The Impact of Unique Meteorological Phenomena Detected by the and ARS Micronet on Automated Quality Control Christopher A. Fiebrich and Kenneth C. Crawford Oklahoma Climatological Survey, Norman, Oklahoma

ABSTRACT

To ensure quality data from a meteorological observing network, a well-designed quality control system is vital. Automated quality assurance (QA) software developed by the Oklahoma Mesonetwork (Mesonet) provides an effi- cient means to sift through over 500 000 observations ingested daily from the Mesonet and from a Micronet spon- sored by the Agricultural Research Service of the United States Department of Agriculture (USDA). However, some of nature's most interesting meteorological phenomena produce data that fail many automated QA tests. This means perfectly good observations are flagged as erroneous. Cold air pooling, " poking," , mesolows, heat bursts, variations in snowfall and cover, and microclimatic effects produced by variations in vegetation are meteorological phenomena that pose a problem for the Mesonet's automated QA tests. Despite the fact that the QA software has been engineered for most observa- tions of real meteorological phenomena to pass the various tests—but is stringent enough to catch malfunctioning sensors—erroneous flags are often placed on data during extreme events. This manuscript describes how the Mesonet's automated QA tests responded to data captured from microscale meteorological events that, in turn, were flagged as erroneous by the tests. The Mesonet's operational plan is to cata- log these extreme events in a database so QA flags can be changed manually by expert eyes.

1. Introduction and personnel at the central processing site for the in Norman, Oklahoma. The Oklahoma Mesonet, funded and maintained The task of manually inspecting each datum from by the state of Oklahoma, is an environmental moni- 157 automated stations, each recording observations at toring network of 115 stations deployed across the 5-min intervals, is insurmountable. Hence, automated state. The United States Department of Agriculture quality assurance (QA) software is an essential tool. and its Agricultural Research Service (ARS) funds the For each datum, the Mesonet's automated software operation of a Micronet of 42 stations in the Little generates a QA flag that indicates the quality of each Washita Watershed in southwest Oklahoma. Data observation. However, some of nature's most inter- from both networks are quality assured by software esting meteorological conditions provide data that fail many QA tests. As a result, some good observations are flagged as erroneous. The purpose of this manuscript is to illustrate how Corresponding author address: Christopher A. Fiebrich, Okla- automated quality control procedures can fail when homa Climatological Survey, 100 E. Boyd St., Suite 1210, microscale phenomena are detected by meteorologi- Norman, OK 73019. E-mail: [email protected] cal observing networks. A remedy for this problem In final form 27 April 2001. involves the use of a database to catalog unique me- ©2001 American Meteorological Society teorological events. The meteorological phenomena

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC that create special problems for automated QA soft- unusual meteorological events. In addition, the use of ware that will be discussed include the following: data quality flags to denote potential data problems is discussed in Snyder and Pruitt (1992). • cold air pooling and "inversion poking"; • mesohighs and mesolows; 1) RANGE TEST • heat bursts; The range test is an algorithm that determines if • snowfall and snow cover; and an observation lies within a predetermined range. The • microclimatic effects produced by variations in allowable ranges are based on sensor specifications vegetation. and annual extremes in Oklahoma; each pa- rameter has a unique set of limits (Table 1). If a da- tum is observed outside of the allowable range, it is 2. The Oklahoma Mesonet's quality flagged with a failure flag. assurance system 2) STEP TEST The automated quality control performed on Mesonet The step test uses sequential observations to de- and ARS data is one part of an extensive QA system. termine which data represent unrealistic "jumps" dur- Together, four components compose the Mesonet's ing the observation time interval for each parameter. QA system: 1) laboratory calibration and testing, 2) on- Observations that exceed the maximum allowed step site intercomparison, 3) automated QA, and 4) manual (Table 1) receive a warning flag. The thresholds for QA. During laboratory calibration and testing, all sen- the step test were determined through repeated test- sors are calibrated in the Mesonet lab to validate or im- ing of the algorithm on Mesonet data. prove upon the calibrations sent from the instrument manufacturer. Next, through on-site intercomparisons, 3) PERSISTENCE TEST instruments deployed to stations across the state are The persistence test analyzes data on a calendar day periodically compared (on average, once a year) to en- basis to determine if any parameter underwent little sure accurate performance by the sensors. Automated or no variation. If the daily standard deviation of pa- QA software (described in detail later in this section) rameter observations is less than or equal to twice the evaluates the data received from remote stations. standard deviation threshold (Table 1), all data are Finally, a meteorologist, trained in state-of-the-art QA flagged with a suspect flag for that entire day. The se- procedures, reviews each day the suspicious observa- verity of the flags is increased to a warning level when tions detected by other components of the QA system. the daily standard deviation is less than or equal to Human judgment is added to complement the auto- the standard deviation threshold. mated QA. For a complete description of the Okla- A second algorithm in the persistence test com- homa Mesonet's QA system, see Shafer et al. (2000). pares the daily range of each parameter to a predeter- mined "delta" threshold (Table 1). When the daily a. Automated QA range of observations is less than or equal to the The Mesonet's automated QA software consists of threshold value, all observations for that parameter five tests: 1) range, 2) step, 3) persistence, 4) spatial, receive a QA flag of warning for the day. All data are and 5) like-instrument. The software evaluates data flagged as suspect when the range of values exceeds calendar day by calendar day. At the end of the process, the minimum threshold, but the observed range is less one of four QA flags is assigned (each with increas- than 1.5 times the minimum threshold. speeds ing severity) to each datum: 0) "good," 1) "suspect," that are stuck near 0 m s"1 (because of freezing ) 2) "warning," or 3) "failure." Such stratification al- or with loose wires are often detected by lows the data user to decide the level of QA they pre- the persistence test. Similar to the thresholds set for fer. Most users request that data receiving warning or the step test, the thresholds used for the persistence failure flags from the QA system be removed from tests were determined by repeated testing of the algo- the datasets they obtain from Mesonet archives. rithms on Mesonet data. Meek and Hatfield (1994) documented algorithms similar to the Mesonet's range, step, and persistence 4) SPATIAL TEST tests, as well as a system of QA flags. They also noted The spatial test performs a Barnes objective analy- that a flagging system could serve to find real but sis (Barnes 1964) for each parameter at each obser-

