REU 2007 Projects

Rachel Butterworth
Mentor: Kevin Kloesel

Communicating Weather Radar Research to Public Audiences: The Proposal

Currently the National Science Foundation (NSF) is seeking proposals to communicate NSF-funded research programs to public audiences. The American public generally supports government funded scientific research. However, despite their favorable attitude, Americans are scientifically illiterate. Weather radar is one of the most important tools for both forecasting and meteorological research. It is also a useful tool for the public. It is crucial the public understands at least some of what is being accomplished in regards to weather radar research because one informed decision during a hazardous weather event may mean the difference between life and death. A proposal was written to use video and web-based material to teach the public about weather radar research, past and present. A literature review was conducted across multiple disciplines prior to constructing the proposal in order to determine the most effective way to teach the public about weather radar. In addition, a survey instrument will be used to determine the public's knowledge-level of weather radar; information that will further shape the content of the proposed video and web-based material.
Doug Crauder
Mentor: Dave Zittel

Evaluation of Faster Scanning Strategies for the WSR-88D with the Combined Range Aliasing Mitigation Techniques

In a recent survey of WSR-88D sites, operational staff indicated a desire for faster volume coverage pattern (VCPs) that also mitigates range aliasing. In 2004, the Radar Operations Center (ROC) fielded a new technique call the Multiple Pulse Repetition Frequency (PRF) Dealiasing Algorithm (MPDA) that combines three sequential Doppler scans with PRFs of about 1300, 1100, 850 Hz, respectively, to mitigate range aliasing. Another technique to mitigate range aliasing, the Sachidananda-Zrnic phase coding algorithm (SZ-2), is being fielded in the summer of 2007. SZ-2, which provides the bulk of the velocity recovery, uses advanced signal processing to separate strong trip from weak trip signal. The two algorithms, when combined, recover an average of 98 percent of velocity area data out to 230 km. One major drawback with the combined technique is that a volume scan takes about five minutes forty-five seconds.

The ROC has investigated using two instead of three Doppler scans at each of the two lowest elevation angles thus reducing the volume scan time by about thirty seconds. Five data cases from the ROC’s test bed WSR-88D were analyzed. Four cases with widespread precipitation were collected during the fall and winter of 2006. The fifth case, a mesoscale convective system (MCS), was collected in June 2007. This case was recorded as Recombined and Super Resolution data. Recombined and Super Resolution products are currently in development at the ROC and scheduled for release in 2008. Using two instead of three Doppler scans yields an average velocity area recovery of 96.87 percent for the four widespread precipitations, 95.87 percent for the Recombined data on the MCS, and 83.83 percent for the Super Resolution data. Velocity dealiasing errors were also scored. Starting with a score of 100, points were deducted for different sized errors in the velocity field. For the widespread precipitation cases, using two instead of three Doppler scans decreased the number of errors slightly (0.045 percent decrease in errors); for the MCS, there was a 1.70 percent increase in errors for the Recombined data and a 2.34 percent increase for the Super Resolution data. These initial results support operationally fielding the combined algorithms with the middle (1100 Hz PRF) scan removed thus helping to meet the needs of radar sites. These would be deployed in the form of two new VCPs called 112 and 122.
Peter Finocchio
Mentors: Dave Hogan, Mark Leidner

Using Ensemble Prediction Systems to Estimate Medium Range Temperature Forecast Error

A salient benefit of an ensemble prediction system (EPS) is its ability to provide a means of estimating forecast error. This study tests the error prediction skill of three EPS features: spread (standard deviation) among ensemble members, consistency between MEX/MOS output and the ensemble mean, and consistency between consecutive 24-hour runs of an EPS. For 27 stations throughout the northeastern United States, 15-day high and low temperature forecasts from calibrated ECMWF ensemble output issued Feb 1 through May 31 2007 are examined. For each of the 15 forecast days, the significance (from r2 statistic) and slope of the error predictor-forecast error relationship is used to determine how valuable each feature is in estimating forecast error.

