© Crown copyright Met Office UK Met Office investigations into laser disdrometers Present Weather Trial at Eskdalemuir, Scotland: Winter 2007/8 Darren Lyth, UK Met Office, 29/11/08

© Crown copyright Met Office Contents This presentation covers the following areas Current UK automatic Present Weather (PWx) observation network Issues with current network Laser disdrometer trial Auto PWx code comparison with current system Using direct measurements from instrument Future analysis

© Crown copyright Met Office Current system: Vaisala FD12P/ PWx Arbiter Employed at 50 UK Met Office stations FD12P PWx output combined with output of other instruments used for elements of SYNOP report Output from FD12P, Eigenbrodt RS85 precipitation detector, Belfort Visiometer, LCBR, and air temperature thermometer fed into a PWx algorithm (known as the “Arbiter”) Arbiter adds value by making changes to the FD12P code (where applicable) System is a combination of bespoke software tailored to off-the-shelf sensors

© Crown copyright Met Office Issues with current system Slight precipitation events can be incorrectly reported due to instrument sensitivity issues, and contamination problems with current precipitation detectors Light drizzle or sleet events not accurately determined (Arbiter is able to reassign some erroneous mist reports as drizzle by use of Eigenbrodt) Automatic systems perform poorly in sleet, and Arbiter cannot effectively improve hit rate A single PWx code output gives little information on uncertainty and no raw information Both the PWx sensor and Arbiter make many assumptions rather than direct measurements – making traceability, calibration and maintenance difficult

© Crown copyright Met Office Present Weather Trial – Eskdalemuir, Scotland (Nov 2007 – May 2008) Disdrometers: Thies LPM OTT Parsivel Campbell Scientific PWS100 Supporting instruments: OTT Pluvio raingauge Lufft R2S Rain Sensor Mierij 205 Rain Detector Metek MRR-2 Rain Radar Gill Windmaster Ultrasonic Anemometer

© Crown copyright Met Office Laser Disdrometers Transmitter produces a flat, horizontal light beam, which the receiver converts into an electrical signal When a hydrometeor falls through the beam in the measurement area, the signal changes Amount of dimming is a measure of hydrometeor size, and along with the duration of the signal change, the fall velocity of the hydrometeor can be derived From this, a Present Weather code type can be reported

© Crown copyright Met Office Disdrometer Types Thies LPM: Low cost, and produces raw data output on drop sizes, fall speeds, etc., along with automatic PW code. Hydrometeor Measurement Specification: 0.16  8mm diameter range; 0.2  20m/s velocity range. Instrument outputs ppn rate, total ppn amount, SYNOP, METAR & NWS PW codes, as well as radar reflectivity. OTT Parsivel: Similar price to Thies LPM, and reports the same products. Hydrometeor Measurement Specification: 0.2  25mm diameter range; 0.2  20m/s velocity range. Campbell Scientific PWS100: Uses both disdrometer and forward scatter techniques to classify precipitation type, & analyse precipitation intensity, drop size distributions, and visibility range. Using four parallel light sheets in a structured detection volume (Doppler Anemometry), the PWS100 discriminates between polycrystalline precipitation and rain, by generating a “pedestal” in the received scatter signal. Hydrometeor Measurement Specification: 0.1  30mm diameter range; 0.16  30m/s velocity range

© Crown copyright Met Office PWx code distribution during trial: FD12P/ Arbiter

© Crown copyright Met Office PWx code distribution during trial: Laser disdrometers

© Crown copyright Met Office Scoring Performance: Snow From the HSS & POD scores, disdrometers are able to report snow codes more accurately than the current PW sensor, on more occasions. The Arbiters high MOSS score shows the benefit of using added sensors to reassign auto PW codes to agree more closely with the observer. For ‘single’ PW instruments (including disdrometers) insensitivity to light snow events results in low MOSS scores. The current PW sensor reports the highest percentage of reports during human observed snow as light snow, but also a large number of ‘no precipitation’ and light rain reports, resulting in a low MOSS score The Arbiter algorithm reassigned many of the current sensor’s ‘no ppn’ codes to ‘intermittent ppn’ Also general insensitivity for disdrometers during snow events (20% of ‘no precipitation’ reports, contributing to a low MOSS score). However, they also reported high numbers of snow codes correctly, which produced high HSS/ POD scores, comparable to the Arbiter score.  Disdrometers correctly reported fewer moderate snow codes, and slightly more light snow codes, during light snowfall events. Sensitivity of all instruments in reporting snow decreases with lighter snowfall “SNOW”HSSPODFARMO Skill Score Current (non-disdrometer) PW % Current PW + Arbiter % Average disdrometer score %

