1 / 17

Langley Calibration

Langley Calibration. Web interface a) Limiting the Lamp Calibration dates b) Creating a Langley Calibration factor from Mauna Loa data c) Calculating the Langley Factor on the fly New Langley Calibration method. Web Interface. Limiting the Lamp Calibration

Download Presentation

Langley Calibration

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Langley Calibration • Web interface a) Limiting the Lamp Calibration dates b) Creating a Langley Calibration factor from Mauna Loa data c) Calculating the Langley Factor on the fly • New Langley Calibration method

  2. Web Interface Limiting the Lamp Calibration Done, needs debugging http://reyes.nrel.colostate.edu:8080/UVB/da_queryUvLamp.jsf Creating a Langley Calibration factor from Mauna Loa data (table: langley_calibrations) Need to add login, needs debugging http://reyes.nrel.colostate.edu:8080/OPER/da_queryLangley.jsf

  3. Calculating Langley on the fly Needs to be done. Logic: First 8 weeks use langley calibration factor from Mauna Loa data Otherwise use daily vnaught created by new langley method java code already in place.

  4. New Langley Calibration method • All data has been reloaded using new method (1997 – May 2010) • Things to do Needs to be verified/examined (1997,2007 done) Holes need to be filled in when possible So far, all software bugs have been fixed but will continue when examining data Better error trapping Determination of good data/ qc codes have changed

  5. Steps in New Langley Calibration Method Step 1: Run la for every day store data in table optical_depths_la air mass is 2 -6 for VIS and 1.2/1.5 – 2.2/3 for UV AM data only New field added to table date_excluded (yes or no) Two sources: RunReloadVoGen.java (previous data) RunNightlyVoGen.java (nightly load)

  6. Previous loads: Reads tracking table for every site to get time period to run so will not span a serial number Get cosine corrected voltages for every day during time period determine by tracking table Creates la config file Checks qc codes/ sunny days to throw out data is needed Runs la and stores in table optical_depths_la Nightly Run: Runs every night after cosine corrected voltages have been loaded. Currently 7:20 am.

  7. Things still to do: What if site did not poll, when to run it? During loadfile.pl maybe? Serial Number logic. What if serial number changed during the day? Need to run twice one for old serial number and one for new serial number. Qc code usage, what is thrown out? Need to verify since qc codes have been updated. Better error trapping with descriptions and less debug statements

  8. Step 2: Trimming out extraneous la values Table: optical_depths_mean nightly_running_mean optical_depths_la Source: LoadVoRunningMean.java(previous data) RunNightlyVoRunningMean.java Logic: At a serial number start: Take first 3 values and create a mean/standard deviation Check the next value against the mean/standard deviation derived from first 3

  9. If (new value – mean value)/mean value > .5 THROW OUT VALUE else if (new value – mean value) > std dev * 2 just a possibility to throw out else keep value Continue updating the mean and standard deviation for 12 values only. If the second criteria above was selected then keep track for the next value. If you have 3 of those values you then want to keep the values. If you do not get three of those values you throw it out.

  10. If you threw value out, update column date_excluded in table optical_depths_la Store the values you kept with mean/standard deviation in table optical_depths_mean

  11. Still to do: Nightly logic for this. Deciding the 3 in a row logic. And also a new serial number logic. Need a table to store the 12 running mean values. When to run it? What about a site that was missed at night? What if no vnaughts for yesterday. Just keep previous running mean? And tag it along? Who is going to keep track if there is no vnaughts. What was the reason/ instrument failure, cloudy day etc. Email message/ what if discovered a problem and a flag is added/ need to rerun?

  12. Step 3: Creating the daily vnaught Table: daily_vnaughts_mean Source: DailyRegrVoLoader.java VoPredictorLinearRegr.java RegressionPredictor.java Logic: PASS 1: Input start date and end data(one day) gets voltage intercept from optical_depths_mean table. Goes back 3 months and forward 3 months in previous data. Trims out 2 standard deviations from least squares regression. Must have a minimum of 5 to continue. If none then use yesterdays prediction and carry it along for 2 months.... More than 2 months -9998....

  13. If more than 5 then get a linear regression(dividing by 1000) to make values more reasonable. At end multiply by 1000. Get predictedValue PASS TWO: Use a weighted sum of values to try and eleviate the noise. Num of days = date wanted - optical depth start sum = value/(num of days*num of days) valsDaySum = 1/(numofdays*numofdays) + sum predictedValue= sum/valsDaysSum Store value in daily_vnaughts table.

  14. To do: Implementing nightly version. When to run. Serial number logic.. Problems when fails. Cannot go forward 3 months. Go back 6 instead?

More Related