NOTICE DATED NOVEMBER, 1999:

United States National Weather Service (NWS) upper-air stations using the Sippican, Incorporated VIZ-B2 radiosonde will implement additional coefficients in the algorithm used to compute relative humidity (RH). NWS supported stations in the Caribbean using this radiosonde will also implement these coefficients. The new coefficients improve the accuracy of RH measurements at high and low RH. Starting about October 15, 1999, NWS stations using the VIZ-B2 radiosonde began phasing over to radiosondes with the new coefficients. This activity will be completed by about December 1, 1999.

NWS UPPER-AIR STATIONS IMPLEMENTING
NEW COEFFICIENTS FOR COMPUTING RELATIVE HUMIDITY

 Sippican Carbon Hygristor Data Processing using H1 and H2

Introduction

The Sippican, Inc.,  VIZ-B2 radiosonde  employs three coefficients to process the relative humidity (RH)  sensor (i.e., carbon hygristor) data.  The three calibration factors are: lock-in, H1, and H2.  These factors provide the original primary calibration setting (lock-in), a new “H1” calibration coefficient to the high end of the RH range (values > 33% RH), and a new “H2” calibration coefficient to the low end of the RH range (values< 33% RH).  The processing algorithm uses these calibration factors to provide RH output that best matches the sensor calibration data.  The factors are required to adjust for process and material fluctuations built into the final sensor. 

Processing

The first of the three calibration factors is the lock-in value.  The lock-in is the sensor’s electrical resistance at 33% RH and 25^O C.  This is the original calibration factor (used for many years) and is the most fundamental of the factors supplied.   For any hygristor, the lock-in is established based on the original resistance read at cut-in (cut-in is the process of bringing the sensor resistance into predetermined range during production by physically cutting the conductive surface to adjust the resistance), and calibration data from a sample of hygristors from the same lot, run in a precision RH calibration chamber.  The use of the lock-in factor without the incorporation of the additional factors provides a good average response algorithm.  However the algorithm does not provide the most accurate response for all sensors produced.

The H1 factor is a value used to remove bias at high RH (values > 33%) found in the calibration data. There is no change from the nominal curve at the high RH end for sensors with H1 set to a value of one.  Hygristors with resistance versus RH higher than normal above 33% RH incorporate a lower H1 factor (less than 1.0) to accurately calculate RH.   Inversely, a higher H1 factor (greater than 1.0) is used to accurately calculate RH when the resistance versus RH curve above 33% RH has a lower resistance than normal.   The effect of H1 is weighted, with no effect at 33% RH and maximum effect at 100% RH (see equation 1).

The H2 factor is a ratio number used to remove any bias at low RH (values < 33%) found in the calibration data.  There is no change from the nominal curve at low RH end for sensors with H2 set to one.  Hygristors with resistance higher than the norm below 33% RH incorporate a higher H2 factor (greater than 1.0) to most accurately calculate RH.   Inversely, a lower H2 factor (less than 1.0) is used to accurately calculate the RH when the resistance vs. RH curve below 33% RH has a lower resistance than the norm.   The effect of H2 is weighted, with no effect at 33% RH and maximum effect at 0% RH (see equation 1).

H1 and H2 are optimized by averaging the absolute error of all 25^oC and –10^oC calibration data points for a given lot and then minimizing that average.

Examples

For B2 radiosondes with H1 and H2 factors optimized at H1=0.9 and H2=0.9 the accuracy improvement over the non-optimized algorithm (H1 and H2=1) is +2.64% RH at 10 % RH and –2.48% RH at 90% RH.

For B2 radiosondes with H1 and H2 factors optimized at H1=1.1 and H2=1.1 the accuracy improvement over the non-optimized algorithm (H1 and H2=1) is –2.77% RH at 10%  RH and +2.39% RH at 90% RH.

Conclusion

Sippican, Inc.,  has developed these algorithms and the related procedures to obtain the required calibration factors from RH chamber testing.  Chamber tests have shown that the RH calculations are improved by implementing the H1 and H2 calibration factors.