© Crown copyright Met Office Scoring Performance: Sleet “SLEET”HSSPODFARMO Skill Score Current (non-disdrometer) PW % Current PW + Arbiter % Average disdrometer score % All automatic instruments show a poor performance in accurate identification of sleet Disdrometers are 5-6 times more sensitive to sleet than either the current system or the Arbiter, however the disdrometer MOSS scores are lower than the Arbiter due to the number of reports of ‘no ppn’, indicating insensitivity to some sleet events. The greatest number of disdrometer reports in sleet events were for slight and moderate non-freezing rain (over one-third of reports) However, when disdrometers do report sleet, the observer on most occasions is also reporting sleet. Over 75% of reports of the current system during observed sleet were for slight and moderate non- freezing rain and ‘no ppn’ (the latter indicating a sensitivity issue during sleet), which resulted in a low MOSS score. The Arbiter reassigned most of these codes as intermittent rain. The Arbiter did not re- assign any auto codes as ‘no ppn’ which would indicate that the existing Eigenbrodt RS85 sensor is more sensitive to sleet than to snow. The current system misidentifies snow as sleet, resulting in a high FAR score.

© Crown copyright Met Office Scoring Performance: Light Drizzle/ Light Drizzle & Rain The scores for all sensors are generally poor for drizzle performance, but especially so for light drizzle events. This indicates either a lack of sensitivity in auto instruments to record a light drizzle code, or that software tolerances are set too high to record it. It may be advisable to incorporate an sensitive precipitation detector into the decision making process, to confirm that light precipitation is occurring. The Eigenbrodt RS85 performed well, enabling the Arbiter to report precipitation which it otherwise would not be able to do when relying on the FD12P precipitation detector alone (as 23% of FD12P reports in drizzle were for ‘mist’ or ‘code 0’). Sensitivity to light events was again an issue for certain disdrometers. However, the majority of the rest of the disdrometer PW reports are correctly assigned as light drizzle events (large number of auto codes 51 & 57 reported). In situations where the disdrometer is reporting light drizzle, it would appear that the sensor underestimates particle size, as in many of these cases, the human observer is reporting slight or moderate rain or rain/drizzle codes. However, the disdrometers show the best hit rate for any single instrument during light drizzle events. “LIGHT DRIZZLE”HSSPODFARMO Skill Score Current (non-disdrometer) PW % Current PW + Arbiter % Average disdrometer score %

© Crown copyright Met Office Scoring Performance: Rain The results for rain (for all automatic sensors) are better than for all other weather types examined. The precipitation detector on the current system shows some insensitivity to light rain events (with 10% of ‘no ppn’ or ‘mist’ reports). The vast majority of these reports are reassigned by the Arbiter, indicating that the Eigenbrodt RS85 at Eskdalemuir works well. The FD12P reports light rain on >60% of occasions. For the disdrometers, over one-third of reports agree with the observer reporting light rain. However, the next highest numbers of reports are for rain/drizzle or drizzle codes. When the disdrometers report light rain/ light rain showers, agreement with the observer is generally good, but with a minority of occasions that the observer reports slight drizzle (<5%). Misinterpretation of multiple coincident drops in the beam as one large particle, or wind affecting fall speed may be the reason for the anomalies. There is also evidence of moderate rainfall being misinterpreted as light rainfall, which may mean the disdrometers are undercounting larger rainfall droplets. Certain disdrometers again shows a sensitivity, or software threshold issue in light precipitation events, which extends also to rain. “RAIN”HSSPODFARMO Skill Score Current (non-disdrometer) PW % Current PW + Arbiter % Average disdrometer score %

© Crown copyright Met Office Scoring Performance: Hail Automatic systems are very poor at reporting hail Both the current system and Arbiter did not report any hail codes The Thies LPM did record a POD score of 0.02, with no misidentification of hail events

© Crown copyright Met Office Raw Disdrometer Output Analysis

© Crown copyright Met Office Example 1: Human observer reports light continuous snowfall (19/3/07)

© Crown copyright Met Office Example 2: Human observer reports light continuous snowfall (10/2/07)

© Crown copyright Met Office Work Still To Do: Analyse how high wind speeds affect PWx code determination of laser disdrometers Presenting the data to users (workshop planned for early 2009) Adding value to system: wind speed (flagging data for users in high wind speed regimes) CCTV profiling (e.g. using a Metek Micro Rain Radar to determine melting layer) Further user-led field trials to take place in 2009: 2 locations (Eskdalemuir, and a military site in the north of the UK)

© Crown copyright Met Office Thank You