# Post-calibration of YSI chlorophyll A Gang of N Production 27 July 2005 Bill Romano – MD DNR Elgin Perry – Statistics consultant Beth Ebersole – MD DNR.

## Presentation on theme: "Post-calibration of YSI chlorophyll A Gang of N Production 27 July 2005 Bill Romano – MD DNR Elgin Perry – Statistics consultant Beth Ebersole – MD DNR."— Presentation transcript:

Post-calibration of YSI chlorophyll A Gang of N Production 27 July 2005 Bill Romano – MD DNR Elgin Perry – Statistics consultant Beth Ebersole – MD DNR Marcia Olson - NOAA

Elgin and I considered four methods Arithmetic mean ratio First subtracting 0.03 x turbidity from YSI chlorophyll then using a temperature adjustment (VIMS) Using temperature and turbidity Temperature only

Data used in the analyses 522 station and time matched pairs of extractive and YSI chlorophyll data Data set was split into calibration and validation data sets using a random number SAS function The first 261 records were assigned to the calibration data set

Arithmetic mean ratio RatioAdj = Extractive chl-a / YSI chlorophyll Calculate the mean ratio for each station Multiply the station mean ratio by YSI chlorophyll from the validation data set Calculate the root mean square error using the difference between post calibrated and extractive chlorophyll from the validation data set RMSE across all stations is 21.17

Turbidity adjusted data Subtract the YSI recommended turbidity interference factor (0.03 μg/L per NTU) from each chlorophyll reading Calculate the log ratio (logCE – logCFtc) Model the log ratio as a function of temperature Post calibrate using predicted ratio and turbidity adjusted YSI chlorophyll RMSE across all stations is 20.79

Temperature and turbidity adjustment Calculate the log ratio (logCE – logCF) Model the log ratio as a function of temperature and turbidity Post-calibrate using the predicted ratio and YSI chlorophyll RMSE across all stations is 20.43

Temperature only adjustment Calculate the log ratio (logCE – logCF) Model the log ratio as a function of temperature Post-calibrate using the predicted ratio and YSI chlorophyll RMSE across all stations is 20.96

Does turbidity contribute that much to the model? The p-value (type I and type III sums of squares) is 0.17 for the turbidity coefficient in the temperature and turbidity model The turbidity coefficient in: CHLa pred = 3.230 + 1.132 x (chl YSI ) – 0.015 x (turb), is half what YSI recommends (-0.03) Should we use it?

How do extractive and YSI chlorophyll compare? A comparison of log ratio (CF – CE) chlorophyll values using the Wilcoxon test indicates that many stations are significantly different Of the twenty-four station differences, only one was positive, so extractive chla exceeds YSI chlorophyll One would expect the opposite result, because the sonde provides a measure of total chlorophyll

Mean of extractive and YSI chlorophyll StationYSI chlchlap-value CCM006919.8420.270.1187 CHE034810.1915.270.1157 CTT000127.7334.650.0775 FRG000211.0013.430.0006 MDR003818.4624.270.0068 MTI00153.604.830.9442 PXT03116.107.290.3258 PXT04555.666.740.9312 SEV0116 (+)27.8926.210.4903 TRQ008819.6731.950.0020 TUV002124.0730.320.0002 WXT00134.164.420.5112

Mean of extractive and YSI chlorophyll StationYSI chlchlap-value XCF902922.9223.820.3484 XCH80978.4210.680.0413 XDE458716.3417.600.4307 XDM448649.4970.930.0225 XED069420.8256.72<0.0001 XHE197316.1822.62<0.0001 XHF371927.7130.420.0191 XIH00778.006.750.1074 XJF42896.4810.020.0195 XJG27187.4513.40<0.0001 XJG43375.697.760.0006 XJG70355.6412.03<0.0001

Simple ratio adjustment factors differ from station to station StationRatio CCM00690.928 CHE03481.203 CTT00011.424 FRG00021.421 MDR00381.344 MTI00150.956 PXT03111.002 PXT04551.195 SEV01160.842 TRQ00881.503 TUV00211.229 WXT00130.992 (see LSD output)

Simple ratio adjustment factors (continued) StationRatio XCF90291.247 XCH80971.451 XDE45871.161 XDM44861.496 XED06941.713 XHE19731.864 XHF37191.223 XIH00770.834 XJF42891.600 XJG27181.696 XJG43371.363 XJG70351.966 (see LSD output)

Observations with RSTUDENT greater in absolute value than 2 may need some attention. Are some values too influential?

Large values of DFFITS indicate influential observations. A general cutoff to consider is 2. Influential data points?

- Ratio Method

Download ppt "Post-calibration of YSI chlorophyll A Gang of N Production 27 July 2005 Bill Romano – MD DNR Elgin Perry – Statistics consultant Beth Ebersole – MD DNR."

Similar presentations