Hazy, stagnant air in Hokkaido

We will launch our 3rd ozonesonde today, Monday, 10 August 2009.  Weather conditions this year have not been so favorable for sampling air from China.  For the first week, high pressure in the Sea of Japan blocked transport from the mainland.  Now a low to the NE, a typhoon to the S, and a high to the E continue to suppress transport from the mainland.  A picture from Friday, 7 August 2009 shows just how hazy it has been here, as the air mass remains in place over Hokkaido day after day, recirculating the air mass.  On the plus side, we’re getting a good chance to measure the local contribution to the pollution profile.

Hazy skies over Hokkaido on Friday, 7 August 2009. Unfortunately, this haze results from a stagnant air mass that has persisted for days.

First 2009 Launch

On Wednesday, 5 August, we launched our first dual sonde of 2009.  The weather was mostly clear and sunny, with temperatures in the mid to upper 20’s (C) — a perfect day for our first launch this year.  We will be launching 10 of the dual sonde instruments this year between now and the end of August.  The data are available on our project web site.  Here’s a photo of the launch team

Hokkaido University team with VU students for 1st launch 2009.

at Hokkaido University, including Valparaiso University students Nathan Kellams (physics) & Ted Pietrzak (meteorology).  And here’s a shot of the balloon just as it leaves the ground.

A view of the first launch from the rooftop observatory at Hokkaido University.

Back in Japan

We have returned to Japan for part 2 of our study.  This year, two undergraduate students from Valparaiso University have accompanied me to help with the balloon launches.  Our plan for this year is to launch 10 dual SO2/O3 sondes during the month of August.  This year will serve as the “control” case for our study, while last year’s launches were done while China was implementing pollution controls in and around Beijing.

We have set up a new page on our project website for this year’s data.  Last year’s data can be found on this web site.  We have improved our trajectory forecast model this year.  For the forward trajectories from Beijing, we start 3 days in the past and run to 1 day in the future.  For the first 2.5 days, the calculations used NCEP reanalysis meteorological data, while for only the last 1.5 days, the calculations use NCEP forecast meteorological data. This should improve the accuracy of our forecasting of air masses from Beijing to Hokkaido.

Our launches will begin during the week of Aug. 3 – 9.

I will also be posting updates on our research progress during the spring, while I was not attending to this blog.  In particular, I will comment on the Special Session at the Spring AGU Meeting in Toronto (see page 50 of the PDF document) that dealt with pollution in and around Beijing, China before, during, and after the Olympics.  You might be interested in a recent paper that came out in AJP by Wang et al.

Stay tuned for more.

Outflow from Beijing: 1998 – 2008

Using the NASA Goddard Kinematic Trajectory model, I have examined flow out of the Beijing region during the month of August from 1998 – 2008 in an effort to put this year’s outflow (see earlier post) in context.  You will find PDF files on the project website summarizing the distribution of the outflow.  The contours indicate the number of parcel-quarter days found in each grid box.  Here’s a link to the outflow from Beijing in the Boundary Layer (< 2.5 km) and  another in the Lower Free Troposphere (2.5 – 5.0 km).

More details of the model: 

NCEP data from 1998 – 2001 are once daily; from 2002 – 2005 are twice daily; from 2006 – 2008 are four times daily.

Beijing box is 5 deg. lat by 8 deg. lon centered on 38.5 N, 115.5 E, with parcels every 0.5 deg. lat. and 0.5 deg. lon.

Gridded map in PDF files runs from 25 N ot 50 N and 105 E to 150 E, with a 1 deg. lat. by 1.5 deg. latitude grid cell.

Observations:

In the boundary layer, flow from China to Japan from 1999 – 2002 was relatively weak, with few areas of Japan experiencing much influence from air over Beijing.  In 1998, however, much of Honshu and all of Kyushu were influenced by Beijing air.  In particular, Northern and Southern Honshu and Kyushu saw substantial flows from Beijing.  In the lower free troposphere, when flows to Japan are strong (1998, 2000, 2003, 2006, 2007), Northern Honshu sees the most air from Beijing.

In the lower free troposphere, 1999, 2000, 2001, 2003, 2005, and 2008 are years with relatively weaker influences in Hokkaido from Beijing.  In the boundary layer, the weaker years are 1999, 2000, 2001, and 2008.  Thus, analysis of this summer’s pollution data need take into account the fact that outflow from China to Japan was relatively weak compared to previous years within the last decade.

