on day 7 to 10 of the forecast.

Clouds

In the top panel cloud cover is visualised. Clouds are classified as low (dark grey), medium (grey), and high (light grey), depending on their height in the atmosphere. Higher cloud covers (measured in fraction of sky) are visualised with thicker clouds here. Yellow suns (why yellow?) are added on every day at noon for orientation. We use the median of the forecast ensemble and do not visualise uncertainty.

Precipitation

Precipitation (or rain) is visualised in a worst-case (meaning higher precipitation) and best-case (lower precipitation) category. The worst-case is the 90th percentile of precipitation in the ensemble forecast meaning the amount of precipitation in the highest 5 out of 50 forecasts. The best-case is the 25th percentile. For every hour the amount of precipitation is visualised by more transparent (=less rain) or less transparent (=more rain) rain drops. Their vertical position has no meaning and is chosen for aesthetic purposes only.

Feels-like temperature

The feels-like temperature is quantified as the Universal Thermal Climate Index (UTCI) which incorporates the effects of windchill (lower experienced temperatures due to higher winds), humidity (higher humidity reduces the body’s ability to lose heat through sweating), and radiation (sunshine during the day and longwave outgoing radiation into space). The feels-like temperature is calculated from these environmental variables (air temperature, wind speed, humidity, mean radiant temperature) and visualised as the daily ups and downs over the course of 10 days. The median across 50 forecasts is the grey line in the centre, the colour shading shows the mininum, 10th, 25th, 75th, 90th percentile and maximum across the ensemble. The colour of the shading is given by its temperature so that 25˚C always gets the same orange and that orange is always associated with 25˚C. The uncertainty visualised by the width of the colour shading is usually much lower in the first days of the forecast, meaning the forecast is more certain what wheather is going to take place, but typically increasing further into the future due to the chaotic nature of the atmosphere.

Wind

Stronger winds are visualised with windsocks at the bottom of the feels-like temperature panel. The wind socks can be angled at 45˚, 67.5˚ or be completely horizontal (90˚) denoting the strength of the expected wind (median), corresponding to 4, 5, and 6 and higher on the Beaufort scale. These are classified as “moderate breeze” (4 Beaufort), “fresh breeze” (5 Beaufort) and “strong breeze” (6 Beaufort) all the way to “Hurricane-force” (12 Beaufort). The transparency of the wind sock is (inversely) proportional to chance of wind speeds exceeding 4 Beaufort. If the 25th percentile of the ensemble distribution exceeds 4 Beaufort then the alpha value will be 0.75 meaning a 25% transparent wind sock as most (75%) of the forecasts agree that winds will be at least that strong. Weaker winds or the wind direction is not visualised.

Contribution to feels-like temperature: Humidity

Human bodies can regulate their heat budget through sweating. However, the efficiency to lose energy by sweating and subsequent evaporation (which cools) of that moisture depends on the surrounding air to be able to take up that additional humidity. If the air is humid, sweating not as efficient to cool down the human body. Consequently, higher humidity levels typically increase the feels-like temperature compared to the air temperature by a few ˚C.

Contribution to feels-like temperature: Windchill

Higher winds transport heat more efficiently away from the human body, consequently reducing the feels-like temperature in comparison to the air temperature. This is visualised in dark blue colour shading with negative temperatures in the bottom panel. If the colour shading reaches down to -10˚C then it would reduce an air temperature of 25˚C to feel like 15˚C. The transparency in the colour shading denotes its uncertainty, visualising minimum, 10th, 25th, 50th, 75th, 90th percentile, and the maximum of the ensemble distribution. If the darkest blue reaches down to -5˚C the windchill effect will be at least that, if the lighter blue shadings reach down to -10˚C it is also possible but not very certain for winds to be that strong to cause -10˚C lower feels-like temperatures due to windchill. The windchill effect is always negative.

Contribution to feels-like temperature: Radiation aka sunshine

Sunny days (but also lesser degree cloudy days) feel warmer than the actual air temperature as shortwave solar radiation transports energy onto your skin, warming you up. (We avoid the term “radiation” in the meteogram so that no on thinks this is about a nuclear power plant in their vicinity, this is just about solar and terrestrial radiation due to their respective temperatures). Daytime feels-like temperatures in the sun can therefore often be 5-10˚C warmer than the actual air temperatures, turning a freezing but very sunny day into an enjoyable outdoor experience. Lower and thicker clouds can considerable reduce that effect. At night, a lot of longwave terrestrial radiation (and radiation from the temperature of a human body) escapes into space cooling you down compared to being indoors with a ceiling that would prevent that and radiate back. Hence the “sunshine” contribution to the feels-like temperature is often negative by 1-2˚C, depending on night-time cloud cover (which acts as a “ceiling”). Uncertainty is visualised here in the same way as for the windchill effect described above. The colour shading for radiation is always orange to distinguish it from the windchill cooling.