For high temperature forecasts in the northeast, ensemble spread and run-to-run consistency are most effective at predicting forecast error through 9-day lead times. Although error prediction skill in both drops off in the longer ranges, ensemble spread is more useful beyond 9-day lead times. Spread and run-to-run consistency are more effective for low temperature forecast error prediction, with ensemble spread still performing best beyond 9-day lead times. Model-to-model consistency is only a moderate to weak error predictor for high temperature forecasts in the short range. For the sake of comparison of error predictability between two regions with disparate climates, data from 21 stations in the southwestern United States are also examined. In this region, all three error predictors are consistently effective in anticipating forecast uncertainty for both high and low temperatures.
Victor Gensini
Mentor: Harold Brooks

Regional Variability of CAPE and Deep Shear from the NCEP/NCAR Reanalysis

Variations in the distributions of parameters that lead to deep moist convection from the National Center for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) 42 year global reanalysis dataset have been analyzed for 3 domains. Although the variability of the distribution of convective parameters is a little higher in the Eastern United States, the Central United States adequately represents the distribution of both domains, and therefore serves as a comparison to the South American domain. CAPE has been roughly increasing in the Central United States since the late 1960’s while South America has been exhibiting a downward trend in CAPE over the period. In fact, from 1970 to 1999 the two regions have exhibited very different characteristics when it comes to the distribution of CAPE. Deep shear in the presence of CAPE has not changed throughout the reanalysis period. Therefore, the increase of the product of CAPE and deep shear can be contributed to the increase of CAPE in the Central and Eastern United States.
Eric Guillot
Mentors: "Lak", Greg Stumpf, Travis Smith, Don Burgess

Tornado and Severe Thunderstorm Warning Forecast Skill and its Relationship to Storm Type

The amount of forecast skill involved when issuing tornado and severe thunderstorm warnings is closely related to the type of storm that causes the severe weather. Storms from eight tornado outbreaks are classified and correlated with tornado warnings and severe thunderstorm warnings. These warnings were verified, missed, or shown to be false alarms by relating them with storm reports that match temporally and spatially with those in the Storm Prediction Center’s database. Certain forecast parameters, including the critical success index (CSI), probability of detection (POD), false alarm ratio (FAR), and warning lead time are calculated for each storm type and for each type of warning. Because it was not practical to manually classify these storms (~50,000 entities), a decision tree was trained on a subset of manually classified storms using Quinlan’s C4.5 algorithm. The decision tree was then used to automatically classify storms as being of one of four types: supercellular, linear, pulse or unorganized. It was found that both tornado warnings and severe thunderstorm warnings issued for isolated supercells and convective line storms have higher CSI, higher POD, and lower FAR scores than those issued for pulse and non-organized storms. Lead times were consistently longer for supercell and line storms, while usually very short for pulse and non-organized storms. We conclude that measures of forecast skill are particularly sensitive to the type of storm. Thus, any measurement of forecast skill, such as the year-over-year skill measure of an individual forecast office, has to take into account the types of storms in that office’s warning area in the time period considered.
Stefanie Henry
Mentors: Mike Douglas, Javier Murillo, John Mejia