Chinese pollution outflow movies

In my last post, I discussed and showed the summary statistics for trajectories out of Beijing and Shanghai during August 2008.  Here, I now add links to the quicktime movies:  Beijing outflow; Shanghai outflow.

The Beijing movie shows several periods during which boundary layer air over Japan likely had origins in China.  31 July – 1 August — most of Hokkaido and Honshu see air from China between 2 and 4 km, some of which may have become entrained in the afternoon mixed layer over Japan, influences surface ozone concentrations.  We will only be able to confirm the influences of such air masses 2 years from now, when all the Japanese surface monitor data becomes public.  2 – 3 August — most of Hokkaido.  3 – 5 August — lower Honshu.  6 – 7 August — Hokkaido and upper Honshu, here with air below 2 km, and on this occasion, we saw elevated ozone in Sapporo, both at the surface and aloft. 10 – 11 August — Hokkaido, above 2 km.  15 August — upper Honshu and Hokkaido from 2 – 6 km.  17 – 21 August — Kyushu and Honshu from 2 – 4 km.  23 – 24 August — nearly all of Japan.  This makes about 9 days in total on which the influence of air from Beijing may have been detectable in Hokkaido, although most of the events brought the air over at 2 – 4 km altitude, the one notable exception being the 6 – 7 August case.

The Shanghai movie also shows several periods of influence over Japan as well:  1 – 6 August — a prolonged period during which the 5-day forward trajectories were over Japan, with many below 2 km.  Early on, the air mass covered from central Honshu north to Hokkaido, mostly above 4 km.  Later on, the air mass covered lower Honshu below 3 km. 15 – 20  August — another wave of air covers most of Japan, from 1 – 4 km.  The further north you go in Japan, the higher the air mass tends to be. 22- 27 August — again, nearly all of Japan appears to have some influence from the Shanghai region, with many of the trajectories arriving below 2 km altitude.  Southern Japan tends to get most of this outflow.

So, the trajectory study would seem to indicate that influences from both the Shanghai region and the Beijing region can be found on the air mass over Japan from the surface up to 4 – 5 km during August.  About half the days in August saw the Shanghai air mass over some part (and frequently most) of Japan.  Also for nearly half the days in August, the Beijing air mass was over some part of Japan, most frequently in the northern part.

The next step is to mine the surface data (two years from now) and the balloon data to identify influences of Chinese pollution.  We can start the analysis of the latter now.

To where did the air go?

To answer this question, I ran Mark Schoeberl’s NASA GSFC trajectory model (Schoeberl and Sparling, 1995) in a kinematic mode, initializing a grid of parcels every 0.5 deg. latitude by 0.5 deg. longitude around Beijing and Shanghai every 500 m vertically from 500 m to 3.0 km.  The Beijing box ran from 36 – 41 N and 111.5 – 119.5, while the Shanghai box ran from 29 – 33 N and 117 – 123 E.  You can see the boxes on my google map here. The model was run in forward mode using the NCEP reanalysis meteorological fields.  Trajectory model output was produced every 6 hours for each 5-day forward run.  A grid from 25 – 50 N and 105 – 150 E with 1 deg. latitude by 1.5 deg. longitude boxes was produced, and counts of the frequency with which each grid box contained air parcels over the period 1 – 31 August 2008 was determined.  Below, I show the plots of the plumes from Shanghai and Beijing below 2.5 km and from 2.5 – 5.0 km.

 

Beijing plume (0.1 - 2.5 km) for August 2008.

Shanghai plume (0.1 - 2.5 km) for August 2008.

 

Beijing plume (2.5 - 5.0 km) for August 2008.

 

Shanghai plume (2.5 - 5.0 km) for August 2008.

At the lower levels, both plumes seem to impact Southern Japan most frequently.  At the higher levels, the plume is dispersed over a wider area, and impacts in Japan are shifted further north.  The Beijing plume seems to center on northern Honshu this year at the upper levels, while the Shanghai plume remains focused on southern Japan (southern to central Honshu).