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC vation time. The analysis is repeated with each sta- A similar method of performing objective intercom- tion successively excluded from the field being ana- parisons between neighboring stations is described in lyzed. For each analysis, an expected value, based on Wade (1987). data from surrounding stations, is calculated for each site. The expected value is compared with the actual 5) LIKE-INSTRUMENT TEST observation at that site. If the difference between the The like-instrument test compares the air tempera- estimate and the observed value is greater than 2 times tures at 1.5 and 9 m at Mesonet sites equipped with the standard deviation of the sample (known bad ob- sensors at both levels. When the two sen- servations are excluded from the sample), the obser- sor values differ by more than 10°C, the data are vation receives a suspect flag from the test. If the flagged as suspect. This threshold was determined by difference exceeds 3 times the standard deviation, a analyzing a of extreme inversions detected warning flag is issued. Minimum standard deviations within the depth of the Mesonet tower (Fiebrich and are also established so that during quiescent times, a Crawford 1998). It is planned for the like-instrument good observation does not become flagged due to a test to expand and compare at vary- very small departure from its expected value (Table 1). ing levels and wind speeds at different heights.

TABLE 1. Ranges and thresholds for the automated QA tests.

Min Observation Range Step Persistence Persistence spatial Parameter interval allowed allowed std dev delta std dev

Relative (%) 5 min 0-103 20 0.1 0.1 20.0

1.5-m air temperature (°C) 5 min -30-50 10 0.1 0.1 3.0

10-min (m s_1) 5 min 0-60 40 0.0 0.0 5.0

10-m 5 min 0-360 360 0.0 0.0 45.0 (degrees)

Wind direction std dev 5 min 0-180 90 0.1 0.1 60.0 (degrees)

Wind speed std dev (m s"1) 5 min 0-20 10 0.01 0.1 5.0

Maximum wind speed (m s"1) 5 min 0-100 80 0.05 0.1 10.0

Rainfall (mm) 5 min 0-508 25 N/A N/A N/A

Pressure (mb) 5 min 800-1050 10 0.1 0.1 1.5

Solar radiation (W m~7) 5 min -1-1500 800 0.1 0.1 400.0

9-m air temperature (°C) 5 min -30-50 10 0.1 0.1 3.0

2-min wind speed (m s"1) 5 min 0-60 40 0.1 0.1 5.0

5-cm soil temperatures (°C) 15 min -30-55 3 0.001 0.0 5.0

10-cm soil temperatures (°C) 15 min -30-50 3 0.001 0.0 5.0

30-cm soil temperature (°C) 15 min -30-50 3 0.001 0.0 5.0

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC 6) DECISION-MAKING ALGORITHM conditions exist, strong low-level inversions may form A decision-making algorithm compiles the results overnight and create a wide range of surface air tem- of the five tests, and logically determines a final flag peratures. Depending on local topography, a station for each datum (Shafer et al. 2000). For example, may observe conditions that are suspiciously cool and when two sequential observations fail the step test, moist or suspiciously warm and dry. both receive warning flags. However, one of the ob- Topographically induced variations in temperature servations may in fact be good (i.e., consecutive air are well known in the meteorological and agricultural temperature values of 10.0°, 10.1°, and 31.2°C will community. Farmers usually are familiar with the result in QA failures for the 10.1° and 31.2°C obser- protected areas of their land that are susceptible to vations, despite the fact that the 10.1°C observation "frost pockets." Operational meteorologists must be may be fine.) Results from the spatial test can help careful to warn for potential freezes in low-lying ar- determine when one of the QA flags from the step test eas on calm nights when temperatures are expected should be decreased in severity. In addition, the re- to remain well above freezing in other areas. Likewise, sults from the spatial and like-instrument tests can areas with higher elevations relative to their surround- determine which temperature sensor is malfunction- ings may sometimes poke above the surface-based in- ing (i.e., either the sensor at 1.5 or at 9 m) when ob- version. Most of the radiationally cooled air generated servations fail the like-instrument test. Finally, the QA at a higher elevation site is advected away, leaving the flags from all the tests are combined so that a final station in warmer air above the inversion layer. In fact, automated QA flag is increased in severity when the the Oklahoma Mesonet has documented many cases datum receives suspect or warning flags from more of strong low-level inversions with a temperature at than one test. 1.5 m near 0°C and a temperature at 9 m in excesses of 10°C (Fiebrich and Crawford 1998). b. The quality of parameters table Two main mechanisms appear to cause localized Trained meteorologists track suspect and bad data cold anomalies. One source of cold temperatures in detected by manual QA methods in a database termed low-lying areas is drainage of cold air into regions the "quality of parameters" (qualparm) table. These with lower elevations. A second source is cool- observations, in many cases, include sensor problems ing in areas sheltered from horizontal mixing. During not detected by the automated QA tests. Wind sen- in situ cooling, elevation differences and katabatic sors with bad bearings or rain gauges with intermit- flow are not necessary for cold air to pool. tently failing switches are two examples. Sensors that Gustavsson et al. (1998) investigated the noctur- produce observations with small biases that, in turn, nal development of the air temperature field and its are identified by long-term manual QA represent an- vertical structure over hilly terrain. They found the other category of events added to the qualparm table. largest temperature gradients developed shortly after The QA software compares the QA flags in the sunset—hardly long enough for cold air to flow any qualparm table with the flags from the automated QA great distance. This fact indicated that in situ cooling tests. The most severe flag (either from the qualparm was dominant over gravity-induced cold air flow. table or from the automated QA) is archived with each Similar to the findings of Gustavsson et al., the great- datum. est spatial temperature gradients observed in the Okla- In the sections that follow, examples where the homa Mesonet and the ARS Micronet usually occur automated quality control procedures failed due to the shortly after sunset. Rapid radiational cooling forms occurrence of small-scale meteorological phenomena strong inversions that prevent mixing in some areas. are discussed. These instances illustrate the impor- Those locales cool quickly, while other areas continue tance of performing manual quality control in conjunc- to mix with warm air from the lower due tion with the automated quality control procedures. to mechanical turbulence. Late in the night, the tem- perature gradients usually relax. At this time during the night, cold air drainage may become a key player 3. Cold air pooling and inversion because a correlation develops between surface tem- poking perature and elevation. The temperature gradients that arise solely from elevation differences are almost al- Meteorological conditions sometimes create a very ways less intense than is the original gradient formed stable boundary layer across Oklahoma. When these by in situ cooling.

