4. AVHRR Level 1b Products Overview

Table of Contents

4. AVHRR Level 1b Products Overview

4.1 The AVHRR/3 instrument

The Advanced Very High Resolution Radiometer/3 (AVHRR/3) is a multipurpose imaging instrument used for global monitoring of cloud cover, sea surface temperature, ice, snow and vegetation cover characteristics and is currently flying on NOAA-15, -16, -17, -18, -19 and Metop-A. The AVHRR/2 version 2 of the instrument was flown on NOAA-7 to -14 in a five-channel version. A detailed account of the instrument technical characteristics is given in [RD41], but we will give in the following sections the basic information necessary for understanding and using the product.

4.1.1 Technical description

The AVHRR/3 is a six-channel scanning radiometer providing three solar channels in the visible/near-infrared region and three thermal infrared channels. The AVHRR/3 has two one-micrometre wide channels between 10.3 and 12.5 micrometres. The instrument utilises a 20.32 cm (8 inch) diameter collecting telescope of the reflective Cassegrain type. Cross-track scanning is accomplished by a continuously rotating mirror directly driven by a motor. The three thermal infrared detectors are cooled to 105 kelvin (K) by a two-stage passive radiant cooler. A line synchronisation signal from the scanner is sent to the spacecraft MIRP processor which in turn sends data sample pulses back to the AVHRR.

The spectral channels of AVHRR/3 are not exactly the same as AVHRR/2, and include an additional channel 3a in the near infrared (NIR). AVHRR/3 has six spectral channels between 0.63 and 12.00 micrometres: three in the visible/near infrared and three in the infrared. Channel 3 is a split channel: channel 3a is in the solar spectral region (1.6 µm) whereas channel 3b operates in the infrared around 3.7 µm.

Although AVHRR/3 is a six-channel radiometer, only five channels are transmitted to the ground at any given time. Channels 3a and 3b cannot operate simultaneously. The transition from channel 3a to 3b and vice versa is done by telecommand and reflected in the science data. For Metop-A, channel 3a is operated during the daytime portion of the orbit and channel 3b during the night-time portion.

The data from the six channels are simultaneously sampled at a 40-kHz rate and converted to 10-bit binary form within the instrument. The data samples from each channel are output in a non-continuous burst of 10 space samples, 2048 Earth samples and 10 internal calibration target samples per scan.

The following table summarises the spectral characteristics of AVHRR/3.

     Channel
    Central wavelength (µm)
    Half power points (µm)
    Channel noise specifications
     
    S/N @ 0.5% reflectance
    NEdT @ 300K
    1
     0.630
     0.580 - 0.680
    9:1
    -
    2
     0.865
     0.725 - 1.000
    9:1
    -
    3a
     1.610
     1.580 - 1.640
    20:1
    -
    3b
    3.740
    3.550 - 3.930
    -
    <0.12 K, 0.0031 mW/(m2 sr cm-1 )
    4
    10.800
    10.300 - 11.300
    -
    <0.12 K, 0.20 mW/(m2 sr cm-1 )
    5
    12.000
    11.500 - 12.500
    -
    <0.12 K, 0.21 mW/(m2 sr cm-1 )

Table 4.1: Spectral characteristics of AVHRR/3

4.1.2 Nominal scanning geometry

AVHRR/3 is an across-track scanning system with a scan range of ±55.37° with respect to the nadir direction. The field of view (IFOV) of each channel is approximately 1.3 milliradians (0.0745°) leading to a square instantaneous field of view size of 1.08 km at nadir for a nominal altitude of 833 km. The scanning rate of 360 scans per minute is continuous (1 scan every 1/6 second). There are 2048 Earth views per scan and per channel for a swath width of about ±1447 km (sampling time of 0.025 ms). The sampling angular interval is close to 0.944 milliradians (0.0541°). The distance between two consecutive scans is approximately equal to 1.1 km.

The following table summarises the scanning characteristics.

