Reports

Contents

Title: Water Level sensor calibration with staff gauge
Date:2018-04-24 - 2020-10-05
Data File: WLcalib_SW.csv
WLcalib_MB.csv
WLcalib_WP.csv
WLcalib_EP.csv
WLcalib_EE.csv
Refers to:SW, WP, MB, EE, EP, sn 20010375, sn 70011229, sn 20010770, 70011236, sn 20010026, sn 70011234, 20010769

We installed metric staff gauges at our 5 wetland sites in spring/summer 2018. These staff gauges are bolted to the tower scaffolding, so their depths should (theoretically) not change through time, as long as the scaffolding does not move. The staff gauge water level was recorded at each site visit.

I used a linear regression to calculate an offset between the staff gauge and the continuous water level sensor (CS450 or CS451). This helps interpret the continuous water level through its multiple shifts in position. Once the water level sensor is benchmarked against the staff gauge, it's a simple offset to any other point in the wetland.

The given conversion from psi to cm is 70.41 cm/psi. In most cases the slope of the calibration was within 5% of the given conversion, so I mainly focused on the offset.

Conclusion:

I have updated the water level sensor offsets for these 5 sites in the smap and range&equation pages.

SMAP: https://nature.berkeley.edu/biometlab/bmetdata/equipment.php?screen=smap

Range & Equation: https://nature.berkeley.edu/biometlab/bmetdata/sfields.php?scrn=rangeeq

For Mayberry, West Pond, and East End, I used this calibration with the staff gauge to benchmark earlier water level sensor offsets. I used field and equipment notes, and in some cases the MATLAB met scripts, to track how the sensor height has changed through time.

For the few sites that have water level sensors but do not have staff gauges (Bouldin Corn, Sherman Wetland Temp Tower), I had to calculate the sensor offsets solely based on field/equipment notes.

Site Calibration start Calibration end Offset from staff gauge (cm)
Regression equation Regression R2 Regression n
Sherman Wetland 2018-07-1 2018-10-18 -8.4 y=0.99-8.4 0.98 8
2018-11-01 2019-06-04 26.3 y=1.01+26.3 0.99 16
2019-06-20 2020-06-16 38.5 y=1.02+38.5 0.98 22
Mayberry 2018-05-16  2020-03-11 9.5 y=0.95+9.5 0.96 30
  2020-04-03 2020-09-10 5.3 y=0.98+5.3 0.99 13
West Pond 2018-07-12 2020-10-05 -35.7 y=1.01-35.7 0.97 49
East Pond 2018-04-24 2020-01-30 -58.9 y=1.00-58.9 0.97 45
East End 2018-05-16 2018-08-23 -65.0 y=0.96-65.0 0.99 6
2018-09-18 2019-09-10 -26.8 y=1.01-26.8 0.72 18
2019-09-18 2020-05-19 17.5 y=0.92+17.5 0.76 12

 

Sherman Wetland

Figure 1. Time series of continuous water level sensor and manual staff gauge measurements at Sherman Wetland. There were 3 time periods for this calibration: 2018-07-11 to 2018-10-18, 2018-11-01 to 2019-06-04, and 2019-06-20 to 2020-06-16. One more time period should be added in the future, but we don't have enough data yet for a confident calibration: 2020-09-10 to present. (Water level sensor was broken 2020-07-30 to 2020-08-28.)

Regression Data

Residuals

Figure 2. Linear regression of staff gauge measurements and water level sensor data at Sherman Wetland. The sensor data has been converted to cm using the given 70.41 psi/cm conversion factor. There are 3 different regressions, 1 for each period of the whole calibration period. The regression slopes are within 1% of 70.41.

 

Mayberry

Figure 3. Time series of continuous water level sensor and manual staff gauge measurements at Mayberry. There are 2 time periods in this calibration: 2018-05-16 to 2020-03-11 and 2020-04-03 to 2020-09-10.

Regression Data

Residuals

Figure 4. Linear regression of staff gauge measurements and water level sensor data at Mayberry. The sensor data has been converted to cm using the given 70.41 psi/cm conversion factor. There are 2 different regressions, 1 for each period of the whole calibration period. The regression slopes are 2-5% different from 70.41.

 

West Pond

Figure 5. Time series of continuous water level sensor and manual staff gauge measurements at West Pond. Amazingly, this sensor has not been moved since 2012, so there is only 1 time period for this calibration.

Regression Data

Residuals

Figure 6. Linear regression of staff gauge measurements and water level sensor data at West Pond. The sensor data has been converted to cm using the given 70.41 psi/cm conversion factor. There are 2 different regressions, 1 for each period of the whole calibration period. The regression slope is within 1% of 70.41.

 

East Pond

Figure 7. Time series of continuous water level sensor and manual staff gauge measurements at East Pond. The East Pond tower scaffolding was adjusted on 2019-09-10. This only affected the offset in the calibration by 1-2cm, so for simplicity I decided to ignore the scaffolding adjustment and treat this whole calibration period as one time period.

Regression Data

Residuals

Figure 8. Linear regression of staff gauge measurements and water level sensor data at West Pond. The sensor data has been converted to cm using the given 70.41 psi/cm conversion factor. There is only 1 calibration period. The regression slope is within 1% of 70.41.

 

East End

Figure 9. Time series of continuous water level sensor and manual staff gauge measurements at East End. There were 3 distinct time periods: 2018-05-16 to 2018-08-23, 2018-09-18 to 2019-09-10, 2019-09-18 to 2020-05-19. One more time period should be added in the future, but we don't have enough data yet for a confident calibration: 2020-06-09 to present.

Regression Data

Residuals

Figure 10. Linear regression of staff gauge measurements and water level sensor data at East End. The sensor data has been converted to cm using the given 70.41 psi/cm conversion factor. There are 3 different regressions, 1 for each period of the whole calibration period. The regression slopes are 1%, 4%, and 8% different from 70.41. However, the larger differences also have more scatter than other regressions, so I think it would be best to keep the given 70.41 cm/psi conversion.