Citation
Comparing CSU-CHILL polari radar rain rate estimations versus precipitation gauge accumulations for thunderstorm precipitation

Material Information

Title:
Comparing CSU-CHILL polari radar rain rate estimations versus precipitation gauge accumulations for thunderstorm precipitation
Creator:
LaRoche, Kendall
Place of Publication:
Denver, Colo.
Publisher:
Metropolitan State University of Denver
Publication Date:
Language:
English

Notes

General Note:
An undergraduate thesis submitted in partial completion of the Metropolitan State University of Denver Honors Program

Record Information

Source Institution:
Metropolitan State University of Denver
Holding Location:
Auraria Library
Rights Management:
Copyright [name of copyright holder or Creator or Publisher as appropriate]. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
Comparing CSU-CHILL Polari metric Radar Rain Rate Estimations Versus Precipitation Gauge
Accumulations for Thunderstorm Precipitation
by Kendell LaRoche
An undergraduate thesis submitted in partial completion of the Metropolitan State University of Denver Honors Program
December 2011
Dr. Richard Wagner
Dr. Keah Schuenemann
Dr. Megan Hughes-Zarzo
Primary Advisor
Second Reader
Honors Program Director


Kendell LaRoche
Comparing CSU-CHILL Polarimetric Radar Rain Rate Estimations Versus Precipitation Gauge
Accumulations for Thunderstorm Precipitation
Introduction
Weather radar has come a long way over the years, but an important characteristic of radar is its ability to estimation rainfall rate and its advantages to rain gauges. Radar has the ability to detect multiple precipitation events over a wide area and the use of a single radar versus a number of rain gauges are just two of these advantages. However, the rainfall rate observed by a radar is an estimation and should be compared with actual rain gauge data for verification of the radar estimates. For this project, precipitation accumulations from Federal Aviation Administration Automated Surface Observing Station (ASOS) tipping bucket gauges and the National Center for Atmospheric Research (NCAR) wire strain gauges are compared minute-by-minute with rainfall rate and precipitation type data from the CHILL weather research radar. The CHILL radar began as a National Science Foundation (NSF) funded joint project from the University of Chicago and the Illinois State Water Survey, (hence the name CHILL). The different precipitation events examined in this paper occurred near Denver International Airport (DIA) in northeast Colorado during the summer months (June, July, August) of 2009 and 2010. Precipitation accumulation was examined for the KDEN ASOS, NCAR DIA1, NCAR DIA2 and NCAR KFTG gauges located near the vicinity of DIA (Figure 1). These single point observations are used to test the accuracy of the rainfall rate estimations used by the CHILL radar. The CHILL radar is used primarily for research but techniques perfected at this radar are implemented on other weather radars nationwide that are used for weather forecasting and public
storm warnings.


The CHILL radar was originally located in Champaign, IL when it was first constructed in 1970 to conduct various field projects. The Illinois State Water Survey was the site of the first radar observation of a tornadic hook echo in 1953. The group that made this observation would later go on to develop the CHILL program and first radar. After reviews from different universities applying to have the CHILL program moved to their campus the National Science Foundation moved the CHILL facility to its current location at Colorado State University in 1990. The CHILL radar is currently funded by NSF and Colorado State University, and is hosted by the Departments of Atmospheric Science and the Department of Electrical and Computer Engineering (Colorado State University, 2009: CHILL History. Available at [http://www.chill.colostate.cdu/w/CHILL_history]).
ASOS station is a surface weather observing station designed to support weather forecasting and aviation operations. The information these stations gather include:
sky conditions with cloud height and sky cover
visibility
present weather information including fog and haze
atmospheric pressure (sea-level pressure and altimeter settings)
Temperature
dew point temperature
wind speed and direction
precipitation accumulation
significant remarks


