Report on AERONET Workshop, Huelva, Spain, May 10-14 2004.

Ross Mitchell

CSIRO's Aerosol Ground Station Network (AGSNet) has been affiliated with NASA's Aerosol Robotic Network (AERONET) since the former's inception in 1998. AERONET consists of a global network of sun photometer stations measuring aerosol properties, with affiliated groups on all continents. AGSNet represents Australia's contribution to AERONET. This workshop brought together 140 AERONET participants from across the globe with the purpose of information interchange, review of operational procedures, and the setting of new directions. I was the sole Australian delegate at the workshop.

Day 1: 10/05/2005

The workshop began with overview presentations from NASA and the major contributing sub networks (Europe, Canada, China, Australia, Siberia and Spain) with only the Chinese delegate unable to attend. In presenting the US/NASA overview, AERONET leader Brent Holben (NASA/GSFC) pointed out a major strength of the network, the fact that it is global and hence contributing to a global understanding of aerosol dynamics and distribution. By the same token, the continued growth of the network has increased the workload at NASA/GSFC to the point where devolution of instrument maintenance and calibration to regional sub-networks is now inevitable.

In my overview presentation I emphasised the different approach to instrumentation and data collection required in view of the remoteness of our sites, and presented result snapshots of the two primary Australian continental aerosol types, smoke from savannah biomass burning, and wind blown dust. An image of the Griffith dust storm of November 2002 created much comment. I pointed out the value of nephelometer measurements in tracking dust storms, over and above the standard AERONET sun photometer measurements that are limited to cloud-free daytime conditions.The exposure through this plenary presentation was vital in facilitating informal contacts later in the meeting.

Subsequent talks on day 1 were primarily from NASA/GSFC AERONET staff giving status reports on the network. Oleg Dubovik discussed retrieval of aerosol size distribution and other microphysical properties from sky radiance inversions, and pointed out the need to use the non-spherical (spheroidal) option for dust aerosol. Tom Eck compared AERONET measurements in SE Asia with the MODIS aerosol product, and concluded that the latter gave biased results in many cases due to the swampy nature of parts of the surface, over which the assumed relation between near infrared and red surface reflectance in the MODIS algorithm breaks down. Other presentations dealt with a variety of operational changes, for example, the move to a relational database for improved web page delivery of products, and to facilitate instrument and site management. There is also a move to integrate AERONET data with the MODIS aerosol product and back trajectory analysis. Day 1 concluded with some novel applications of AERONET including measurement of solar flux, cloud optical depth, and the status of a network of Micro Pulse Lidars (MPLNET), the latter of undoubted potential if collocated with AERONET sun photometers, but plagued by laser reliability problems.

Day 2: 11/05/2004

Day 2 started with Yoram Kaufman discussing the synthesis of AERONET measurements and the MODIS satellite aerosol product. There is a clear low bias in AERONET measurements of dust, which Kaufman suggested as being due to erroneous rejection of dust events as being cloudy. This issue was revisited later in the meeting with Sasha Smirnov, the architect of the AERONET cloud screening algorithm. Didier Tanre showed that for much North African aerosol, smoke and dust are mixed. He reported that during the dry season, there is more absorption in the large size mode than the small mode. The reason for this is unclear as the large mode is expected to be dust with lower absorption than the fine mode (smoke).

Paola Formenti reported on dust measurements on the Mediterranean basin. Single scattering albedos ranged from 0.92 to 0.99 at 660 nm, with iron mass fractions in the range 4 to 7%. Jim Spinhirne (NASA/GSFC) reported on global aerosol profiling using the GLAS satellite. Persistent high aerosol loadings were seen over Northern Australia during the late dry of 2003 (Oct-Nov). Later discussions with Jim will lead to collaborative work on this, bringing together the sun photometer data from the three northern AGSNet sites with the GLAS. Chris Cattrall (Arizona) reported on the assembly of lidar ratios for different aerosol types in preparation for the space borne lidar CALIPSO mission, and found that coal burning emissions from China and SE Asia required separate classification from previously considered aerosol types. Yevgeny Derimian (Ben Gurion) looked at aerosol classification using the difference between single scattering albedo at 440 and 1020 nm, with positive differences indicating pollution and negative flagging dust. This general result is also seen in the negative correlation between Fe content and black carbon. Dust at Sede Bokker (Negev Desert) has 2~4% Fe and 0~2% BC by mass, and an Angstrom coefficient <0.3. Pollution aerosol has <2% Fe, >3% BC and Angstrom coefficient >0.7. Overall, Fe makes up between 1 and 3 % of total particulate mass, almost all confined to the coarse size mode. Dust collected from Cape Verde has somewhat higher Fe content (~5.5% TPM). Dubovik (NASA/GSFC) attempted to retrieve fine mode aerosol sources from MODIS data using the GOCART transport model, while Stephan Kinne (Hamburg) explored the synthesis of AERONET and satellite data in the assembly of a global aerosol model.