A Method for Mapping Cloud Forests Using High-Resolution Satellite Imagery

This paper describes an algorithm to describe the distribution of cloud forests in the tropics using Moderate-Resolution Imaging Spectroradiometer (MODIS) imagery. As an important component to global biodiversity, the conservation of tropical cloud forests is a high priority. Unfortunately, it is difficult to map these forests from space because they tend to look similar to other tropical forests. Our approach uses high-resolution satellite imagery (MODIS) to determine the average cloudiness in a region, and to relate these mean cloudiness patterns to the underlying topography. In addition, we developed techniques to distinguish the cloud forests based not only on the total amount of cloudiness, but its seasonal and diurnal variability. Currently, the Central American region was most closely depicted by averaging monthly mean images into annual mean images for both morning (Terra satellite) and afternoon (Aqua satellite). The algorithm was created to narrow the favorable conditions for a cloud forest to exist. For this, each corresponding pixel for each month was averaged together to create a mean annual frequency cloudiness image for Terra and Aqua separately. The difference of maximum and minimum pixel cloud brightness was then calculated in order to locate the lowest variability locations necessary for persistent cloudiness throughout the year. This helps to eliminate any season dependence that may occur with frequent cloudiness locations. Lastly, maintaining the combined overlapped product of Terra and Aqua provided the lowest amount of diurnal variability essential for an ideal cloud forest to exist. This acts to prevent non-ideal daytime dependence. The results demonstrated the combined product of the thresholds of the Central American MODIS sector, and are placed among topography in order to analyze the detailed locations with the 250 meter spatial resolution. The cloud forest regions are along the northeastern mountain slopes of Central America, which are favorable locations due to the persistent trade winds.
Laura Kruep
Mentor: Jim LaDue

Errors in the WSR-88D ZS (Snow) Algorithm

The WSR-88D’s ZS (snow) Algorithm gives current estimates of snow accumulations. However, underestimations from the radar beam overshooting the dendrite production zones and riming zones can occur. Overestimations can occur from bright banding and sub-beam evaporation/sublimation. This study seeks out a method for forecasting when and where these errors might occur and tests its accuracy.

Ground observations used as ground truths and corresponding radar data was collected for three snow events: Wisconsin for March 1-3, 2007, Pennsylvania for March 16-17, 2007, and Colorado for January 29-31, 2005. Comparing the ground truths to the radar’s snow water equivalent (SWE) rate showed the ZS algorithm is inaccurate in its SWE estimates. A method for forecasting the sources of the previously mentioned errors was devised, but after analysis, it too was considered inaccurate. Regressions were then run on the data to see what factors would lead to better predictors of SWE rates. Out of all the data tested, it is believed that basing SWE rates off the distance between the bottom of the radar beam and the -3°C and -12°C layers. If better predictors could be found, improvements can be made to the ZS snow algorithm.
Luke McGuire
Mentors: Ned Snell, Mark Leidner

Development of Methodology and Tools for Determining the Impact of Cloud-Cover on Satellite Sensors

Several studies have addressed the problem of optimizing field of view (FOV) size and sampling area of infrared sensors with the goal of achieving a higher percentage of cloud-free measurements. This study focuses on developing a tool to use global cloud analysis data in order to better understand the effects that different FOV sizes and satellite tracks have on the percentage of cloud-free measurements and the expected altitude of clouds that distort the signal of interest. This paper specifically discusses the situation of a satellite taking nadir measurements with a square FOV. The probability of a cloud contaminated measurement is estimated within 12-km grid boxes, making up a domain centered over the continental United States, using cloud fraction, cloud top altitude, and cloud base altitude values. The data confirms that the probability of a cloud contaminated FOV increases with an increase in FOV size. Compared to seasonal and diurnal variations, data suggests that FOV size has a relatively small effect on the expected value of cloud top and base altitudes. Increased understanding of factors effecting cloud contamination can improve scanning strategies and future satellite-based sensor designs.
Scott Powell
Mentor: Dan O'Hair

Communicating Weather Information to the Public: People's Reactions and Understandings of Weather Information and Terminology

This study examines the general public's use and understanding of weather information in the southwestern United States. Through a paper questionnaire, 769 participants responded to items relating to their uses of and reactions to weather information, sources of obtaining weather data, and emotional response to certain types of weather. In this study, demographic information is used to predict behaviors such as trust in weather data and forecast agencies, planning for severe weather, and ignoring pertinent information. The study also evaluates participants' understanding of weather terminology. In relating predictable behaviors to demographics such as gender, age, race, and state of residence, specific groups of individuals may be targeted to most effectively transmit and convey important weather information in order to reach the largest audeince possible.