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC Thompson (1986) also noted that drainage of cold air temperature of 7.5°C at 0600 UTC (Fig. lb), air is not necessary for cold pools to develop. In fact, whereas the expected air temperature (from the spa- he noted that katabatic flows usually follow the for- tial test) was 16.6°C. This difference of-9.1°C (ob- mation of strong temperature inversions and spatial servation minus expected) was greater than 3 times gradients—they do not precede them. As long as the the standard deviation of the air temperature field. surface-based air is prevented from mixing with Thus, the spatial test produced a warning flag for the warmer air above, the cooling rate at the surface is data from A167. Knowing that the average station increased. Therefore, even sheltered areas (i.e., a clear- spacing in the ARS Micronet is approximately 5 km, ing in a forested area) can prohibit mixing and can be it is remarkable that temperature gradients near 10°C susceptible to rapid cooling. can exist over such short distances. However, these After several years of looking at small-scale fea- strong gradients are common during the tures revealed by Mesonet data, a list of sites sus- months. ceptible to cold air pooling and inversion poking has A time series plot of the standard deviation of the been developed. When these stations report observa- air temperature field in the ARS network on 25-26 tions that are spatially inconsistent with neighboring October 1999 is depicted in Fig. 2. The air tempera- sites on calm, clear nights, there is usually little cause ture field across the Micronet remains fairly homo- for concern. The disappearance of the anomalies geneous (standard deviation of approximately 0.5°C) around sunrise also is a good indicator that a real event during the late afternoon (2045-2300 UTC), but it occurred, rather than a sensor began to malfunction. a. Case study 1: Intense radiational cooling of 26 October 1999 Surface temperatures across the Mesonet on 26 October 1999 at 0600 UTC (to convert to local standard time for Oklahoma, subtract 6 h) are shown in Fig. la. In west central Oklahoma, Erick (ERIC) observed a temperature of 5.7°C, while the neighboring site at Cheyenne (CHEY) reported 15.6°C. Similarly, the Nowata (NOWA) site in northeast Oklahoma observed 4.3°C, while its neighbor at Skiatook (SKIA) reported 14.5°C. In addition to these two dramatic examples of large spatial tem- perature variations, countless sets of neighboring sites differ by 5°C on this map. Some of the variations can be at- tributed to in situ cooling, while others more closely correlate with differences in topography. With an average station spacing of only 30 km in the Oklahoma Mesonet, it is not surprising that the spa- tial QA test would erroneously flag many observations. High variability in nighttime surface temperatures also has been found to oc- FIG. 1. (a) Mesonet station plot of the surface temperature field (°C) across Oklahoma at 0600 UTC 26 Oct 1999. Radiational cooling caused large varia- cur frequently in the ARS Micronet. For tions in temperature between many neighboring stations, (b) As in (a) except instance, on the same night as discussed for the ARS Micronet. Average station spacing is only 5 km in the Micronet, above, ARS station A167 observed an yet some neighboring sites differ by as much as 10°C.

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC deviation slowly decreases after midnight (0600 UTC), but decreases rapidly with the onset of sunrise. These observations support the theory that in situ cooling dominates over cold air drainage and creates large gra- dients in surface temperatures.

b. Case study 2: A comparison between two stations that experienced topographically induced variations in temperature On the same day as in the previous case study, data from the Newkirk (NEWK) station in north-central Oklahoma revealed that the site "poked" above the inversion layer (Fig. 3a). At 0600 UTC, the expected FIG. 2. Time series plot of the std dev of surface air tempera- air temperature at Newkirk (NEWK) was 7.3°C. tures (°C) across the Micronet for 2045 UTC 25 Oct 1999 However, NEWK observed an air temperature of through 2245 UTC 26 Oct 1999. The std dev increased at sun- 15.4°C. On almost any other occasion, this difference set due to in situ cooling and decreased sharply at sunrise due of 8.1°C (between the observed and expected air tem- to increased vertical mixing. perature) would indicate a problem thermistor. As a result, an automated QA flag labeled the temperature quickly becomes stratified as sunset approaches. The data as suspect. Because NEWK is located at a higher largest increase in the standard deviation of the air elevation than are surrounding stations (Fig. 3b), temperature field occurs between 0000 and 0245 UTC NEWK remained in the well-mixed layer above the (1 h prior to sunset and 2 h afterward). The standard cold inversion.