To better see the apportionment of air arriving in Japan from these two important source regions, a ratio was taken between the air Beijing frequency in each box and the Shanghai frequency.  It should be noted at this stage that the Beijing box has ~60% more particles, which results in a ratio of 1.6 if influences from the two regions are identical.  Below are maps of the ratios for both the lower (0.1 – 2.5 km) and upper (2.5 – 5.0 km) air masses.  The thick lines are the 1:1 lines.  Contours are shown for ratios of 2.0, 3.0, 4.0, 5.0, 10.0.  Negative numbers indicate the dominance of the Shanghai plume while positive numbers indicate the dominance of the Beijing plume.

 

Ratio of the Beijing to Shanghai plume frequencies (0.1 - 2.5 km) for Aug. 2008.

 

Ratio of the Beijing to Shanghai plume frequencies (2.5 - 5.0 km) for Aug. 2008.

Not surprisingly, at the lower levels, southern Japan tends to be dominated by the Shanghai plume while northern Japan tends to see more of the Beijing air.  At the higher levels, the dominance of the Shanghai plume is even greater over southern Japan, and its influence is equal to or stronger than that of the Beijing plume throughout Honshu.  Important for this study is that the Beijing plume tends to dominate the Shanghai plume in Hokkaido (where we took our measurements this summer).

The next step will be to identify individual days on which the influence of the Beijing plume dominated over Japan, and to identify if any of the ozone features observed in our ozonesonde data for August 2008 can be linked to plumes from either of these major industrial centers in China.

Kasatochi eruption simulation

I have recently run Mark Schoeberl’s NASA Goddard Trajectory Model in kinematic mode using the NCEP reanalysis data for the meteorological fields to simulate the spread of the SO2 cloud from the eruption of Mt. Kasatochi.  Click here for the simulation.  The model was run with 0.025 day time-steps forward from 7 August 2008 through 22 August 2008.  I’ve injected parcels on a 2 deg. lat by 4 deg. lon grid centered on Kasatochi (52.18 N, 175.51 W) from 12 – 15 km on 7 – 8 Aug, then from 6 – 12 km on 8 – 9 Aug, based on observations posted on the Alaska Volcano Observatory web site.  While the simulation is not perfect, it does match many of features seen in the OMI SO2 data.

Another data reprocessing

I caught a couple of errors in the plotting routine for the 7 flights with SO2 data.  I have since reproduced the plots and IONS/INTEX-B formatted data files that appear on the project web site.  The problems in the originals were the following:  1) the potential temperature scale that appeared at the top of the SO2 plots was off; and 2) the ozone data was from the unfiltered sonde unit and had not been corrected for the SO2 measurement — thus, it was showing [O3] – [SO2] instead of [O3].  Both problems have been corrected.

Data reprocessing and update

I have now had a chance to go through each and every ozonesonde flight from Sapporo this past August, reprocessing the data to correct for the following:

1) pressure registration errors — due to a bias in the pressure reported by the Vaisala RS80-15N radiosondes, a disagreement that increases with altitude between the pressure altitude and the GPS alittude grows.  The problem has been corrected on several flights, for which pressure of up to a few hPa were observed.

2) background current — usually the adjustments make only a fraction of a ppb difference in the mixing ratios, but when integrated through the whole column, can make a difference of a few DU.  Ozonesonde columns were compared with Microtops and OMI column data and background currents adjusted to provide improved agreement.  On dual flights, the relative background currents of the two instruments is an important factor in the magnitude of the difference in the filtered vs. unfiltered ozone profiles.  Since the difference of these two is used to compute SO2 concentrations, the background currents are important.  While background currents are measured in the laboratory before flight, it is believed that the high ozone exposure preformed to demonstrate good agreement between the filtered and unfiltered sondes prior to launch may have shifted the background current somewhat.  Again, the background currents have been examined in the context of the resulting SO2 profiles, the independent SO2 surface measurements from Sapporo, the sonde total columns (from both instruments), the SO2 total column, the microtops total column ozone, and the OMI SO2 and ozone columns.

3) Erroneous wind data on descent — due to an error in the STRATO-Dual code used for the original measurements, the wind data on descent were erroneous.  The latitude/longitude/altitude registration was just fine.  Thanks to Holger Volmel, this error has been corrected and the flights rerun through a new version of STRATO that has corrected the problem.