    Characteristics Value Unit
    Scan direction East to West (northbound) -
    Scan type continuous -
    Scan rate 0.025 ms
    Sampling interval (duration) 0.1667 s
    Sampling interval 0.0541 deg
    Pixels/scan 2048 -
    Swath ±55.3 deg
    Swath width ±1446.58 km
    IFOV 0.0745 deg
    IFOV type square -
    IFOV size (nadir) 1.08 km
    IFOV size (edge) - across track 6.15 km
    IFOV size (edge) - along track 2.27 km
    Scan separation 1.1 km

Table 4.2: Nominal scanning characteristics of AVHRR/3

On the NOAA satellites, the on-board processor samples the real-time AVHRR/3 data to produce reduced resolution Global Area Coverage (GAC) data (Figure 4.1). Four out of every five samples along the scan line are used to compute one average value, and the data from only every third scan line are processed. As a result, the spatial resolution of GAC data near the subpoint is actually 1.1 km by 4.4 km with a 2.2 km gap between pixels across the scan line, although generally treated as 4 km resolution. All of the GAC data computed during a complete pass are recorded on board the satellite for transmission to Earth on command. The 10-bit precision of the AVHRR data is retained. The following table summarises the different resolution/grid characteristics of the data.

    Characteristics Value Unit
    Pixels/scan 409 -
    Sampling size (nadir) 4.4 (across-track) x 1.1 (along-track) km
    Sampling grid (nadir) 5.5 (across-track) x 3.3 (along-track) km

Table 4.3: Resolution and grid characteristics of AVHRR/3 GAC data

Figure 4.1: Simulated earth-surface footprints for AVHRR/3 showing relation between full resolution data (Local Area Coverage, LAC, in blue) and reduced resolution Global Area Coverage (GAC, black outlines).
[Based on NOAA original. Compare with equivalent schematic in Section 4 of the ATOVS Level 2 Product Guide.]

4.1.3 Instrument calibration

The AVHRR/3 calibration is different for the visible and the IR channels.

4.1.3.1 Visible and near-infrared channels

There is no on-board calibration for the visible channels (channels 1 and 2) and channel 3a. The visible and near-infrared channels of the AVHRR/3 are calibrated prior to launch following a protocol which has evolved over the past two decades, by means of using a calibrated light source and varying the level of illumination as desired for the different channels. At each level of illumination, measurements of the signal issuing from the AVHRR are made, and the mean and standard deviation recorded, and converted to digital counts. Pre-launch calibration results in the form of a simple linear regression relationship between the measured AVHRR signal, expressed in counts, and the albedo of the light source at different levels of illumination are then given.

During the ground processing, the gain and intercept values are selected and applied according to the count values. For AVHRR/3 a dual slope/gain function is used for the visible (channels 1 and 2) and near-infrared (channel 3a) channels to enhance the radiometric resolution at low radiance or reflectance values. For every channel and for every one of the two gain regimes, a set of pre-launch calibration factors (slope and intercept) is provided. It should be noted that the specifications of the split gain ranges are not fixed but may alter from instrument to instrument and during the life time of the instrument.

The AVHRR visible/NIR channels do not have effective on-board calibration, and are known to decrease in response as a function of time, as well as due to launch processes. Pre-launch calibration is carried out to confirm the linearity of the detectors, and to establish baseline calibration coefficients. Post-launch, they will be calibrated against stable surface regions and against other satellites, following various well-established techniques, which constitute what is known as vicarious calibration.

4.1.3.2 Thermal infrared channels

As for the visible channels, a pre-launch calibration is carried out to confirm the linearity of the detectors for different instrument operating temperatures and for the full range of expected Earth target temperatures.

During each in-orbit scan line, the AVHRR views three different types of targets. It first outputs 10 counts when it views cold space, then a single count for each of the 2048 Earth targets (pixels), and finally 10 counts when it views its own internal black body target. The cold space and internal black body target views are used to calibrate the AVHRR, because a radiance value can be independently assigned to each target. The internal black body radiance is estimated from the internal black body temperature, measured by four platinum resistance thermometers (PRTs) embedded in the AVHRR instrument, and the space radiance is computed from pre-launch data.

4.2 Overview of the ground processing and calibration

The Level 1 ground processing chain is illustrated in Figure 4.2 below.

The first goal of Level 1 ground processing for AVHRR is the generation of the AVHRR Level 1b product, containing as the main geophysical parameter reflectivity (for channels 1, 2 and 3a) and calibrated radiances (for channels 3, 4 and 5). This processing is data driven and is applied to every science source packet to generate Level 1b products.

Additionally, scenes analysis is performed within the Level 1b processing, with the main goal of assessing cloud contamination for every pixel. This information will be used later on within the processing of ATOVS and IASI Level 2. Only the cloud analysis information is included in the output AVHRR Level 1b product distributed to users. AMSU-A Level 1b data is needed as input to the scenes analysis. In Section 10, the context of the ATOVS, IASI and AVHRR processing chain interactions is provided for information.

The ground processing is applied to data from the AVHRR/3 instruments on both Metop and NOAA satellites. In the case of NOAA, GAC frames are the input raw data flow. These data do not have the same resolution, but the respective output Level 1b products from the EUMETSAT CGS have the same structure, contents and format.

