When we look at a real-time rainfall map, much of what we see comes from a technology that transformed meteorology in the second half of the twentieth century: weather radar. In Spain, the State Meteorological Agency (AEMET) runs one of the most complete radar networks in Europe, central to detecting rainfall and issuing warnings. This article explains how the technology works, from the basic physics to its operational uses.
The physics: pulse, backscatter and reflectivity
RADAR is an acronym for Radio Detection And Ranging. A weather radar emits very short pulses of electromagnetic radiation in the microwave band. When those pulses meet obstacles in the atmosphere — raindrops, snowflakes, hail, even insects — part of the energy scatters in all directions. The fraction returning towards the radar is called backscatter.
The radar measures two fundamental things about that returning signal: its intensity, meaning how much energy comes back, and its travel time, meaning how long it takes to go and return. Time gives the distance to the target; intensity relates to the quantity and size of the particles doing the scattering.
The fundamental quantity a radar measures is reflectivity (Z), expressed in decibels of reflectivity (dBZ). Technically, Z is proportional to the sum of the sixth power of the diameters of all the drops within a sample volume. That means large drops dominate: a 4 mm drop produces 64 times the reflectivity of a 2 mm one.
From reflectivity to rainfall: Z-R relations
What hydrologists and forecasters actually want is not reflectivity but the rainfall rate (R) in millimetres per hour. Converting Z to R relies on empirical formulas known as Z-R relations, the best known being the Marshall-Palmer equation of 1948, in which Z equals 200 times R raised to the power of 1.6.
That relation assumes a specific drop-size distribution and is not universal. Snow, hail and convective rainfall each need different Z-R relations with different coefficients. Choosing the right one is a major source of uncertainty in quantitative precipitation estimation from radar.
The AEMET radar network
AEMET operates a network of 15 C-band radars, with a wavelength of roughly 5 cm at 5.6 GHz, spread across the Iberian Peninsula and the islands. C-band is a compromise between resolution, which is better at shorter wavelengths such as X-band, and range and penetration through heavy rain, which are better at longer wavelengths such as S-band. The radars generally sit on high ground to maximise coverage, with a typical range of 240 km for reflectivity and 120 km for Doppler data.
Each radar completes a full volume scan roughly every 10 minutes. During that cycle the antenna rotates 360° at a series of elevations, typically between 0.5° and 25° above the horizon, building a three-dimensional picture of the precipitation within range.
Doppler capability: measuring the wind
AEMET’s modern radars include Doppler capability, which exploits the Doppler effect to measure the velocity of precipitation particles. When raindrops move towards the radar the frequency of the returned signal rises slightly; when they move away it falls. By measuring that frequency shift, the radar calculates the radial velocity of the drops — the component of their motion along the beam.
Doppler information is extraordinarily useful for:
- Detecting mesocyclones: rotation patterns within storms that can indicate supercells and tornado risk.
- Estimating wind profiles: reconstructing the wind field at different heights, complementing atmospheric soundings.
- Detecting convergence: areas where low-level winds converge tend to force ascent and new convection.
- Filtering fixed echoes: zero Doppler velocity identifies returns from mountains, buildings and other stationary objects so they can be removed.
Dual polarisation: classifying hydrometeors
The most significant upgrade to AEMET’s radars in recent years has been dual polarisation. A conventional radar transmits horizontally polarised pulses; a dual-polarisation radar transmits horizontally and vertically at the same time.
Large raindrops are not spherical: air resistance flattens them into an oblate shape, wider than they are tall, so they reflect more energy horizontally than vertically. Snowflakes, ice needles and hailstones have different shapes and produce different polarimetric signatures. The main parameters dual polarisation provides are:
- Differential reflectivity (ZDR): the ratio between horizontal and vertical reflectivity. High values indicate large, flattened drops and heavy rain; values near zero indicate spherical particles such as drizzle or hail.
- Specific differential phase (KDP): how the phase of the signal changes as it passes through precipitation. It is very useful for estimating rainfall intensity and is robust against calibration errors.
- Correlation coefficient (ρHV): how similar the horizontal and vertical signals are. Values close to 1 indicate uniform particles such as pure rain; lower values indicate a mixture, such as wet hail or the melting layer.
