Dr Philip Rosenberg

Apply Now
Instrument Support Level 2
Instrument Support Level

Instrument Support Level 2

Manufacturer & Model

RPG Radiometer Physics Gmbh, HATPRO Radiometer, G5

CEDA Archive Name


Data Products

boundary-layer-temperature-profiles, brightness-temperature, boundary-layer-height full-troposphere-temperature-profiles, iwv-lwp, moisture-profile, stability-indices, surface-met

Insurance Value


Physical Dimensions

105cm x 40cm x 100cm, 70kg


See page for details

Daily Facility Charge



Calendar 2:
HATPRO Radiometer

HATPRO Radiometer

The HATPRO instrument is a stand-alone system for automated weather-station use under nearly all environmental conditions. Full atmospheric profiles are derived, retrieved data, as well as raw data, are stored. A variety of retrieval algorithms (custom designed or global standard algorithms) can be selected.

The system is passive with two frequency reception bands: 22-31 GHz (7 channel filter bank humidity profiler and LWP radiometer) and 51-58 GHz (7 channel filter bank temperature profiler). A range of data are retrieved including:

  • Vertical profiles of atmospheric temperature
  • Vertical profile of atmospheric humidity (relative and absolute humidity)
  • Liquid Water Path (LWP)
  • Integrated Water Vapour (IWV)
  • Stability indices
  • Surface pressure, temperature, relative humidity, and rain flag


The system is capable of standalone operation but we supply a host PC to facilitate data archiving and custom setup of the instrument. The HATPRO GPS provides the measurement timing and also position. The environmental temperature range for operation: – 60°C to + 45°C and the automatic rain-mitigation system, consisting of a hydrophobic coating and strong blower, prevent rain settling on the radome. The 1.8kW heater module prevents the formation of dew and fog condensation on the radome. A new radome is fitted before any deployment and replaced, while on deployment, every 6 months or if compromised. (Mechanical damage, salt build-up, etc all result in the hydrophobic properties of the radome being compromised).

Optical Resolution: 3.5° (2.5°) HPBW at 22 (51) GHz
Radiometric resolution: 0.3 – 0.4 K RMS at 1.0 s integration time
Absolute system stability: 1.0 K
Receiver and antenna thermal stabilisation: < 0.02 K
Repetition rate: filter bank receiver produces one atmospheric profile per second
Humidity profile performance:

Vertical resolution
200 m (range 0-2000 m)
400 m (range 2000-5000 m)
800 m (range 5000-10000 m)

0.4 g/m3 RMS (absolute hum.)
5% RMS (rel. humidity)

Temperature profile performance:

Vertical resolution
BL-Mode 30 m -50 m (range 0-1200 m)
Z-Mode 200 m (range 1200-5000 m) / 400 m (range 5000-10000 m)

0.25 K RMS (range 0-500 m)
0.50 K RMS (range 500-1200 m)
0.75 K RMS (range 1200-4000 m)
1.00 K RMS (range 4000-10000 m)

Accuracy: +/- 20 g/m2
Noise: 2 g/m2 RMS

Accuracy: +/-0.2 kg/m2 RMS
Noise: 0.05 kg/m2 RMS

Atmospheric profiles of temperature, humidity, wind direction and speed are typically measured by radiosondes. Their operation is expensive and requires extended logistics, and hence results in a poor spatial (several hundred kilometres at best) and temporal (about twice a day) coverage. Remote sensing of temperature and humidity profiles from satellites yields better spatial coverage especially over oceans and sparsely populated land areas, however, the obtained horizontal and temporal resolution is coarse. Due to their viewing geometry, the vertical resolution is good in the upper troposphere but deteriorates towards the surface. Because clouds strongly absorb in the infrared spectral region several satellite instruments (e.g. the Advanced Microwave Sounding Unit AMSU and the Special Sensor Microwave/Temperature SSM/T sounder) operate in the microwave region where clouds are semi-transparent.Profiling is achieved by measuring the atmospheric emission along the wings of pressure broadened rotational lines. The 60 GHz oxygen absorption complex is typically used for temperature profiling while the 183 GHz water vapour line is used for the humidity profile. Because the atmospheric opacity is high for both bands, the problem of the unknown surface emission is eliminated.