Note: The landmark navigation results are not yet used operationally.

Figure 4.2: Functional overview of the AVHRR Level 1 ground processing chain

4.2.1 Pre-processing

Basic raw data validation checks are applied and the instrument telemetry and other auxiliary data are also validated and related to the input raw data flow.

After that, calibration data need to be processed to retrieve the calibration functions that will allow the Level 1b generation. Concerning the visible and NIR channels, calibration coefficients are calculated based initially on on-ground characterisation information, and later on during the mission lifetime on updated characterisation information from vicarious calibration. Each pixel count value determines the gain regime and the slope and intercept of the linear regression.

Concerning the calibration for the IR channels, the on-board calibration system is used. Black body temperature measured by the PRT is extracted for every pixel from the AVHRR instrument source packet. Black body radiances are computed from this temperature taking into account the spectral response function of the channel. Average black body estimated radiances and counts, as well as on-ground characterised space radiance and average space counts, provide the two points for linear interpolation to estimate radiances corresponding to each Earth target pixel along the scan line. An additional non-linear correction of this radiance based on pre-launch calibration data is further applied. A minimum of 55 scan lines are necessary to obtain a compete set of calibration coefficients.

Finally, geolocation for tie points along the scan line is performed, as well as their satellite and solar zenith and azimuth angles. Tie point geolocation is estimated by means of a satellite ephemeris model and an instrument scanning model. This first step in the navigation is based on default attitude values, directly after data acquisition.

The calculation of the satellite zenith and azimuth is done by applying a transformation matrix to the Earth fixed satellite position coordinates previously obtained during the tie point geolocation. The solar azimuth and zenith are obtained taking into account the actual solar declination, for which an accurate time stamp for the scan line is previously estimated.

4.2.2 Level 1b processing

4.2.2.1 Generation of the Level 1b product contents

Calibration coefficients are applied to both visible/NIR and IR channels, in order to convert channel count values into reflectivity and radiances, respectively. These are the geophysical parameters which constitute the AVHRR Level 1b products.

Using the tie points geolocation information, each individual pixel is geolocated, following either linear interpolation or Lagrangian interpolation. The same interpolation schemes are used to estimate satellite and solar zenith and azimuth angles for every pixel. This information is also included in the AVHRR Level 1b product.

Using a high resolution coast line data set and the geolocation information estimated above, a surface type is assigned to each pixel. For Metop full resolution data, an automatic adjustment based on landmark position processing is also performed, which will give us also a good assessment of the positioning accuracy, as well as an accurate platform attitude.

4.2.2.2 Scenes analysis

The main functionality of the scenes analysis algorithm is to determine whether a pixel is contaminated by clouds or not. Partially cloudy pixel or pixels covered with semi-transparent clouds will be declared as cloudy. The algorithm also identifies clear pixels which may be covered with snow or ice. In addition, for pixels identified as clear, the surface temperature is determined, and for pixels identified as cloudy the cloud top temperature is computed. The scenes analysis algorithm is based on a threshold technique and works nominally on a pixel-by-pixel basis. The threshold technique compares the image data with thresholds which mark the border between the physical signal (i.e. brightness temperature and reflectance factor) of a pixel without clouds and a pixel containing clouds. The scenes analysis algorithm uses a prediction, based either on forecast data of the current AVHRR Level 1b scene or on climatological values from a database. Also, spatial information (mean, standard deviation) is used to supplement the scenes analysis process. The threshold technique makes use of the spectral information provided for each pixel with the measurements in all available channels.

Mainly, there are six steps performed in the scenes analysis:

In more detail, the different types of cloud detection tests are briefly listed below, where A1 and A2 are respectively the albedo computed from channels 1 and 2, and T3.7, T11 and T12 are respectively the brightness temperatures estimated from channels 3b, 4 and 5.

The thresholds for the different tests depend on season, geographical location, daytime, satellite viewing angle and the availability of distinct data sets (e.g., forecast data and/or climatological data).

As output from the scenes type identification, cloud cover information is retrieved and included in the Level 1b product. Retrieval quality information is also included in the Level 1b product. The scenes analysis outputs are forwarded within the EPS CGS to the ATOVS and IASI Level 2 processors.

4.2.3 Post-processing and quality control

This function covers both the radiometric and the geometric quality assessment. The radiometric quality assessment consists of the production of a detailed set of radiometric characteristics of the data for each detector/channel, this for different imaged scenes during the dump (day/night sides, calibration viewing, etc.).