Results indicate that behaviors do indeed vary based on demographic factors, especially geography, age, and gender. Particularly, Californians report lower levels of planning, readiness and trust in weather information, including that from the National Weather Service. Additionally, data shows that over one-third of the sample population does not know the difference between a severe weather watch and a warning, clearly signaling the need to educate the public in such matters.
Jessica Ram
Mentors: Paul Schlatter, Bob Johns, Elizabeth Quoetone

National Weather Service Warning Performance Associated with Watches

The National Oceanic and Atmospheric Administration’s (NOAA) National Weather Service (NWS) is responsible for alerting the public to the threat of severe weather by issuing severe weather watches and warnings. The NWS Storm Prediction Center can issue severe thunderstorm watches, tornado watches, or particularly dangerous situation tornado watches while Weather Forecast Offices (WFO) issue severe thunderstorm warnings and tornado warnings. It is vital for these warnings to be accurate and illicit an effective response from those likely to be affected. Although many factors affect the warning decision process, it isolating and examining each factor is an important step towards improving the process. Data were collected using online archives for information on all of the watches, warnings, and events that occurred between January 1, 2006 and April 19, 2006. Several WFOs were given surveys to help determine some of the human factors that might lead to an association between watches and warning issuance. When combining all of the information, watches and watch type appear to be correlated with warning performance in a positive way. The forecasters who issue warnings have also indicated through the surveys that watches influence how forecasters make decisions, especially while issuing warnings.
Bo Tan
Mentors: Mike Douglas, Javier Murillo, John Mejia

Developing and Non-Developing African Easterly Waves and their Relationship to Tropical Cyclone Formation

As African (Tropical) Easterly Waves (AEW) form in eastern to central Africa, convective storms propagate across Northern Africa. Tropical cyclone genesis is a tough question; however, under stronger tropical wave circumstances, hurricanes are more likely to develop. During summer of 2006, two field experiments called African Monsoon Multidisciplinary Analyses (AMMA) and NASA AMMA gathered valuable tropical upper-air data that is available to further study AEW. The observations permit the distinctions of the stronger AEW to the weaker ones. Meridional wind anomaly plot shows strong signals of AEWs’ propagation. The plot will be analyzed to determine the exact date of trough axis passage. Using the dates, Infrared images will be analyzed to determine the differences of developing and non-developing AEWs. This study confirms some of previous studies’ findings on structures of AEW. This study shows that the stronger meridional wind relates to the Cape Verde Storms. All the lesser meridional wind waves did not develop. Satellite imagery shows that the differences of convective cloud fields are not easily noticeable between developing and non-developing waves.
David John Gagne II
Mentors: Amy McGovern, Jerry Brotzge

Automated Classification of Convective Areas in Reflectivity Using Decision Trees

This paper presents an automated approach to classifying storms based on their structure using decision trees. When dealing with large datasets, manually classifying storms quickly becomes a repetitive and time-consuming task. An automated system can more quickly and efficiently sort through large quantities of data and return value-added output in a form that can be more easily manipulated and understood. Our method of storm classification combines two machine learning techniques, k-means clustering and decision trees. Kmeans segments the reflectivity data into clusters and decision trees classify each cluster. We chose decision trees for their simplicity and ability to screen out unimportant attributes.

We used a k-means clustering algorithm derived from Lakshamanan (2001) to divide the reflectivity into different regions. Each cluster was sorted as convective or stratiform based on reflectivity. Each convective cluster was hand labeled at both a general and a specific level. The two general classifications were storm cells and linear systems. The specific classifications for cells were isolated severe, isolated non-severe, and circular Mesoscale Convective System (MCS). The specific classifications for linear systems were trailing stratiform, leading stratiform, and no or parallel stratiform. We used the Waikato Environment for Knowledge Analysis (WEKA), a machine learning suite, to develop the decision trees (Witten and Frank, 2005).