The usefulness of ground-based microwave radiometry for the retrieval of temperature and humidity profiles has been proven for quite some time. Due to the low maintenance requirements of microwave radiometers, continuous atmospheric profiles can be measured which have the highest vertical resolution close to the ground in the planetary boundary layer. This feature is extremely important for the evaluation of (and incorporation into) high resolution numerical weather forecast models of the future. Due to technical improvements and the intensifying search for alternatives to radiosondes, multi-channel microwave radiometers for the operational profiling of tropospheric temperature and humidity have been developed in the last few years.

An additional advantage of ground-based microwave radiometers is their sensitivity to cloud liquid water. Over the land, passive microwave remote sensing is by far the most accurate method to measure the vertically integrated liquid water content (liquid water path, LWP) other than sporadic and expensive in-situ measurements from research aircraft. More than two decades ago two channel radiometers were shown to achieve high accuracy in the retrieved LWP and the integrated water vapour content (IWV).

In the last few years, further improvements to the LWP retrieval have been made by the inclusion of additional microwave channels and the combination of microwave radiometer measurements with other ground-based instrumentation.

Satellite based remote sensing of LWP over the oceans is a well established method, however, the inhomogeneous distribution of clouds within the satellites field of view (typical several kilometres), can lead to substantial errors. This effect has mostly been neglected for ground-based radiometers whose viewing geometry is often assumed to behave as a pencil beam although the spatial and temporal variability of clouds is high even on scales below the resolution of most radiometers. With a typical wavelength of about 1 cm, practical considerations about the antenna aperture size (about 20 cm) lead to half-power beam widths from 2° to 4° for conventional radiometers. These beam widths correspond to footprints of up to several 100 m at cloud base heights.

Atmospheric water vapour profile information is derived from frequency channels covering 6 GHz of the high frequency wing of the pressure broadened, relatively weak water vapour line (22-28 GHz). With a pressure broadening coefficient of about 3 MHz/hPa information between approx. 300 and 1000 hPa can be resolved with the spectral measurements. In the centre of the oxygen absorption complex the atmosphere is optically thick and the measured radiation originates from regions close to the radiometer. For frequencies further away from the line centre the atmosphere gets more transparent and the channels receive radiation which originates from regions more distant to the radiometer. Due to the known mixing ratio and the temperature dependence of the absorption coefficient of oxygen, information about the vertical temperature distribution is contained in the channels spanning the 8 GHz of the low frequency side.

For a ground based radiometer pointing to zenith, well defined weighting function peaks for each frequency are observed. If the elevation angle is lowered, (and hence the atmospheric path is increased), the peaks shift to lower altitudes. This demonstrates the radiometer’s superiority in the retrieval of the planetary boundary layer temperature.

The cloud liquid water contribution to the microwave signal increases roughly with the frequency squared. It depends on temperature and is proportional to the third power of the particle radius. Therefore measurements at two channels, one influenced mainly by the water vapour line and one in the 30 GHz window region lead to good estimates of LWP and IWV.

Artificial neural networks (ANN) are increasingly used for the retrieval of geophysical parameters from measured brightness temperatures. They can easily adapt to nonlinear problems such as the radiative transfer in the cloudy atmosphere. Additionally, the input parameters of a diverse nature can be easily incorporated into neural networks.

RPG use a standard feed forward neural network where the cost function is minimized employing the Davidon-Fletcher-Powell algorithm. The architecture of the ANN used for the retrieval includes an input layer consisting of simulated brightness temperatures for the HATPRO frequencies, a hidden layer with a certain number of neurons (nodes) and an output layer with the atmospheric variable of interest (LWP, IWV, temperature, or humidity profile). To derive the weights between the nodes of the different layers we generated a data set comprising about 15,000 possible realizations of the atmospheric state, which was divided into three subsets; the first for training, the second for generalization (finding the optimum number of iterations to avoid overfitting), and the third for evaluating the retrieval RMS. For each output parameter the optimal network configuration – number of nodes in the hidden layer, number of iterations and initial weight – was derived and the retrieval performance was evaluated using the third data subset. Generally, it can be stated that all algorithms developed show no systematic errors.