The geometric quality control extracts from the Level 1b data areas corresponding to geographical areas of interest (landmarks), applies the projection using the attached tie-point information and compares this with landmarks extracted from a high-accuracy digital map. The produced information is used to generate detailed quality statistics for analysis purposes. Note that the set of landmarks and statistics produced is different for each instrument chain, as the characteristics of the instrument make a common approach not practical.

Finally, statistics produced by the quality control function are used to perform trend analysis and to derive information on the misalignments between instruments and mis-registration between channels. Updates of the model parameters for the platform/instrument being processed are then estimated and this information allows compensating for slow drifts and changes in these parameters.

4.2.4 Nominal and degraded processing modes

The following table summarises non-nominal processing situations, corresponding to either corrupted/missing Level 0 data, missing auxiliary information and/or instrument ancillary data, missing channels, or invalid calibration information.

    Description of anomaly Influence on processing
    Instrument anomalies
    Missing motor telemetry data No processing
    Missing electronics telemetry data No processing
    Missing Channel 1 data Landmark navigation without NDVI test, scenes analysis without reflectance test over land and with degraded snow/ice detection
    Missing Channel 2 data
    Missing Channel 3a data Scenes analysis with degraded snow/ice test
    Missing Channel 3b data Scenes analysis without relevant brightness temperature difference tests
    Missing Channel 4 data Degraded landmark navigation and scenes analysis without the relevant brightness temperature difference tests
    Missing Channel 5 data
    Missing voltage calibrate status No processing
    Missing status of cooler heat, scan motor and/or Earth shield Degraded Level 1b processing, including scenes analysis
    Time sequence
    Bad time field, but it can be inferred from previous good time Degraded geolocation information and scenes analysis
    Bad time field, and it can't be inferred from previous good time No geolocation information, no angular relations, no scenes analysis
    Time discontinuity detected Degraded geolocation and scenes analysis
    Repeated scan times
    Scan time not corrected for clock drift
    Earth location
    No satellite position and velocity No geolocation, no angular relations, no scenes analysis
    Degraded satellite position and velocity Degraded geolocation, angular relations and scenes analysis
    Calibration
    Degraded or incomplete input data for channels 3, 4 and 5 calibration Use of previous or pre-launch calibration data / Degraded IR radiances and brightness temperatures, landmark navigation and scenes analysis
    Navigation
    Degraded orbit ephemeris data Use of latest available ephemeris file / Degraded geolocation and scenes analysis
    No information on Earth location No geolocation, no angular relations, no scenes analysis
    Degraded satellite attitude Degraded geolocation, scenes analysis

Table 4.4: Summary of non-nominal processing situations

All the situations above are adequately flagged within the Level 1b product. More details on the relevant flag fields are to be found in the product description sections.

An additional non-nominal processing situation is the edge of a dump or possible gaps in a continuous measurement sequence. In that case, the first/last 55 lines of data are used. If a continuous measurement sequence contains less than 55 lines, all the available lines are used for the processing of that sequence, and the calibration cycle contains then less than 55 lines. A degraded calibration for any of these reasons is also flagged within the product.

4.3 AVHRR Level 1b product characteristics and use

4.3.1 General characteristics

Table 4.5 summarises the main characteristics of AVHRR Level 1b products available to users. All products contain quality control and other information about the retrieval and their use, which are important to know when you choose the product needed for your application. Two Level 1b products are generated, from Metop and from NOAA/GAC data.

     Product
    Main geophysical parameter
    Accuracy
    Resolution
    /grid spacing (nadir)
    Swath width
    Coverage
    Generated
    AVHRR Level 1b from Metop
    Geolocated reflectivity from visible and NIR channels, and radiances for IR channels / cloud coverage information
    Radiometric: 1 K for IR channels
    /
    Geolocation: 1 km
    /
    Channel to channel mis-registration: <0.1 mrad
    1.1 km x 1.1 km
    /
    1.1 km x 1.1 km
    2893 km
    Global and continuous
    EPS CGS
    AVHRR Level 1b from NOAA/GAC
    4.4 km (across-track) x 1.1 km (along track)
    /
    5.5 km (across-track) x 3.3 km (along track)

Table 4.5: Summary of the main characteristics of AVHRR/3 Level 1b products

Apart from the main geophysical parameter given for each pixel, navigation information is given for each scan line, as well as angular relations for every navigation point. Calibration data are also appended in the product (slope and intercept) for both visible/NIR and IR channels.