This information is updated every minute year round. These stations are often located near airports including DIA (National Weather Service Office of Meteorology, 1999: Automated Surface Observing System. Available online at [http://www.nws.noaa.gov/ost/asostech.html]).
Literature Review
A brief overview of radar is as follows. Radar systems use a transmitter to generate a high-frequency signal that will bounce off of objects away from the radar. The radar antenna sends out the generated signal and receives the reflected echo. Then the signal goes to a receiver where all the calculations are made and data can be interpreted. Finally the data is displayed so the observer can understand with the radar has detected (Rinehart 2008).
One of the more common radar displays is radar reflectivity (Z) which is measured in decibels (dBZ). The intensity of the precipitation is proportional to the strength of the returned radar beam signal. These signals are displayed using a color coordinated system where light blue indicates weak returned signal and very light precipitation (mist and light drizzle) up to dark red or purple for a strongly returned radar signal indicating heavy precipitation (heavy rain and hail <2 inches).
What makes the CHILL radar unique from other forecasting radars is its dual-polarization capability. The CHILL can send out a signal in the horizontal and vertical plane simultaneously. The horizontal orientation is usually displayed as a positive number and the vertical orientation is displayed as a negative number. This radar signal orientation technique is called polarization. A radar that emits polarized signals in both the horizontal and vertical position is called a dualpolarization radar. Raindrops falling from a cloud are not symmetrical circles or even the classic teardrop shape. These falling drops flatten horizontally due to surface tension and falling


terminal velocity which causes a larger horizontally returned radar signal then a vertical signal (Figure 2) (Rinehart 2008). For a dual-polarization radar signals passing through an area of large raindrops the horizontal signal lags behind the vertical signal. The difference in the phase shift between the horizontal and vertical return signals is called phidp, and this quantity will change in the portions of the radar beam path that contain a high rain rate. Specific differential phase or Kdp is the range derivative of phidp and is an important variable in radar rainfall estimation equations. The larger the phidp phase shift the larger the Kdp magnitude (Wang and Chandrasekar 2009).
Polarization can also indicate if there is any hail present in the sampling area using different polarization parameters. Hail from a storm could cause errors when trying to measure rainfall with a rain gauge. A hailstone usually tumbles as it falls and is seen by the radar as a more symmetrically shaped object than a raindrop. The horizontal signal is usually equal to or very close to the vertical signal. One parameter for hail detection in a storm is reflectivity depolarization ratio (ZDR) which is used to determine if the precipitation is symmetrical or not. ZDR is measured in decibels. A measurement of 0 decibels means the precipitation is nearly spherical, positive for horizontally oriented precipitation, and negative for vertically oriented
precipitation. ZDR is defined as Zdr = 10log10() where Zh is the horizontally oriented radar
zv
signal and Zv is the vertically oriented radar signal. ZDR can be used to indicate which storms contain spherical precipitation that might be hail and could cause errors with rainfall measurements (Rinehart 2008). However ZDR alone is not a good indicator of hail because not every kind of spherical object observed by ZDR is hail. Very small drops are nearly spherical and could be mistakenly used as hail. Hail Differential Reflectivity or HDR is another parameter that is used along with ZDR to give a better indication of hail. HDR is defined as HDR = Zh


f{ZDR) where HDR is in dB. Radar observations of HDR greater than 0 dB indicate the presence of hail. Larger HDR values increases the certainty that the radar reflectivity is not from raindrops but hailstones (Aydin et al. 1986). Previous studies by Aydin et al. (1995) using the CHILL radar have shown that collected gauge precipitation measurements were within 10% agreement of the radar rainfall estimation measurement for storms that contain hail. Some of the data that was examined for this project contained hail signatures, and were compared to other data to see if there is a noticeable discrepancy.
Brandes and Wilson (1979) noted that a predominate source of error that affects the radar rainfall estimation is the changes that affect precipitation as it falls from the cloud base. This shift in precipitation that causes the estimated rainfall error relates to the rain rate history such as storm size, storm intensity, duration, and storm speed and direction. Because of this shift the estimated rainfall amount can be different than the actual gauge measurement. Because rain gauges have been around longer and thus studied longer, they are usually the measurement that is assumed to be the correct amount. A radar beam will increases with height as it moves away from the radar source because of the curvature of the earth. The farther away from the source the more the radar beam will ascend and pass through a higher part of the atmosphere. Looking at a precipitation that is high in the atmosphere could affect the rainfall estimation rather than looking at the surface. All the precipitation gauges used for this study were not far enough from the radar site that radar beam height will cause significant errors to radar rainfall estimation. One problem that often affects gauge measurements is the turbulence created by strong horizontal winds flowing over the top of the gauge. Horizontal winds can cause rain to fall at an angle, reducing the amount that falls in the collection bucket. Strong winds could also affect precipitation falling from a storm by changing the location where the precipitation hits the ground. Brandes and


