Comparison of Hobo Weather Station and Kestrel Drop

Comparison of Hobo Weather Station and Kestrel Drop Temperature Data: ATOC Sky. Watch - 29 Jan to 5 Feb 2019 John Cassano ATOC 4500

Objectives • Compare temperature observed by Hobo weather stations and Kestrel Drop sensors from a 1 week deployment at ATOC Sky. Watch • Under what conditions do these sensors report similar / dissimilar temperatures • Is a solar radiation error present in the observed temperature? • If so, how large is this error? • What conditions minimize / maximize this error? • What corrections and uncertainty estimates should be applied to these sensors for future analyses? • What is the mean difference between each sensor and a reference sensor? • How much scatter is present in the observations relative to a reference sensor?

Comparison of Hobo and Kestrel Sensors For Hobo sensors: • Most of scatter is between +/- 1 deg C • Afternoon positive temperature bias for T-RH sensor on Hobo 01 (and to a lesser degree for Hobo 02) For Kestrel Drop sensors: • Most of scatter is between +/- 2 deg C • But, Kestrel sensors have large positive bias in afternoon • Afternoon bias can be 10+ deg C • D 09 has inconsistent bias signal and appears to be unreliable

Comparison of Hobo and Kestrel Sensors vs D 01 • Comparing sensors to D 01 increases spread for Hobo obs to +/- 2 deg C and decreases spread for Kestrel sensors to +/- 1 deg C • Afternoon warm bias is still present for Drop sensors and is not the same between sensors suggesting small-scale differences in position are important • Afternoon cold bias for Hobo sensors consistent with Drop sensors having positive radiation error relative to shielded Hobo sensors • Recommend calculating offsets between sensors based on Kestrel Drop instead of Hobo sensors

Temperature Differences vs Wind Speed • Largest scatter between observations for light wind speeds • Scatter between observations greatly reduced for wind speed > 5 m/s • When well ventilated all sensors report similar temperatures

Temperature Differences vs Time of Day (UTC) • Reduced scatter between observations during overnight period (00 to 14 UTC) • Large scatter during daytime (14 to 24 UTC) • This suggests that the Hobo and Kestrel sensors are well suited for nighttime studies but may not be reliable for daytime analysis • These results suggest comparison between different sensors at night will produce scientifically useful results • Comparison between different sensors during the day will not be reliable

Temperature Differences vs Solar Radiation Note: These plots are for differences from Hobo 01 T • Increased scatter occurs for solar radiation > 100 W/m 2 for Kestrel Drops and > 150 W/m 2 for Hobo sensors • Little evidence for solar error on other Hobo sensors aside from Hobo 01 T-RH • Drop sensors have warm bias of up to 10+ deg C for solar radiation > 100 W m-2 • Time series plots shown previously indicated reduced positive bias on afternoon of 4 Feb • This day had strong winds which provided ventilation that reduced solar radiation errors in unshielded Kestrel sensors

Average Difference • For all data: • Average difference for Kestrel Drops is mostly less than 0. 4 deg C • Average difference for Hobo sensors is within +/- 0. 2 deg C • These results are consistent with the reported accuracy of the sensors (0. 5 deg C for Kestrel and 0. 2 deg C for Hobo) • For data filtered by solar radiation: • Differences for all sensors is reduced to < 0. 2 deg C (except Hobo 02 T-RH) • There is little change for no radiation or radiation < 100 W m-2 • Recommend calculating sensor corrections for solar radiation < 100 W m-2 and relative to D 01 sensor

Standard Deviation • For all data: • Standard deviation for Drop sensors is between 1. 5 and 2. 5 deg C for differences from Hobo • Drop standard deviation decreases to between 0. 5 and 1. 5 deg C for differences from D 01 • For data filtered by solar radiation: • Standard deviations mostly less than 0. 5 deg C for all sensors • These results suggest that ”nighttime” data will have an uncertainty of approximately 0. 5 deg C • This is the minimum meaningful temperature difference we can identify between any pair of sensors

Conclusions • Compare temperature observed by Hobo weather stations and Kestrel Drop sensors from a 1 week deployment at ATOC Sky. Watch • Under what conditions do these sensors report similar / dissimilar temperatures ØSmallest differences between sensors occur: ØAt night (00 to 14 UTC) ØWith solar radiation < 100 W m-2 ØWith winds greater than 5 m s-1 ØKestrel D 09 sensor is unreliable • Is a solar radiation error present in the observed temperature? ØYes for the Kestrel Drop sensors ØNo for the Hobo sensors except Hobo 01 T-RH • If so, how large is this error? ØKestrel Drop radiation error can be 10+ deg C • What conditions minimize / maximize this error? ØRadiation error is negligible for solar radiation < 100 W m-2 ØStrong winds (4 Feb) reduced but did not remove the radiation error

Conclusions • What corrections and uncertainty estimates should be applied to these sensors for future analyses? • What is the mean difference between each sensor and a reference sensor? ØMean differences are mostly less than 0. 2 deg C once observations are filtered for solar radiation < 100 W m-2 • How much scatter is present in the observations relative to a reference sensor? ØThe standard deviation of sensor differences is mainly less than 0. 5 deg C once observations are filtered for solar radiation < 100 W m-2

Conclusions: Recommended Corrections Relative to D 01 Sensor Correction (deg C) Hobo 01 T -0. 2 Hobo 01 T-RH -0. 2 Hobo 02 T-RH -0. 5 Sensor Correction (deg C) D 06 0. 3 D 11 0. 0 D 02 0. 0 D 07 0. 0 D 12 0. 0 D 03 0. 2 D 08 n/a D 13 0. 2 D 04 0. 1 D 09 -4. 8 D 14 0. 1 D 05 0. 1 D 10 0. 2 D 15 0. 2