We constructed multiple decision trees with both morphological and reflectivity attributes for both the general and specific classifications. The training and test data sets came from Advanced Regional Prediction System (ARPS) simulated reflectivity data (Xue et al., 2001, 2002, 2003), and we created an additional data set from a collection of composite reflectivity mosaics from the CASA IP1 network (Brotzge et al., 2006). Overall, the best accuracy for the general type trees stayed in the 90% range for all three test sets indicating a very reliable classification tree. By verifying the trees learned on simulated data with observations from the CASA network, we demonstrated that the knowledge gained from simulation can be applied to real situations. For the specific type, the accuracy ranged from 55% to 80% across the test sets, implying additional work is needed for improvement.
Kyle Howe
Mentors: Jerry Brotzge, Keith Brewster

Verification of Low Level Vorticity in a High Resolution Forecast Model Using Radar Data

The Center for Collaborative Adaptive Sensing of the Atmosphere (CASA) first integrative project (IP1) has provided researchers with radar data having high temporal and spatial resolution. These data are currently assimilated to produce short-term mesoscale forecasts of severe weather events. This case study considers a tornadic event that passed within range of the CASA network; multiple high-spatial and temporal resolution forecasts of these severe weather areas are examined. Five model runs were done, each using various combinations of NetRad (CASA) reflectivity, NEXRAD reflectivity, and Doppler radial velocity data. These high-resolution forecasts include areas of low-level vorticity, which were subsequently tracked and compared to verification data from the NEXRAD and CASA radar networks.

This case study provides a baseline for future research in this area as well as showing a direct and useful application of CASA radar data. Most of the models were skilled in predicting the location of these low-level rotation areas even two or three hours out. While it is hard to statistically verify these results, it does show that a high resolution forecast assimilating high-resolution radar data can do quite well in predicting severe weather.
OCS Summer Research Program
John Barr
Mentors: Jeff Basara, Brad Illston

Analysis of Surface Energy Budget Data Over Varying Land-Cover Conditions

Identifying and predicting the influences and changes in surface energy budget data has been a subject of interest, particularly in monitoring latent heat fluxes. Burba and Verman (2005) conducted a field experiment in north central Oklahoma from 1996 until 2000. Evapotranspiration was monitored for the time period over prairie and wheat field terrains, and was compared to a modified Priestly and Taylor model. Data collected from the field experiment was useful in depicting how the magnitude of evapotranspiration overestimation in the modified version of the model was drastically reduced. This paper presents the results of the analysis of surface energy budget data from two different land-cover types from the Little Washita watershed in southwestern Oklahoma: rangeland settings, and winter wheat fields.
Aaron Gleason
Mentors: Jeff Basara, Brad Illston

Analysis and Verification of Soil Moisture Measurements from the Oklahoma Mesonet

The need for accurate and robust measurements of soil water content has grown as numerical weather models have increasingly incorporated soil moisture variables into simulations of land-atmosphere interactions. To meet this need, measurements of soil moisture have been collected by the Oklahoma Mesonet since 1996. Currently, automated soil moisture sensors are installed at over 100 of the Mesonet sites at depths of 5, 25, and 60 cm. From the output of the sensors, observations of variables such as soil matric potential (MP), fractional water index (FWI), and volumetric water content (WC) are calculated.

During the summer of 2007, soil cores were manually extracted from over 20 Mesonet locations at various intervals. Each sampled location was chosen specifically for its soil texture classification. The soil cores were divided into 5 cm increments from the surface to a depth of 30 cm, and 10 cm increments from 30 to 60 cm. From these cores, the volumetric water content of the cores was determined and compared to the values reported by the automated sensors at the Mesonet sites. Preliminary analysis has shown that biases exist in the automated sensors when the soil is both extremely wet and dry. A significant percentage of the manually determined WC had a large degree of variability in the range of WC reported at the same location, depth, and time. To augment the 2007 soil core analysis, data from the 2003 Soil Moisture Experiment (SMEX) was also included.