FIG. 3. (a) Mesonet station plot of the surface temperature field (°C) across northern Oklahoma at 0600 UTC 26 Oct 1999. The automated spatial QA flagged the NEWK observation as suspect due to the warm anomaly there, (b) As in (a) except the terrain elevation (m). (c) Mesonet station surface temperature field (°C) across eastern Oklahoma at 1215 UTC 4 Nov 1999. The WIST observations also received erroneous QA flags due to the cool anomaly there, (d) As in (c) except the terrain eleva- tion (m).

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC In contrast to the NEWK site, Wister (WIST) fre- on the trailing edge of . A second type quently experiences cold air pooling. Despite the fact of mesolow (often called a "pre- low") is that southeast Oklahoma usually observes the warm- sometimes located in advance of a convective line. est overnight temperatures, WIST recorded the Although the pre-squall-type mesolow is often very second-lowest temperature in the state at 1215 UTC pronounced, it is rarely intense enough so that pres- on 4 November 1999 (Fig. 3c). WIST's elevation of sure data from a site fail the automated QA. In con- 143 m makes it one of the lowest stations above sea trast, data from mesohighs and wake lows frequently level in the Mesonet (Fig. 3d). The valley location of fail the spatial test because these features are usually WIST created undisturbed radiational cooling during more intense. the hours after sunset. Additional cooling via cold air The region of a is usually characterized drainage during the night contributed to a low tem- by a cool, dense lower troposphere. A dense air mass perature of-1.4°C, whereas many neighboring sta- near the surface creates higher surface pressures tions observed sunrise temperatures that were greater (Fujita 1959). When heavy rainfall occurs in the than 9.0°C. mesohigh region, the weight of the hydrometeors also contributes to the strength of the mesohigh (Sanders and Emanuel 1977). 4. Mesohighs and mesolows The physical processes that create wake lows are not fully understood. The main mechanism for the Mesohighs and mesolows are mesoscale pertur- appears to be an unusual warming of the bations in the surface pressure field. They generally lower troposphere. The scientific consensus is that the are associated with convective precipitation. warming is caused by strong subsidence. Scientists Mesohighs are usually centered along a convective have connected the subsidence with processes that in- line, whereas mesolows (or "wake lows") are located clude divergence of cold air near the surface, the rear

FIG. 4. (a) Mesonet station plot of the sea level pressure field (mb) at 0700 UTC 25 May 2000. At this time, the spatial QA test began to flag the anomalously low pressure at WYNO. (b) Base reflectivity from the Inola radar at 0659 UTC 25 May 2000. Note that WYNO lies on the back edge of the precipitation, (c) As in (a) except for 0730 UTC. The wake low has moved southeast and is being observed by the SKIA station, (d) As in (b) except for 0730 UTC. SKIA now lies on the west edge of the .

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC inflow into a convective system, evaporational cool- good indicator that the pressure anomaly is a real ing, gravity waves, and microphysical processes event. within the stratiform region (concepts summarized by Haertel and Johnson 2000). a. Case study 3: A wake low event Mesohighs and mesolows occur during all months A mesoscale convective system with an intense of the year in Oklahoma. At any given time when these wake low caused a number of observations to be in- features are underway, these mesoscale phenomena appropriately flagged by the spatial QA on 25 May affect only one or two stations. Thus, the spatial test 2000. The first observations to be flagged were from often places suspect or warning flags on the data from Wynona (WYNO) in northeast Oklahoma at 0700 UTC the station observing the pressure perturbation. As the (Fig. 4a). The wake low was spatially small in scale, pressure perturbation advances along with the convec- and caused the sea level pressure at WYNO to drop tion, a second and third station is oftentimes impacted. to 1003.6 mb; meanwhile stations immediately sur- In these situations, radar and rainfall data are helpful rounding WYNO were 4.0-6.2 mb higher. The base to diagnose when pressure observations have been in- reflectivity image from the Inola radar at 0659 UTC appropriately flagged by the automated QA. In addi- indicated that WYNO was immediately west of a large tion, the occurrence of very strong at and near area of convection (Fig. 4b). In addition to the small the site observing a suspicious pressure reading is a area of the wake low, a large, elongated mesohigh was located underneath the most intense con- vection in far eastern Oklahoma. The wake low remained well defined as it drifted east-southeast along with the convection. The Skiatook (SKIA) sta- tion next observed the mesolow at 0730 UTC with a similar pressure drop to 1003.2 mb (Fig. 4c). For comparison, the radar image at 0730 UTC is shown in Fig. 4d. Thirty minutes later, the Claremore (CLAR) station observed a sea level pressure of 1003.6 mb as it was influenced by the mesolow at 0800 UTC (not shown). In each instance, the wake low was observed to be on the west edge of the precipitation. Thus, the wake low of 25 May 2000 caused the pressure ob- servations from WYNO, SKIA, and CLAR to receive a number of suspect and warning flags from the automated QA-

b. Case study 4: A mesohigh event On 1 June 1999, a strong mesohigh developed under an intense convective in southwest Oklahoma. In 20 min, the sea level pressure at Fort Cobb (FTCB) rose 4.02 mb. Due to the small spatial scale of the intense mesohigh, the pressure data from FTCB were FIG. 5. (a) Mesonet station plot of the sea level pressure field (mb) at flagged as warning by the automated 0235 UTC 1 Jun 1999. The FTCB station was determined to be observing an intense mesohigh. (b) Base reflectivity from the Frederick radar at 0236 UTC 1 QA. Figure 5a depicts the sea level pres- Jun 1999. Note that FTCB was located beneath a large area of radar echoes which sure field over southeast Oklahoma at were 50-60 dBZ in intensity. 0235 UTC. The FTCB pressure was