The data set is based on atmospheric profiles of temperature, pressure and humidity measured by radiosondes. In order to analyze profiles of cloud liquid water content (LWC) from the radio soundings, we chose a relative humidity threshold of 95 % as a threshold for the presence of clouds and calculated a modified adiabatic LWC-profile as proposed by Karstens et al. Radiation transfer calculations were performed for each radio sounding using MWMOD. A random noise of 1 K was added to the resulting brightness temperatures to simulate radiometric noise. Realistic noise was also added to the other potential input parameters like the standard meteorological measurements (ground level temperature (Tgr), pressure (pgr), relative humidity (qgr)) and the cloud base temperature (Tcl) as derived by an infrared radiometer (if provided).

It should be noted that a limitation to ANN algorithm, as to all statistical algorithms, is that they can only be applied to the range of atmospheric conditions, which is included in this data set. When extrapolations beyond the states included in the algorithm development are made, ANNs can behave in an uncontrolled way, while simple linear regressions will still give a reasonable, although erroneous, result. Quadratic regressions offer the robustness of a linear regression retrieval with the advantage to model nonlinearities much better than linear regressions. In many cases where unusual atmospheric conditions are likely the quadratic regression is the best choice.

The HATPRO (and related radiometers) supports two temperature profiling modes: Full troposphere profiling (frequency scan across the oxygen line) and boundary layer scanning (elevation scan @ 54.9 and 58 GHz). 22.4 GHz WVL humidity profiling is only available for the full troposphere mode due to the lack of opaque channels on the water vapour line at 22.4 GHz.

For boundary layer temperature profiling the radiometer beam is scanned in elevation between 5° and zenith (figure 3). At the frequencies in use (54.9 GHz and 58 GHz) the atmosphere is optically thick. The frequencies weighting functions peak at 500 m (58 GHz) and 1000 m (54.9 GHz), see figure 2. The receiver stability and accuracy have to be optimized due to the small brightness temperature variations that must be resolved in the elevation scanning method. In the RPG-HATPRO models, the receiver’s physical temperature is stabilized to better than 30 mK over the whole operating temperature range (-45°C to 50°C) to guarantee high gain stability during measurements (>200 sec). The receiver noise temperature is minimized to be better than 700 K which optimizes the overall noise level.

From the weighting functions corresponding to the various water vapour line and oxygen line profiling frequencies, the vertical resolution of the retrieval outputs can be derived:

  • Tropospheric temperature profiles (0-10000 m): 200 m (<5000 m altitude), 400 m above, profile accuracy: +/- 0.6 K RMS (0-2000 m), +/- 1.0 K RMS (>2000 m)
  • Boundary layer temperature profiles (0-1200 m), 30 m vertical resolution on the ground 50 m between 300-1200 m, profile accuracy: +/- 0.7 K RMS
  • Tropospheric humidity profiles (0-5000 m), 200 m vertical resolution (0-2000 m), 400 m (2000 m – 5000 m), profile accuracy: +/- 0.4 g/m3 RMS

The HATPRO is capable of fully autonomous operation and with power recovery enabled will continue in the operational mode that was active when power was lost. The host PC acts as the gateway to facilitate archiving of data off the HATPRO and for visualising the data in real-time. The host PC can be connected to a network if one is available. The host has been set up to allow remote access via VNC.

The HATPRO GPS provides the time stamp for the data and to set the embedded PC within the HATPRO: the time on the host PC is not used in any way.

The GPS also provides positional information and this is logged.

The system has no internal tilt sensors if operating on a moving platform an external inertial motion pack will be required if motion correction is to be applied.