Maria Peronne presented sun photometer measurements from Lecce on the Heel of Italy. Results indicated the presence of the iron oxide illite but not hematite in Saharan dust events, with imaginary parts of the complex refractive index < 0.005 and Angstrom coefficients in the range 0.3-0.9. The analysis concerning iron oxides was hotly contested in questions after the talk. Monica Campanelli (Rome) reported that the Italian network uses several Japanese Prede instruments. She described an in situ calibration technique developed in collaboration with Terry Nakajima that yielded standard uncertainties in system calibration constant lnV0 of between 1-2.5% in a single day. [Note that this is not adequate for Australian conditions where values of 0.3-0.6% are achieved and indeed required since the standard uncertainty in optical depth is u(lnV0)/m where m is the optical airmass (1 at zenith). Hence background optical depths (0.03) observed at m=1 require u(lnV0)=0.3% to achieve 10% coefficient of variation (standard uncertainty/mean)]. Nataly Chubarova (Moscow) examined absorption optical depths due to NO2 and found values in the range 0.08 to 0.12 at 440 nm under highly polluted conditions near Moscow. By contrast, background values over unpolluted areas (eg the Australian Outback) are typically <0.002 (Mitchell and Forgan 2003).

Day 3: 12/5/2004

Day 3 began with Norm O'Neill (Canada) discussing spectral techniques for separating fine and coarse mode aerosol. If the spectral aerosol optical depth consists of the sum of a spectrally neutral (coarse) and spectrally varying (fine) component, then it is possible to separate these under certain circumstances. This technique may have an important application in screening smoke aerosol from thin cirrus, a situation often arising in Northern Australian sites. Norm went on to extend this method to cases where the size distribution is evolving due to particle growth but the form of the size distribution remains fixed - so called 'log translatable growth'.

Takamura and Aoki presented results from Japan's SKYNET. This network is based on Prede instruments and hence is operationally incompatible with AERONET. Other instruments including in situ sampling and nephelometers are also operated. Takamura found that single scattering albedos calculated from in situ sampling were higher by about 0.1 compared with those found by combining absorption coefficients measured using a Particle Soot Absorption Photometer (PSAP) with scattering coefficients measured by a M903 nephelometer. The reason for this large difference is not clear.

Finally, Victoria Cachorro (Spain) showed that, on occasion, calibrations performed by AERONET are clearly erroneous, leading to negative aerosol optical depth in some cases, and apparent diurnal cycling, both well known artefacts of assuming an incorrect instrument calibration. The problem was realized by AERONET who then suggested a new calibration based on one day's in situ Langley analysis. However, this subsequent calibration was also found to be in error. In any event, the data were unable to be raised to level 2, even though instrument performance was otherwise normal. Cachorro described a post facto method of correcting this without recourse to the raw data. While valid, Ellsworth Dutton pointed out in a question after the talk that this was unlikely to be any better than going back to the raw data and applying Langley analysis to a number of clear days in the deployment period, a course of action that couldn't be followed because in the standard AERONET data collection system contributing stations do not have access to the raw data.

Day 3 continued with a demonstration of new instruments by Cimel Electronique, followed by a poster session. Cimel's new sun photometers are evolving along two lines. First, the spectral coverage of the standard model is being extended into the near infrared with a channel at 1.6 microns. This requires one of the silicon detectors in the old model to be replaced by an InGaS detector. Also, it is being achieved without sacrificing a spectral channel, by placing the 1.6 micron filter in a position formerly blanked. Second, the polarized version of the instrument was always hamstrung by the fact that only one spectral channel (870 nm) was polarized, and that this occupied three slots on the filter wheel, drastically reducing spectral coverage. The new polar version overcomes this by having a polarization wheel rotating independently from the filter wheel, thus allowing polarization measurements in all spectral channels.

In addition to hardware developments, informal discussions with Cimel engineer Marius Canini indicated a capacity for flexible development of the software of existing instruments. For example, Jim Freemantle (Canada) raised the need for more frequent direct sun observations than the default 15 minute sampling that the Cimel currently uses when not in the Langley airmass range (2 - 6). After some discussion, Canini agreed to implement a version of the software with 3 minute default sampling,
and make it freely available to interested members of the AERONET community.

Day 4: May 13, 2004

This day was devoted to workshop sessions as opposed to the more formal presentation style of preceding days. There were two concurrent morning sessions: Calibration and Future measurements/operations. I attended the former, which began with Tom Eck (NASA/GSFC) describing current calibration procedures employed by AERONET, in brief, Langley calibration of a small number of reference instruments at Mauna Loa, Hawaii, followed by shipping back to NASA/GSFC and transfer of the calibration to field instruments on the roof by simultaneous operation under appropriate conditions. Sasha Smirnov than spoke about cloud screening operations, and cited the Canberra fires as being the most testing case ever encountered, with only 40% of the data being identified as not cloud affected even though it was known from direct observation that the sky was filled with smoke, not cloud, during the week following the fires. The high rate of rejection was caused by the algorithm focusing on temporal variations, which in this case were caused by variations in the smoke thickness rather than cloud. I suggested that it would be useful to include a check on Angstrom coefficient in the procedure, and to 'reprieve' rejected periods where the temporal variability was high but the Angstrom coefficient indicated small (smoke) particles rather than large water droplets.