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC 6.5 mb greater than that at the nearby Hinton (HINT) Heat bursts are thought to originate from air that site located 40 km north. The base reflectivity from has subsided dry adiabatically from midlevels of in- the Frederick radar at the same time is depicted in nocuous . They are usually accompanied Fig. 5b. Note that the FTCB site was located beneath by downdrafts from decaying areas of convective pre- a large area of radar echoes that were 50-60 dBZ in cipitation. Williams (1963) was the first scientist to intensity. Downdrafts of cool air and almost 25 mm document the occurrence of heat bursts. Johnson of rain in 20 min were observed during this mesohigh (1983) studied in detail a event from 29 May event at FTCB. Such ancillary information provided 1976, and found that the lower troposphere was con- strong support that the pressure anomaly was a real ditionally unstable in five special soundings released event, despite the fact that the data were flagged as that day. Those soundings also suggested that evapo- erroneous by the spatial test. rative cooling played a large role in triggering the me- soscale subsidence necessary for heat burst formation. Johnson et al. (1989) suggested that the strong 5. Heat bursts downdrafts responsible for heat bursts could be due to circulations associated with the stratiform region Heat bursts are mesoscale phenomena that usually of a mesoscale convective system. occur in association with dying thunderstorms. On When heat bursts occur in Oklahoma, they are occasion, heat bursts produce anomalously warm usually detected by only a limited number of stations. surface temperatures. Heat bursts have historically Because of the small spatial scale of a heat burst, the been difficult to detect due to their small spatial and observations from an event are often flagged as erro- temporal scale. However, the Oklahoma Mesonet neous. Along with a temperature increase, sites that and the ARS Micronet have made it possible to ob- detect heat bursts usually observe a sharp decrease in serve these "rare" phenomena on numerous occasions. relative humidity and a rapid increase in wind speed. Perhaps the most intense heat burst ever captured by These additional parameters plus radar data, when an observational network occurred on 23 May 1996. considered as a set, can lead to a more accurate deter- This event impacted a large portion of southwest mination of data quality than is produced by auto- Oklahoma and created damaging winds as air tem- mated routines. peratures rose to near 40°C (MacKeen et al. 1998). Based upon more than 75 heat burst events detected Case study 5: A northwest Oklahoma heat burst within the Mesonet since 1994, the most dramatic heat A localized heat burst occurred in northwest Okla- bursts have occurred at night when temperatures homa on 20 September 1998 (Fig. 6a). On that morn- sometimes rise higher than the afternoon maximum ing, the temperature at the Buffalo site (BUFF) rose temperature. to 32.8°C as winds gusted to 23.6 m s"1 (Fig. 6b).

FIG. 6. (a) Mesonet station plot of the air temperature field (°C) across northwest Oklahoma at 1145 UTC 20 Sep 1998. A heat burst is underway at the Buffalo (BUFF) station, (b) As in (a) except for the wind gust field (m s"1)- During the heat burst event, the BUFF station recorded wind gusts in excess of 23 m s_1.

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC By 1145 UTC, the heat burst was well underway at the BUFF station (cf. Fig. 6 with the radar image in Fig. 7b). Because the radar echoes decreased significantly in intensity, the storm likely collapsed and produced the heat burst. As the moved east of BUFF, the contin- ued decay of the convection is evident at 1244 UTC (Fig. 7c).

6. Snowfall and snow cover

When wintry precipitation visits Oklahoma, data quality problems often develop. During ice storms, the automated QA is very successful in determining that wind sensors have accumulated ice and are frozen in place. The automated QA also is good at identifying barom- eter tubes that become sealed from the atmosphere due to ice accumulation. On the downside, accurate observations are sometimes incorrectly flagged due to the unique meteorological conditions that accompany snowfall.

a. Case study 6: A persistent snowfall A long duration snow event on 27 January 2000 became a problematic case for the persistence test in the automated QA system. Light to moderate snow fell FIG. 7. (a) Base reflectivity image from the Vance AFB radar at 1046 UTC continuously for most of the day over a 20 Sep 1998. A convective storm is approaching the BUFF station, (b) As in (a) except for 1145 UTC. The convective storm began to collapse as BUFF re- large portion of central Oklahoma. As is corded a heat burst, (c) As in (a) except for 1244 UTC. The storms continued to evident from an inspection of a map of decay as they moved east of the BUFF station. daily minimum and maximum tempera- ture across the Micronet (Fig. 8a), the snowfall kept most stations at -2° to These observations seemed inconsistent with nearby -3°C for the entire UTC day. The standard deviation observations from stations in the area. As a result, the of the daily temperature (rounded to tenths) across the automated QA flagged the data as erroneous. Micronet (Fig. 8b) revealed many stations observed Within 45 min, the heat burst dissipated, winds de- standard deviations of only 0.2°C (based on the 5-min creased, and the temperature decreased rapidly to data throughout the day). When the standard devia- 22.7°C. Associated with the initial temperature rise tion of air temperature throughout the day is less than and gusty winds, a small decrease in pressure and or equal to 0.2°C, data for an entire 24-h period are a distinct decrease in dewpoint temperature were flagged as suspect by the persistence test. This qual- observed. ity assurance threshold was met at Micronet stations: Figure 7 provides a 3-h radar history of the decay- A134, A149, A158, A162, A163, A164, A167, and ing convection as it moved across the BUFF station. A168. Thus, data from those sites were inappropri- The first image (Fig. 7a) in this series depicts a mod- ately flagged by the automated QA. erately strong storm approaching BUFF at 1046 UTC.