This instrument needs no special licence to operate but will need an initial 25 litres of liquid nitrogen for calibration purposes.

It is recommended that a liquid nitrogen absolute calibration be performed when the radiometer is initially deployed and at the end of a project. With extended deployments, additional calibrations every 6 months are recommended. This calibration requires 25 litres of liquid nitrogen to cool an externally mounted cold target, and all calibrations are stored by the system.

Continual monitoring during deployment and the system diagnostics allows the instrument scientist to determine if the system is operating optimally or requires manufacturer servicing and intervention. The instrument undergoes a manufacturer’s “health check” annually.


The user will need to supply

  • Liquid Nitrogen for the calibration: AMF supply the dewer
  • Any radomes required DURING the deployment
  • Instrument Insurance
    • This system must be insured by the user for £190K
  • Public Liability Insurance
    • We are not liable for any damage or injury arising from the deployment or operation of this instrument when unattended by the instrument scientist.
  • Shipping Expenses
    • The user is liable for all costs arising from the shipping of the instrument both to and from a deployment.
  • IS T&S
    • The user is responsible for coving the travel and subsistence expenses of the instrument scientist while attending the instrument.

The system when packed ready for shipping consists of four flight cases.

  • Radiometer (wheeled shipping case)
    Shipping dimensions: 150 cm (L) x 75 cm (D) x 100 cm (H)
    Shipping weight: 160 kg
  • Accessories box 1 (Host PC, PC power supply & cables, HATPRO cables, manual, fixings)
    Shipping dimensions: 100 cm (L) x 75 cm (D) x 50 cm (H)
    Shipping weight: 50 kg
  • Accessories box 2 (L2N calibration kit, Heater, blower)
    Shipping dimensions: 80 cm (L) x 60 cm (D) x 61 cm (H)
    Shipping weight: 50 kg
  • Accessories box 3 (Table)
    Shipping dimensions: 165 cm (L) x 80 cm (D) x 70 cm (H)
    Shipping weight: 100 kg
  • 25 Litre Liquid Nitrogen Dewar
    Shipping dimensions: 70 cm (H) x 50 cm (Diameter)
    Shipping weight: 10 kg (Shipped empty)

The HATPRO should be deployed on as level a surface as possible, the four feet of the table should be adjusted to ensure that the table top is level before the radiometer itself is attached.

When considering deployment locations some thought needs to be given to obstructions along the line of site. If the instrument is to be operated in zenith mode (looking vertically) only there should be, under most circumstances, no problem; however if the ‘sky-dip’ (tip curve) calibration feature is enabled then the radiometer will scan to an angle of 20° above the horizontal. In this case no obstacles should block the beam within a distance of 10km. If operated in boundary layer mode then the radiometer will scan, potentially down to the horizontal, on either side of the instrument. The most common scan pattern employed has a minimum elevation angle of 5°; no obstacles should block the beam within a distance of 1 km.

2 x 30m power cables are supplied terminated by UK plugs. The heater cable has an inbuilt RCD and it is recommended that the radiometer power cable be plugged in via an RCD

1 x 20m signal cable is supplied to allow a host computer communicate with the instrument. The host system is usually a laptop computer with accompanying power supply and requires approximately 50 cm x 50 cm of bench space that is protected from the environment and has access to power

Although requiring two people to move the system will require securing to a surface to prevent movement due to wind impact. It should also be located on a secure site to avoid theft and vandalism.

The HATPRO will require a liquid nitrogen calibration to be performed. This will require the HATPRO to be in a well ventilated area, the area around the instrument clear of obstacles that may constitute a trip hazard, and a source of liquid nitrogen available to fill a 25 litre dewar (supplied).

At sites where animal activity is likely precautions will need to be taken to prevent chewing\pecking of both cables and the radome.