Composites and operational products
No single radar covers the whole country. To build a complete picture, AEMET generates national radar composites by merging data from all 15 radars. The process involves:
- Time synchronisation. The scan cycles of each radar are not simultaneous, so data is interpolated to a common reference instant.
- Choosing the best elevation. Where coverage overlaps, the system picks the return most representative of rainfall at ground level, avoiding beams that are too high or blocked.
- Mosaicking and projection. Data originally in polar coordinates is projected onto a regular cartesian grid for display and for use in models.
The best-known product is the PPI, or Plan Position Indicator: the familiar coloured image over a map showing rainfall intensity seen from above. Others include CAPPI, a horizontal slice at constant altitude; Max Z, the maximum reflectivity in the vertical column, useful for spotting hail; and VIL, the vertically integrated liquid water content of the column.
What radar cannot do
For all its usefulness, radar has important limitations that must be borne in mind when reading its output.
Terrain blocking. Mountains intercept the beam, creating shadow zones where precipitation cannot be observed. In a country as mountainous as Spain this matters, especially in the Pyrenees, the Cantabrian range and the Bétic sierras.
Ground clutter. Returns from terrain, buildings or wind turbines can be mistaken for rain. Doppler filters remove much of it, but anomalous propagation caused by temperature inversions can bend the beam towards the ground and generate extensive false echoes.
Attenuation in heavy rain. At C-band, very intense rain absorbs part of the radar energy, so precipitation behind a strong convective core appears weakened or invisible — a problem precisely in the most dangerous situations.
Distance and the curvature of the Earth. As the beam travels outwards it rises, because the Earth curves away beneath it. At 200 km even the lowest elevation is sampling several kilometres up, far above where rain actually reaches the ground. What falls aloft does not always reach the surface, and vice versa.
Radar compared with rain gauges
Rain gauges measure directly and accurately how much water falls at a point. Their limitation is that they are point measurements within a spatially variable rainfall field. Radar, conversely, offers continuous spatial coverage but carries the uncertainty of the Z-R conversion.
The best solution is to combine both: gauges calibrate and correct the radar, and the radar interpolates between stations. This approach, known as radar-gauge merging, is what AEMET and the SAIH systems of the river basin authorities use to obtain the most accurate rainfall fields possible.
Satellite products
Complementing radar, geostationary weather satellites such as Meteosat Second Generation provide continuous images of the atmosphere. The SEVIRI instrument observes in 12 spectral channels every 15 minutes, or every 5 minutes in rapid scan mode, allowing rainfall to be estimated from cloud-top temperatures in the thermal infrared.
Satellite rainfall estimates are less accurate than ground radar, but they cover areas with no radar coverage, such as the open sea, and give a synoptic view of the whole weather system.
How WhatAWeather uses radar
WhatAWeather integrates radar data through RainViewer, a service that gathers and processes information from radar networks worldwide to deliver real-time rainfall imagery refreshed every few minutes. The radar tiles overlay the interactive map, letting you see where rain is falling and how heavily.
That view is combined with weather station data, river levels, reservoir status and numerical model forecasts to give a joined-up picture of the situation in real time.
Nowcasting versus forecasting
Radar is the primary tool of nowcasting: very short-range prediction, from zero to six hours. By extrapolation, the movement of observed rainfall structures is projected into the immediate future. For the next one to two hours this approach is generally more accurate than numerical weather models.
Extrapolation assumes, however, that rainfall keeps its shape and intensity, which is not always true: storms grow, weaken, merge and split. Beyond two or three hours, numerical models such as AEMET’s HARMONIE-AROME or the ECMWF’s IFS outperform extrapolation, because they simulate the physics and can anticipate rainfall that has not formed yet.
The most advanced nowcasting techniques blend the two, weighting radar extrapolation heavily in the first hours and transitioning gradually to the numerical model for longer horizons. Machine learning methods based on convolutional neural networks are also showing promising results, learning storm evolution patterns directly from historical radar data.
Weather radar, with all its strengths and limitations, remains an irreplaceable link in the chain of monitoring and warning for heavy rainfall. Its continued evolution — dual polarisation, higher resolution, integration with satellites and models — keeps it at the centre of flood protection systems.