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC b. Case study 7: A localized heavy snow event an hour (as detected by the Mesonet's ). Heavy snowfall is usually a localized phenomenon Moreover, an irrigated field upwind of a station can in Oklahoma. On a number of occasions each win- increase significantly the measured relative humidity. ter, the Panhandle and western portions of the state An extreme example of this influence occurs each receive snowfall, while the rest of Oklahoma receives winter as the wheat fields in western Oklahoma cre- rain. On 4 December 1999, a band of heavy snow fell ate a moist anomaly across a 6-7 county area. The across Ellis, Harper, Woodward, and Woods counties higher relative humidity values observed result from in northwest Oklahoma. In these areas, 15 to 20 cm abundant transpiration that occurs in the nearby wheat of snow was reported (National Climatic Data Cen- fields. The same agricultural fields have a surprisingly ter 1999). The visible image from 7 Decem- strong impact on air temperature observations in those ber 1999 depicted the narrow band of snow still on counties (normally, Mesonet sites are much cooler). the ground across that region days later (Fig. 9a). The distinction between snow-covered and bare ground a. Case study 8: The impact of native sod also was evident in the surface air temperature field characteristics on soil temperatures across northwest Oklahoma (Fig. 9b). Buffalo The high degree of spatial variability observed in (BUFF) dropped to -11.9°C at 1000 UTC on 6 De- soil temperature makes it a very difficult parameter cember due to a deep snowpack at that location. This to quality control. A wide range of soil temperatures observation was 5.6°-8.6°C colder than were the ob- is observed across the statewide Mesonet and the servations at Slapout (SLAP) and Beaver (BEAV), smaller Micronet, especially during the station locations west of the snowpack. BUFF was also 4.4°-9.5°C colder than were stations south and southeast of the greatest snow depth. Due to these spa- tial differences in temperature, the BUFF data were flagged as suspect by the automated QA.

7. Vegetation influences on

The characteristics of vegetation in and around a meteorological observing station sometimes dramatically impact the observations of several meteorologi- cal parameters. The soil temperature, understandably, is sensitive to the char- acteristics of the soil. Darker natu- rally warm more than do light-colored soils under similar conditions of solar ra- diation. In addition, the vegetation grow- ing above the soil temperature sensors modulates the amount of insolation re- ceived by the soil; in effect, vegetation has a strong impact on the variability of soil temperature. Likewise, the vegetation of surround- ing land areas also influences the meteo- FIG. 8. (a) Micronet station plot of the maximum and minimum temperature (°C) observed on 27 Jan 2000. Most stations observed a minimum variation in rological observations at a Mesonet site. temperature during the day due to persistent snowfall, (b) As in (a) except for A grove of trees on the eastern horizon the std dev of the daily temperature (°C). Many stations observed std dev as may delay sunrise from a few minutes to low as 0.2°C.

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC (Fig. 10c). The vegetation at both sites was similar to the surrounding land area (an objective for all Mesonet/Micronet sites). Thus, data from both sites were deemed to be "quality" data, despite the fact that both sets of data failed the au- tomated QA test.

b. Case study 9: A cooling effect from an agricultural field Each summer, a number of users ex- press concern about a cool temperature anomaly at the Altus (ALTU) site in southwest Oklahoma. A snapshot of sur- face temperatures in that region at 2300 UTC on 8 August 1998 is shown in Fig. 11. Temperatures across the counties of southwest Oklahoma were nearly uniform at 38°C, except at ALTU which was 4°C cooler. As indicated in Fig. 12, this cool bias is not always present. In fact, ALTU and nearby Tipton (TIPT) observed similar after- noon air temperatures on 7 August 1998 while the afternoon wind speed averaged less than 1.6 m s_1. However, as average wind speeds exceeded 5 m s_l on 8-9 FIG. 9. (a) Visible satellite image from 7 Dec 1999. Snow cover from the 4 August 1998, a cool bias at ALTU be- Dec 1999 snow event remains evident across northwest Oklahoma, (b) Mesonet comes apparent. station plot of the air temperature field (°C) across northwest Oklahoma at This cool anomaly can be explained 1000 UTC 6 Dec 1999. The deep snow cover at the BUFF station caused the by using site photos posted online overnight temperature to decrease to -11.9°C. (http://okmesonet.ocs.ou.edu). A site photo from ALTU reveals a large irri- months. For example, the Micronet soil temperature gated cotton field surrounding most sides of the sta- at 5 cm under sod (TS05) on 17 August 1999 is shown tion (Fig. 13). As surface air moved across the cotton in Fig. 10a. The observations at 1900 UTC have a fields toward the ALTU station, temperatures at the range of over 17°C. As a result, the application of the site were cooler (due to increased latent heat flux and spatial test created numerous suspect and warning decreased sensible heat flux) when compared with air flags at several Micronet stations. temperature data from Tipton (20 km to the south- Because Micronet stations A146 and A163 rep- east). Despite the fact that the ALTU temperature data resented the extremes in TS05, Mesonet scientists vis- appeared suspect, in reality, the data were deemed to ited both sites to determine whether the sensors were be quality observations. functioning properly. To document the vegetation, In contrast to other case studies in this manuscript, site photos of the soil temperature plots were taken. the transient cool anomaly at ALTU has never been Not surprisingly, A146 (with a temperature of 46.2°C intense enough for the temperature data to fail the at 1900 UTC) and A163 (with a temperature of automated QA. A temperature difference of 6°C com- 28.0°C at 1900 UTC) had dramatically different veg- pared to surrounding sites would be necessary for the etation. The ground cover at A146 had patches of spatial test to suggest poor quality data. However, the near-dormant grass around the soil temperature plot ALTU observing site represents a good example of (Fig. 10b). Station A163, in contrast, had green grass how land use affects meteorological observations and 30-50 cm tall covering its soil temperature plot make quality control more difficult.