I then presented material covering calibration operations at CSIRO, beginning with the role of the EOC Calibration Facility in AGSNet/AERONET. I showed that preliminary sky radiance calibrations performed using the Labsphere URS-600 uniform radiance source agreed with NASA calibrations to better than 2% (standard uncertainty/mean) in 3 out of 4 channels and to better than 6% in the other, a very promising start. I then turned to the direct sun calibrations, where the difficulty of frequent return of Australian instruments to GSFC for calibration has meant that much high quality data has not been raised to level 2, similar to the situation related by Cachorro for the Spanish network on the previous day. I pointed out that Australian Outback sites provide ideal conditions for in situ calibrations, with aerosol optical depths low (~0.02 at 500 nm) and stable through the winter months. In addition, advanced calibration techniques have been developed that provide smaller uncertainties than those derived from the current AERONET approach, as published by Mitchell and Forgan (2003) (MF). I suggested that for sites such as these, the AERONET protocol needs to be expanded to accept well-documented alternative calibration strategies.

Following these presentations, there was a session charged with working out how to set up the distributed calibration facility concept. Tom Eck began by reiterating the current AERONET requirements (high altitude reference calibrations combined with transfer to field instruments at another suitable site). I responded by suggesting that this approach was not foolproof, as shown by the Spanish experience, and required extension to include high quality in situ calibrations as detailed in my earlier talk. Tom responded by calling for validation of these calibrations against the high altitude method. After some discussion it was agreed to select datasets from a number of sites that had been raised to level 2 through standard AERONET calibration, then reprocess these using the MF scheme, and finally compare the resultant calibrations and aerosol optical depths over an extended period. If the agreement was within expected uncertainty envelope, then the MF technique would be adopted as an alternative AERONET calibration method.

The meeting then turned to laboratory calibration of the sky radiance channels.
The position at the outset was that AERONET would provide a small number (2-3) travelling standard Cimel instruments that would be rotated between Distributed Calibration Facilities, and would serve to tie all sky calibrations to a common reference. However, what that reference would be is not yet clear and was complicated by some teething troubles with the URS-600 at NASA GSFC.

Misha Sorokin (NASA/GSFC) had earlier reported a discrepancy between the spectral radiance of the AERONET URS-600 as supplied with the source, as compared with spectral measurements made at GSFC. This discrepancy has delayed adoption of the URS-600 into the AERONET processing chain. Currently, sky calibrations are performed using the 30-inch sphere. However, this has to be serviced regularly and is offline for a substantial fraction of time, making an alternative source an attractive proposition. On the other hand, the URS-600 has only a 38mm port, requiring separate calibration of the two collimators on the Cimel, whereas both can be calibrated simultaneously using the 30-inch sphere; moreover, three instruments are routinely lashed together and calibrated en bloc. There was also a perception that the output beam of the URS-600 was spatially non-uniform across the port and varied strongly with distance from the source. However, later examination of the Labsphere document on which this view was based showed that it is the irradiance (or flux density) that depends on the distance from the source, as is obvious from the required integration over solid angle inherent in the definition of the irradiance. The quantity of importance here, the radiance, is spatially uniform across the exit port to typically 1-2%, and does not vary with distance from the source, provided that back reflection into the sphere is taken into account when the instrument is close to the source, and the port fills the field of view of the radiometer. Given our experience in this area, I was given the task of developing a detailed calibration protocol including the design of instrument and cal source mounts to be used at all Distributed Calibration Facilities.

There is some diversity in calibration arrangements across the different sub-networks. For example, the European PHOTON network has many polarized instruments and thus has developed a calibration system ('PolBox') to deal with these. The Spanish intend purchasing a large integrating sphere, possibly for the calibration of instruments additional to the Cimels. Brent Holben was strongly in favour of uniformity across all regional calibration facilities, with each operating a Labsphere URS-600 as their primary source for Cimel sky calibrations. I agreed with this view to the extent that common instrumentation leads to standardization of procedures and results. However, I also accept the point later made by Tom Eck, that use of the smaller source raises the workload at AERONET because each collimator has to be calibrated separately. My considered view is that the choice of light source is less important than the methods used to tie the spectral radiance output of the various labs together. I suggested that using a bandpass filtered Quantum Efficient Detector (trap detector) would provide a means of achieving this, in addition to the travelling Cimels already agreed on. Completion of the development of this detector was noted as an action item on me, with the plan that they become one of the calibration tools used to tie together the sources used across the Distributed Calibration Facility.

Summary
This workshop was of great value in raising the profile of aerosol activities in Australia, particularly in connection with the CSIRO AGSNet. I was pleased to be able to provide significant input in the area of instrument calibration, and look forward to closer international ties and greater application of aerosol data from the Australian network by the global aerosol community.