Resistance-to-RH Algorithm

The most important variable used in the resistance-to-RH algorithm is the ratio of sensor resistance, R_s, to sensor lock-in resistance, R_lock-in,

                                                      Ratio = R_s / R_lock-in.

R_lock-in is the sensor resistance at +25°C and at a RH with respect to water of 33%.  R_lock-in is specific to each sensor and is “locked in” during the manufacturing pro­cess to have a nominal value around 10 KΩ.  R_s has a relatively small dependence on the temperature, and historically it has been found that at 33% RH the response of the sensor to changes in temperature is particularly small.  This is the reason why it is the 33% RH that is “locked in”.  We therefore charac­terize the overall behavior of the sensor resistance using the value of Ratio instead of  R_s.

The main function in the resistance-to-RH algorithm is a family of curves (the quantity raised to the k^th power in Equation No. 1 below), with one member of the family for every temperature.  Each member describes the response of the sensor to RH at a particular temperature.  The following two equations have been found to accurately match this family of curves and so are used to convert sensor resistance into relative RH for all temperatures.  

Equation No.1

                                               6
                        RH =   A - (B /Σ D_k * [ H * g(T) * ln(Ratio) ]^k)          
                                               ^k=0

                                        where,

RH = calculated percent relative humidity with respect to water

A = A₁,  H * g(T) * ln(Ratio) ≥ -0.2

    = A₂,  H * g(T) * ln(Ratio) < -0.2

B = 69

D_k = D_1k,  H * g(T) * ln(Ratio) ≥ -0.2

     = D_2k,  H * g(T) * ln(Ratio) < -0.2

g(T) = a function of temperature given in Equation No. 2; this function specifies the member of the family for curves at temperature T

H = H₁,  Ratio ≥ 1  (constant in the calibration data with VIZ system software)

    = H₂,  Ratio < 1  (constant in the calibration data)

T = temperature in degrees C                                                                                                                          

Equation No. 2                                             

                                            g( T ) =  Σ C_k * T^k
                                                          k=0

where,                   

               C_k = a set of six calibration constants, each assuming one value when

               Ratio ≥ 1 and another when Ratio < 1

               T = temperature in degrees C.

The sensor response to RH at +25°C is particularly well documented and this curve is especially important.  The H and D_k coefficients in Equation No. 1 are fitted  to experimental data on this +25°C curve.  Because the g in Equation No. 2 equals 1.00 when T =+25°C,  Equation  No. 1 reduces to this best-fit curve at that temperature.  The C_k’s in Equation No. 2 are best-fit coefficients to experimental data at tem­peratures other than at +25°C.

Appendix

The carbon-based RH sensor response can be accurately matched using the following coefficients in Equation No. 1.:

Equation No. 1A                                                   Equation No. 1B                    

                H * g(t) * ln(Ratio) > -0.2                                       H * g(t) * ln(Ratio) < -0.2

        D₀ =  1.000E+00                                                  D₀ =  1.000E+00
        D₁ =  7.290E-01                                                   D₁ =  7.290E-01 
        D₂ = -5.580E-02                                                   D₂ =-5.580E-02
        D₃ =  7.480E-03                                                   D₃ =  7.480E-03
        D₄ =  1.010E-02                                                   D₄ =  1.010E-02 
        D₅ = 0.0                                                              D₅ =  0.0                                            
        D₆ = 0.0                                                              D₆ =  0.0                                            
        A = 1.02E+02                                                       A = 1.02E+-2                    
        B = 69                                                                 B = 69                                                

  The following coefficients are used in Equation No. 2:

 Equation No. 2A                                       Equation No. 2B

  R_s > R₃₃  (Ratio >1)                                  R_s < R₃₃  (Ratio < 1)

 C₀ =  7.885E-01                                         C₀ =  9.243E-01
  C₁ =  9.286E-03                                         C₁ =  3.059E-03
  C₂ = -2.462E-05                                         C₂ = -1.188E-06
  C₃ = -3.368E-07                                         C₃ =  0.0
  C₄ =  0.0                                                   C₄ =  0.0
  C₅ =  0.0                                                   C₅ =  0.0

NOTICE DATED DECEMBER, 1998:

United States National Weather Service (NWS) upper-air stations and three NWS supported station in the Caribbean will begin reporting the cloud data group in the World Meteorological Organization (WMO) upper-air coded messages at 00:00 UTC, February 1, 1999. NWS stations are not meeting the requirement for the cloud data group as stated in the WMO No. 306 Manual on Codes, Vol I.1, FM35 TEMP, Section 8. The list of stations that will report the cloud data is attached.