Manual handling
  • When in its packing case it is recommended that four people be used when lifting. Once out of the case the instrument requires two people to lift and or move it. At least two cables have to run to the instrument and so users should be aware that both the instrument and the power cable constitute a trip hazard and users should take appropriate actions to minimise this.
Electric safety
  • Under no circumstances should any attempt be made to open the up the main body of the instrument. Should the radome require replacing the instrument should first be shutdown, unplugged, and moved inside.
Attended operation
  • There is no requirement for the system to be attended during operation from a safety standpoint.
Liquid Nitrogen
  • Liquid nitrogen at -196°C is kept in double-walled steel vessels specifically designed for the storage and transport of cryogenic gases.
  • When dispensing quantities necessary for handling cryo-preserved materials, or for cooling purposes, or when transferring liquid nitrogen from one container to another, observe the following rules.
  • Work in a well-ventilated area. Nitrogen gas, continuously evolved during handling, may build up causing dizziness and, potentially, asphyxiation by displacing the air in the room. Always use alarms /O2 sensors if provided.
  • Wear protective goggles, or a face-shield, thermal gloves, and a laboratory coat. Absorbent material close to the skin (e.g. gloves) should not be exposed to contact with liquid nitrogen. Do not wear open-toed shoes or sandals when working with liquid Nitrogen.
  • Decant liquid nitrogen slowly, especially into vessels at room temperature, since rapid vaporisation sprays cold droplets into the atmosphere until equilibration of temperature is reached. The extremely cold vapour released during such procedures is particularly damaging to the eyes and the immersion of tubing should be avoided, as the spray issuing from the open end may constitute a hazard to those nearby.
  • Containers, other than large storage Dewar’s, should be of rigid polystyrene (‘Styrofoam’) or double-skinned metal construction. Do not use glass or plastic ‘thermos’ flasks not designed for cryogenic gases, as there is an implosion risk from thermal shock during filling.
  • Do not touch any non-insulated surface cooled to liquid nitrogen temperatures, as adhesion of the skin will occur, due to freezing of the moisture layer at the interface, resulting in contact burns. Handle all cooled objects with tongs or forceps, and do this without undue delay, as these will also cool rapidly by conduction.
  • ALWAYS replace stoppers or lids, loose-fitting only – never seal vessels containing liquid nitrogen. Ensure vented stoppers do not ice up as this could cause excessive build up of pressure inside the vessel and thus create a risk of an explosion.
  • First Aid Treatment for Cold Skin Burns
    • Flush the area of skin with tepid water.
    • Do not apply direct heat or hot water.
    • Do not use a forceful flow of water as this can cause tissue damage
    • Move the causality to a warm place and seek medical attention.
    • If burn severe call an Ambulance
      • While waiting for medical attention continue to flush with tepid water and remove any tight jewellery.
      • Do not allow patient to smoke or offer hot beverages.

The HATPRO is deployed on a fold out table that must first be levelled (spirit level required).

Table dimensions: 105 cm (L) x 40 cm (W) x 100 cm (H)
Table weight: 40 kg

When unpacked and deployed the HATPRO is:

Footprint (with no blower & heater): 90 cm (L) x 43 cm (W) x 59 cm (H)
Weight (not including shipping case): 60 kg
Footprint (with blower & heater): 112 cm (L) x 45 cm (W) x 65 cm (H)
Weight (not including shipping case): 70 kg

The HATPRO is attached to the table with the M8 bolts supplied (Allan key required)


Supply: 100 – 240V AC (at 47 – 63 Hz)
Consumption (Radiometer): 350W @ turn-on power, 150W under stabilised operational conditions
Consumption (Heater): additional 1.8kW


Operation temperature: -60°C to 45°C


Radiometer power cable is 30m long and terminated by UK plug and RCD

Heater power cable is 30m long and terminated by UK RCD plug

Signal cable is 30m long and terminated by a 9pin D connector for connection to the serial port of the host PC

The output files stored on the host PC are user selectable and can be changed in real-time without interrupting the continuous operation of the radiometer.

Field Data
  • The instrument produces a range of out files and all are text format.
  • The user can download (but not delete) this data from the instrument but it should be noted that this data will not have been quality controlled.
Archive Data