Wilson estimated that the amount of precipitation missing the gauge in strong thunderstorm winds (10-35 m/s) could be as much as 20-40%. The possibility that the gauge data is missing precipitation due to strong winds is accounted for as the accumulated gauge data is plotted alongside the mean wind speed. Areas of stronger wind speeds that have less recorded precipitation in the rain gauges could indicate missing precipitation.
The CHILL radar uses a variety of techniques and algorithms to minimize rainfall errors. Using techniques that utilize the shape of the precipitation rather than the intensity give more accurate rainfall measurements. These techniques depend on Kdp, ZDR, and Zh. These measurements perform well in storms that have heavy precipitation and a high rain rate. The CHILL radar uses a program called CSU-HIDRO (Hydrometeor Identification-based Rain-rate Optimization) that is used to guide rainfall estimations based on precipitation identification. This program has different algorithms that are used for storms that contain hail and other ice particles with storms that do not. This reduces errors from previous algorithms (CSU-ICE algorithm) where small, spherical drops were seen by the radar as spherical ice particles causing rainfall errors for light rain. The algorithm logic is shown in Figure 3. This algorithm is based on three precipitation parameters:
Liquid precipitation
Ice precipitation
Mix precipitation
The numbers at the bottom of Figure 3 indicate which rainfall rate equation the radar would use after the algorithm logic was completed. The rain rate equations are


2. R(Kdp)=40.5(Kdp)'85
3. R(Zh, ZDr)=6.7x10-3(Zh)0'9271 o("0 343ZDR)
4. R(Zh)=0.0170(Zh)-7143
After the most appropriate precipitation parameter is selected a proper equations can be employed to give the most accurate rainfall estimation possible for this radar (Cifelli et al. 2010).
The two different types of rain gauges used to collect precipitation were tipping bucket and weighing bucket rain gauges. The KDEN ASOS station uses the tipping bucket gauge which is a precipitation recording device where precipitation falls/melts through a funnel into a pair of identical buckets that are horizontally balanced. When a predetermined amount of water is collected and tips the bucket over, the spilled water amount and rate are recorded. Then the other bucket moves into position under the funnel for the next movement. This accumulation and tipping process is repeated throughout the precipitation collection event (Gilckman 2000). The Denver International Airport NCAR stations DIA1, DIA2 and KFTG use a wire-strain weighing bucket. This precipitation recording device is a bucket being weighted by a vibrating wire. The frequency of the empty bucket is known and as the buckets weight increases from precipitation the frequency of the vibrating wire changes. This change in frequency can be used to determine the amount of frozen or liquid precipitation collected (GEONOR Inc., 1998: T-200B Series Precipitation Gauge. [Available online at http://www.geonor.com/precipitation_gauge.html]).
Tokay et al. (2009) studied the performance between the tipping bucket and weighing bucket for rain events along the East coast. Their study showed that weighing bucket gauges have higher performance for both daily and monthly rainfall collections, then the tipping buckets. Whereas the tipping buckets are not as dependable for daily rainfall observation but are


reliable for monthly collections. Tokay et al. concluded that since maintenance and data retrieval were the same for both types of rain gauges, the weighing bucket is superior for accurate precipitation observation. This was something that was taken into account in our study to see if the weighing bucket gauges are more accurate then the tipping bucket.
Methodology
Both the gauge data and the CHILL radar data came in text format that was loaded into Microsoft Excel 2007. For each precipitation event the precipitation gauge time intervals that most matched up with the CHILL radar rainfall estimation were selected and any invalid data was changed to 0.
The NCAR wire strain gauges record the precipitation amount value after one minute.
As precipitation accumulated in the bucket the recorded amount increased. The NCAR gauges still had precipitation recorded from previous rain event so the starting precipitation amount was subtracted so the gauge readings at 0 inches. Since the CHILL radar measurements were in millimeters per hour the NCAR gauge units were changed from inches to millimeters (mm). The KDEN ASOS data come from the National Climatic Data Center records and was recorded at individual one minute intervals. To develop a precipitation accumulation curve versus time this data was converted from inches to mm and the data summarized over time to give an accumulation curve that could be compared to the NCAR data.
Even with the versatility of the CHILL radar, the time it would take to make area sweeps and gather data over the rain gauges varied between 1 to 5 minutes. The rain gauges gathered data at one minute intervals which meant the CHILL data was not usually in exact time agreement with the gauge data. Because of the unevenness of the time intervals a time