NWS UPPER-AIR STATIONS TO REPORT THE WMO CLOUD DATA GROUP
EFFECTIVE 00 UTC, FEBRUARY 1, 1999

 

 

Contiguous United States

WMO Index  & Station 

72659 Aberdeen, SD

72518 Albany, NY

72365 Albuquerque, NM

72363 Amarillo, TX

72764 Bismarck, ND

72318 Blacksburg, VA

72681 Boise, ID

72501 Brookhaven, NY

72250 Brownsville, TX

72528 Buffalo, NY

72712 Caribou, ME

74560 Central Illinois, IL

72208 Charleston, SC

72649 Chanhassen, MN

74494 Chatham, MA

72251 Corpus Christi, TX

74455 Davenport, IA

72261 Del Rio, TX

72469 Denver, CO

72387 Desert Rock, NV

72451 Dodge City, KS

72582 Elko, NV

72376 Flagstaff, AZ

72249 Fort Worth, TX

72634 Gaylord, MI

72768 Glasgow, MT

72476 Grand Junction, CO

74389 Gray, ME

72776 Great Falls, MT

72645 Green Bay, WI

72317 Greensboro, NC

72747 International Falls, MN

72235 Jackson, MS

72206 Jacksonville, FL

72201 Key West, FL

72240 Lake Charles, LA

72340 Little Rock, AR

72597 Medford, OR

72202 Miami, FL

72265 Midland, TX

72327 Nashville, TN

72305 Newport, NC

72357 Norman, OK

72562 North Platte, NE

72493 Oakland, CA

72215 Peachtree City, GA

72520 Pittsburgh, PA

72797 Quillayute, WA

72662 Rapid City, SD

72489 Reno, NV

72672 Riverton, WY

72694 Salem, OR
72572 Salt Lake City, UT

72293 San Diego, CA

72364 Santa Teresa, NM

78526 San Juan, PR

72230 Shelby County, AL

72248 Shreveport, LA

72233 Slidell, LA

72786 Spokane, WA

72440 Springfield, MO

72403 Sterling, VA

72214 Tallahassee, FL

72210 Tampa Bay Area, FL

72456 Topeka, KS

72274 Tucson, AZ

72558 Valley, NE

72402 Wallops Island, VA

72632 White Lake, MI

72426 Wilmington, OH

Alaska

70273 Anchorage

70398 Annette

70026 Barrow

70219 Bethel

70316 Cold Bay

70261 Fairbanks

70326 King Salmon

70350 Kodiak

70133 Kotzebue

70231 McGrath

70200 Nome

70308 St. Paul Island

70361 Yakutat

Caribbean

78583 Belize City, Belize

78954 Barbados

78384 Grand Cayman

Pacific Islands

91217 Guam, Mariana Islands

91285 Hilo, HI

91408 Koror, Palau Islands

91165 Lihue, HI

91376 Majuro, Marshall Islands

91765 Pago Pago, Samoa

91348 Ponape, Caroline Islands

91334 Truk, Caroline Islands

91413 Yap, Caroline Islands

The code group on cloud information will  be found between the 31313 and 51515 groups in the Part B or "TTBB" message and will be encoded as follows:

41414 N_hC_LhC_MC_H

Where,

N_h = Amount of all C_L cloud present or, if no C_L is present, the amount of all C_M cloud present in eighths of cloud cover. Code figures 0 through 9 and / are used to define cloud types as defined by a code table.

C_L = Type of low cloud present.                        

h = Height above surface of the lowest cloud seen as defined by a code table.

C_M = Type of mid-level cloud present.

C_H = Type of high-cloud present.

Reference: WMO No. 306 Manual on Codes, Vol 1.1, Sections A and B.

Back to the Upper-air Observations Main Page