I have also expanded the posted data files to include an estimated column SO2 and estimated SO2 concentrations (at the ppbv level) in the tropopshere (but no higher than 15 km altitude).  As mentioned, these estimates come from differencing the ozone concentrations recorded by the two sondes:  one with the SO2 filter and one without.  The resulting profiles are then examined individually to remove any detectable remaining bias.  After correcting for the bias, the data are multiplied by an efficiency factor as determined in laboratory tests at Hokkaido University (1/0.87 — see earlier post regarding SO2 filter tests).  SO2 concentrations have been rounded off to the nearest 1 ppb, and negative values have been allowed to provide any potential data user with a sense of the variability of this differential measurement.  While I am quite pleased with the resulting profiles, users should be aware of the limitations of the data.  The data are probably more reliable if smoothed vertically.  Given a 20 – 25 second time constant for ozonesonde instrument response, I would recommend 1 minute (~300 m) smoothing of the SO2 data.  I have NOT smoothed the data that appear in the data files on the project website — another reason to leave the negative values alone — in order to allow any potential user to process the data as he/she wishes.  If you would like to work with or end up wanting to publish and results that use this data set, please contact me first.

Below is a summary, flight by flight, of the reprocessing details:

Flight SE033: pressure correction of 1.2 hPa

Flight SE034:  pressure correction of 1.3 hPa

Flight SE035:  pressure correction of 1.3 hPa

Flight SE036:  adjust primary sonde background from 0.027 to 0.03 (= prep background #1); adjust secondary sonde background from 0.032 to 0.03 (=prep background #1); pressure offset +0.3 hPa; SO2 maximum altitude at 13.0 km; SO2 zero range for bias determination = [3, 8]; made secondary sonde ozone between 13 and 14 km the ozone reading, as the primary sonde shows a large positive excursion that is not seen in the secondary sonde nor by either on descent.

Flight SE037:  background #1 on secondary sonde was reset from 0.04 to 0.02 as per prep sheet; ozone background for secondary sonde set to 0.025 from 0.037; SO2 max alt. at 11.9 km; SO2 zero range = [1.5, 4]

Flight SE039:  pressure offset of +1.8 hPa; background #1 on secondary sonde was reset from 0.05 to 0.03 as per prep sheet; ozone background for secondary set to 0.035 from 0.046; SO2 max. alt. at 15 km; SO2 zero range = [4,8]

Flight SE040:  pressure offset of +1.7 hPa; ozone background for secondary set to 0.03 from 0.032; SO2 max. alt. at 12.5 km; SO2 zero range = [4, 7]

Flight SE041:  pressure offset of +2.4 hPa; ozone background for primary set to 0.025 from 0.022; ozone background for secondary set to 0.03 from 0.035; set ozone data to bad flag for primary ozone between 3.974 and 4.037 km and for secondary ozone between 3.974 and 4.190 km; SO2 max. alt. at 14.3 km; SO2 zero range = [ 3, 8]

Flight SE042:  pressure offset of +1.2 hPa; ozone background for primary set to 0.020 from 0.03; ozone background for secondary set to 0.037 from 0.047; SO2 max. alt. at 11.0 km; SO2 zero range = [3,6]

Flight SE043: ozone background for primary set to 0.03 from 0.024; ozone background for secondary set to 0.03 from 0.026; SO2 max. alt. at 15.0 km; SO2 zero range = [11,14]

The new data appear on the project data site.

Frontier Research Center for Global Change

I arrived for my first day of work in Yokohama on Wednesday.  I will spend the next two months here analyzing the data we gathered in Sapporo as well as the NASA satellite data related to pollution in East Asia.  I am grateful to my hosts here who had already prepared a desk and internet connection for me prior to my arrival, so my transition to this work environment has been quite smooth.

My initial focus will be on reprocessing the Hokkaido ozonesonde data using the new STRATO software from Holger Volmel — this to correct errors with the winds reported on the descent of the GPS sondes.  I will also try to finalize my SO2 product, implementing the correction for the filter efficiency.

On Oct. 23rd, I will be giving a seminar to the group here to provide an overview of my research program.  A larger community will arrive at FRCGC the following week for a conference at which I will present a poster on the same subject.  So, I have much to do over the next couple of weeks.  I will continue to update my blog as I make progress on my data analysis.