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC 8. Cataloging unique meteorological events in the qualparm table

The cases of cold air pooling, inversion poking, mesohighs, mesolows, heat bursts, snowfall and snow cover, and variations in vegetation discussed in sections 3 through 7 illustrate instances when unique meteorological phenomena im- pacted the Mesonet's automated quality control procedures. To combat these problems, there are plans for the Mesonet to use its existing qualparm table to cata- log the start and end times of events that created unwarranted QA flags by the au- tomated tests. The qualparm table represents a da- tabase that uses human intervention to flag data as erroneous regardless of results from the automated QA tests (section 2b). It has been successfully used since 1996 to increase the severity of QA flags be- FIG. 10. (a) Micronet station plot of the soil temperatures at 5 cm under sod cause trained personnel sometimes rec- (°C) at 1900 UTC 17 Aug 1999. Despite an average station spacing of only 5 km, ognized erroneous data while automated many neighboring sites differ by 5°-10°C. (b) and (c) Site photos on 17 Aug 1999 depicting the vegetation cover over the 5-cm soil temperature plot. The routines did not. In the future, the qualparm vegetation at A146 was sparse and near dormant, whereas the vegetation at A163 table will also be used to decrease the se- was thick and green. verity of QA flags when real meteoro- logical events are known to have occurred. In addition to the QA flags of 0-3 (good through designed quality control system is vital. Automated failure), two new levels of flags will be added. A flag quality assurance software provides an efficient means of 5 would dictate that the final automated QA flag to sift through large volumes of data and to flag ob- be reduced by 1 (i.e., a suspect QA flag would be decreased to good, or a warn- ing QA flag would be decreased to sus- pect). A flag of 6 would dictate that the QA flag be reduced to 0 (i.e., the data should be deemed as good no matter what the automated QA determined.) These additional levels will allow a me- teorologist to manually downgrade the severity of QA flags for a specific sta- tion at a specific time when appropriate, and allow more data to reach scientists without unwarranted flags.

9. Conclusions FIG. 11. Mesonet station plot of the air temperature field (°C) across south- west Oklahoma at 2300 UTC 8 Aug 1998. The Altus (ALTU) temperature ap- To ensure quality data from a meteo- pears to have a cool bias, but in fact, the cool anomaly is real and results from rological observing network, a well- the influence of a large agricultural field upwind of the site.

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC flagged good observations during peri- ods of steady snowfall. Although not discussed in this manu- script, the step test, range test, and like- instrument test also have erroneously flagged good data. For instance, the hot, dry summer of 1998 created some soil temperatures in excess of the maximum range of 50°C used at that time. As air temperatures soared repeatedly above

FIG. 12. Time series plot of the air temperature at Altus (solid line) and Tipton 40°C in 1998, it became obvious that flags (starred line) on 7-9 Aug 1998. The two stations observed similar temperatures from the range test for soil temperatures on 7 Aug when afternoon wind speeds averaged less than 1.6 m s_1. When aver- were unwarranted. That same summer, soil age wind speeds increased to greater than 5.4 m s_1 on 8-9 Aug, Altus devel- temperature observations occasionally oped a 4°C cool bias due to increased advection over irrigated cotton fields. failed the step test when rain fell on hot soils (> 50°C) and created rapid cooling. servations that truly are erroneous. While employing The step test also "failed" solar radiation data on algorithms to flag "bad" data, it is difficult to avoid occasions when bright sunshine sporadically broke flagging data as erroneous when small-scale phenom- through puffy cumulus . Although rare, solar ena occur. radiation has changed more than 800 W m-2 between Data from many micro- and mesoscale meteoro- consecutive observations during midsummer. Data logical events are difficult to quality control. For ex- collected during low-level temperature inversions that ample, when the spatial test compares data acquired were unusually strong also have been flagged by the from close neighboring stations, perfectly good ob- like-instrument test when temperature differences servations may differ enough to fail the test. Cold air exceeded 10°C between 1.5 and 9 m. pooling, "inversion poking," mesohighs, mesolows, Despite these problems, the Oklahoma Mesonet heat bursts, variations in snow cover, and variations has implemented a QA system that represents a com- in vegetation have all caused the spatial QA test used promise between having automated QA tests stringent by the Oklahoma Mesonet to flag good data as erro- enough to catch bad data, but relaxed enough to per- neous. Likewise, the persistence test often incorrectly mit meteorological observations of real phenomena to be correctly flagged in the archives. Most importantly, the automated QA tests em- ployed by the Oklahoma Mesonet are thought to be rigorous enough to ensure that research-quality data are collected on a routine basis. By cataloging the me- teorological events that fail the auto- mated QA in the qualparm table, even better quality assurance of Mesonet and Micronet data will be accomplished. During the one to two years that fol- low the publication of this manuscript, there are plans for the Mesonet to imple- ment an upgraded suite of QA software. This software will include a number of new automated algorithms. In addition, the capability of a QA system that can readjust inappropriate automated QA flags via the qualparm table will be FIG. 13. Site photo from the Altus station. Much of the land area around the implemented. site consists of irrigated cotton fields.