4.3.2 Quality information in the products

A number of quality flags are generated during the Level 1b processing, associated with individual scan lines. The following are the most relevant with respect to data use. A full list and detailed explanation of all flags is given later in the Level 1b spatial averaged products content and format description.

4.4 Summary of AVHRR Level 1b product current and potential applications

The main internal use of AVHRR Level 1b product is for further processing in the ATOVS Level 2 and IASI processors in the EUMETSAT CGS. AVHRR/3 radiances and geolocation information included in the AVHRR Level 1b product are passed to the IASI Level 1 processor in order to support IASI navigation and radiative surface analysis. The cloud analysis is further used in the IASI and ATOVS Level 2 processors.

Originally, NOAA designed AVHRR for the following tasks: Channels 1 and 2 were to be used to discern clouds, land-water boundaries, extent of snow and ice, and the inception of snow/ice melting, and to monitor terrestrial vegetation employing the computation of the NDVI; Channels 3, 4, and 5 were to be used to measure the temperature of clouds and the sea surface, and for night-time cloud mapping. Several decades of availability of AVHRR data have proved its usefulness for a range of applications in meteorology, oceanography and terrestrial sciences, extending far beyond these original objectives. Most of these applications imply the derivation of geophysical parameters beyond the contents of the AVHRR Level 1b product (which are basically sensor radiances and cloud information) and that higher-level product derivation is partially covered by products/software generated by the SAFs.

A good reference for the AVHRR instrument is [RD42], which includes a thorough review of all AVHRR applications.

4.4.1 Applications in meteorology

Day and night cloud mapping is the main application of AVHRR data in meteorology, especially at high latitudes where data from geostationary satellites are severely distorted due to Earth curvature. The AVHRR/3 Level 1b contains basic cloud map information necessary for the processing of higher-level ATOVS and IASI products. Additionally, the NWC SAF develops an end-user software package for derivation of a cloud mask from AVHRR images.

Other important applications of AVHRR in meteorology are in combination with information from the ATOVS sensors (HIRS, AMSU-A and MHS) flying on the same platform. Together, these systems provide a suite of infrared and microwave channels that can be used to profile atmospheric temperature and humidity. Such meteorological applications include interpreting cloud top temperatures and heights for predicting and monitoring storms, differentiating ice, water, shadow and other aspects of clouds, deriving polar winds from monitoring cloud motions, water vapour content of the lower atmosphere, and the study and monitoring of tropical cyclones.

Finally, surface radiative fluxes are an essential geophysical parameter for climatological studies which can be derived from AVHRR data. Both the OSI SAF and the LSA SAF generate products containing this information.

4.4.2 Applications in oceanography

Multi-Channel SST (MCSST), computed from channels 4 and 5 of the AVHRR, is the main geophysical parameter of use in AVHRR oceanographic applications.

Infrared AVHRR imagery has also proven very useful in mapping mesoscale ocean features in terms of their SST signatures. Major ocean currents, such as the Gulf Stream, are readily visible by their marked SST gradients. Techniques have been developed for mapping ocean current variability from their signatures in the AVHRR SST imagery. The OSI SAF is developing an operational AVHRR-based global SST product suitable for these purposes.

Another oceanographic application of AVHRR data is in the study of sea ice. Properly filtered for clouds over ice, AVHRR imagery can be used to compute sea ice concentration, type and ice edge location. The OSI SAF develops such a product, based not only on AVHRR imagery, but also on additional passive and microwave sensor information. Finally, a sequence of AVHRR images, either visible or thermal infrared for polar winters, can be used to compute ice motion.

4.4.3 Applications in terrestrial sciences

The AVHRR has evolved into an invaluable resource for studying the land surface. AVHRR's frequent day/night synoptic coverage and high horizontal resolution are features that make the system unique for such applications.

In the area of monitoring terrestrial vegetation, the AVHRR-derived NDVI has proven to be a very robust and useful quantity to monitor vegetation, land cover and climate. The index has been produced and utilised globally and regionally. The NDVI is related to the health of the vegetation growth, and has therefore been used for drought forecasting, crop growth monitoring and to map forest fire fuel potential.

Multi-channel imagery from the AVHRR has also proven to be useful in snow cover mapping. The frequent coverage of the AVHRR is again the prime advantage in being able to distinguish clouds from snow cover with their similar albedo signature. Combined with topographic relief information, snow cover from AVHRR can be converted to snow-water equivalent to give an estimate of the amount of water reserve represented by the winter snow pack.

The LSA SAF develops vegetation products from AVHRR and other sensors, which can be used for the above applications.