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC Acknowledgments. The integrity of data from the Oklahoma Johnson, B. C., 1983: The heat burst of 29 May 1976. Mon. Wea. Mesonet and ARS Micronet is due in large part to the work of Rev., Ill, 1776-1792. Scott Richardson, David Grimsley, Leslie Cain, Gavin Johnson, R. H., S. Chen, and J. J. Toth, 1989: Circulations as- Essenberg, James Kilby, Ken Meyers, and Bill Wyatt who main- sociated with mature-to-decaying midlatitude mesoscale con- tain both networks. The authors also acknowledge Mark Shafer, vective system. Part I: Surface features—Heat bursts and Derek Arndt, Tim Hughes, Sherman Fredrickson, and the pre- mesolow development. Mon. Wea. Rev., 117, 942-959. vious QA managers for their work in building a strong founda- MacKeen, P., D. L. Andra, and D. A. Morris, 1998: The 22-23 tion for quality Mesonet data. In addition, the Critical Computer May 1996 heatburst: A severe wind event. Preprints, 19th Operations Group at OCS includes Sridhar Khulasekharan, Conf. on Severe Local Storms, Minneapolis, MN, Amer. Jessica Thomale, Jared Bostic, and the Mesonet operators who Meteor. Soc., 510-513. are a vital component of data management for the Mesonet. The Meek, D. W., and J. L. Hatfield, 1994: Data quality checking authors are grateful for the Software Development Group at for single station meteorological databases. Agric. For. Me- OCS, which includes Mike Wolfinbarger, Brad Stanley, Robb teor., 69, 85-109. Young, Billy McPherson, and Justin Greenfield whose software National Climatic Data Center, 1999: Climatological Data Okla- made possible the graphical visualizations used in this manu- homa. Vol. 108, No. 12, Department of Commerce, 31 pp. script. Finally, we thank the anonymous reviewers whose atten- Sanders, F., and K. A. Emanuel, 1977: The momentum budget tion to detail greatly improved the quality of this manuscript. and temporal evolution of a mesoscale convective system. J. Atmos. Sci., 34, 322-330. Shafer, M. A., C. A. Fiebrich, D. S. Arndt, S. E. Fredrickson, and T. W. Hughes, 2000: Quality assurance procedures in References the Oklahoma Mesonet. J. Atmos. Oceanic Technol., 17,474- 494. Barnes, S. L., 1964: A technique for maximizing details in nu- Snyder, R. L., and W. O. Pruitt, 1992: Evapotranspiration data merical map analysis. J. Appl. Meteor., 3, 396-409. management in California. Irrigation and Drainage, Sav- Fiebrich, C. A., and K. C. Crawford, 1998: An investigation of ing a Threatened Resource—In Search of Solutions, T. E. significant low-level temperature inversions as measured by Engman, Ed., American Society of Civil Engineers, 128- the Oklahoma Mesonet. Preprints, 10th Symp. on Meteoro- 133. logical Observations and Instrumentation, Phoenix, AZ, Thompson, B. W., 1986: Small-scale katabatics and cold hol- Amer. Meteor. Soc., 337-342. lows. Weather, 41, 146-153. Fujita, T. T., 1959: Precipitation and cold air production in Wade, C. G., 1987: A quality control program for surface mesoscale systems. J. Meteor., 16, 454-466. mesometeorological data. J. Atmos. Oceanic Technol., 4, Gustavsson, T., M. Karlsson, J. Bogren, and S. Lindqvist, 1998: 435-453. Development of temperature patterns during clear nights. J. Williams, D. T., 1963: The thunderstorm wake of May 1961. Appl. Meteor., 37, 559-571. National Severe Storms Project Rep. 18, 23 pp. [NTIS PB- Haertel, P. T., and R. H. Johnson, 2000: The linear dynamics of 168223.] mesohighs and wake lows. J. Atmos. Sci., 57, 93- 107.

Bulletin of the American Meteorological Society 219 7

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC GLOSSARY OF

Forty-one years ago, the AMS published the . Containing 7900 terms, more than 10,000 copies have been sold over four decades through five printings. It is a tribute to the editors of the first edition that it has withstood the test of time and continued to be among the leading reference sources in meteorology and related sciences.

Now, over five years in the making, the second edition is available. The volume contains over 12,000 terms, including those from "new" disciplines, such as satellite meteorology and numerical weather tM^ prediction. In addition, related oceanographic and hydrologic terms are defined.

The Glossary of Meteorology, Second Edition, was produced by an editorial board comprised of 41 distinguished scientists and the participation of over 300 contributors. The CD-ROM version is compatible with Windows, Macintosh, and most UNIX platforms, and features hyperlinked cross-references.

electhojet e WSR-88D

ISBN 1-878220-34-9, approx. 850 pp., hardbound. To place an order, refer to the pricing chart below and submit your prepaid orders to: Order Department, AMS, 45 Beacon Street, Boston, MA 02108-3693; call 617-227-2425 to order by phone using Visa, Mastercard, or American Express; or send e-mail to [email protected]. Please make checks payable to the American Meteorological Society. oret& HARDBOUND BOOK CD-ROM $80 List $95 List $55 AMS Members $65 AMS Members $30 AMS Student Members $35 AMS Student Members

2180 Vol. 82,, No. 10, October 2001

Unauthenticated | Downloaded 10/11/21 09:50 AM UTC