interpolation procedure was used on the CHILL data. This interpolation was constructed by first indentifying all the whole minute time values between each pair of successive radar sweeps over the rain gauge locations. Then using the two successive radar observations the fractional time location of each whole minute was calculated. These time fractions were then used to interpolate the observed radar data values to whole number values. The interpolation would output the CHILL data in one minute time values so a more accurate comparison could be made between the radar and gauge data (Figure 4). This interpolated data units were mm/hr. To convert the
(771771\
) *
(\hv \
---- 1 min. interval. Like the KDEN ASOS, the CHILL data was recorded at individual
60 min.J
one minute intervals, and the data was summarized to give a comparable accumulation curve.
The data was then plotted on line graphs and scatter plots. The line graphs were plotted to show the visual comparison between the variables. Scatter plots were made to show a mathematical comparison between the precipitation gauge accumulation and the CHILL radar
7 SSE
rainfall accumulation estimation. Microsoft Excel uses the equation R = 1--------where
1 SST
SSE = 'Ziyi ~ 9i)2 ar|d SST = Ziyi y)2 to display the R2 correlation value. This value was plotted on the scatter plot to display the correlation coefficient between the gauge and radar estimation accumulation. R values closer to one indicate a good correlation between the variables while values closer to zero indicate the values are not in good correlation.
Radar images were used from the VCHILL radar database. VCHILL can be found on the CHILL web homepage.


Results
July 30-31 R2 Wind speed Max Speed Remarks
average (m/s) Average (m/s)
DIA1 0.9940 9.62 11.36 -
KFTG 0.9933 5.49 7.01 -
KDEN ASOS 0.9961 - - -
DIA2 - - - Data Corrupted
June 23, 2009 R2 Wind speed average (m/s) Remarks
KDEN ASOS 0.9888 2.73 Hail present
June 26, 2009 R2 Average Wind Speed (m/s) Average Max wind speed (m/s) Remarks
KDEN ASOS 0.7167 2.70 - -
DIA2 0.9899 - - -
KFTG 0.1238 8.22 10.74 Noise in data
June 7, 2009 R2 Average wind speed (m/s) Average 10m wind speed (m/s) Average max wind speed (m/s) remarks
KDEN ASOS 0.8885 2.37 - - -
DIA2 0.9845 - 7.13 - -
KFTG 0.9908 6.72 - 8.41 -
August 6, 2010 R2 Average wind Average max wind remarks


speed (m/s) speed (m/s)
KDEN ASOS 0.8291 - - -
DIA1 0.9619 7.07 8.53 -
The following presents a summary of the results of the gathered data for each rain event. For each precipitation event the R correlation along with wind speed and event remarks are displayed in Tables 1-5. All the figures that show plots are displayed at the end of the paper.
For the July 30-31, 2010 precipitation event the gauge accumulation and CHILL radar estimation accumulation for each station look to be in reasonable agreement with each other.
The line plot for the DIA1 station has the actual gauge accumulation greater than the CHILL estimation accumulation for most of the plot (Figure 5). For stations KFTG and KDEN ASOS the CHILL estimation accumulation is slightly higher than the actual gauge measurement (Figure 6, 7). The scatter plots show a linear trend line that intersects the majority of the data for all 3 gauge stations (Figures 8, 9, 10). Table 1 shows that the correlation for all three gauge stations are strongly correlated with the R correlation value for DIA1 being 0.9940, KFTG being 0.9933, and KDEN ASOS being 0.9961. To see how the CHILL rainfall estimation accumulations compare Figure 11 shows the precipitation accumulation estimation for these stations on one plot. The KDEN ASOS looks to be within reasonable agreement with the DIA1 and DIA2 station estimations. A plot of wind speeds versus actual gauge accumulations to see if there is evidence that strong horizontal winds caused the gauges to record less precipitation than what actually fell. For DIA1 the 3m wind speed (m/s) and maximum wind speed (m/s) were plotted versus the recoded gauge accumulation amount times 10. The gauge accumulation is multiplied by 10 to move the accumulation values higher on the y-axis so they can be more easily compared to the wind speeds. DIA1 the gauge accumulation starts to record precipitation as the peak wind speed occurs at the station around 23:31 UTC to 23:34 UTC (Figure 12). As the wind speed starts to decrease around 23:35 the accumulation has a steep increase as the gauge continues to


take measurements. Because the storm was moving northwest to southeast the accumulation occurred at KFTG at a later time than the DIA and KDEN gauges. For the KFTG gauge the greatest increase in precipitation occurred during most of the highest wind speeds between 00:13 through 00:22 UTC July 31. As the wind speed decreased after 00:21 UTC and the rate of accumulated precipitation decreases (Figure 13).
The June 23, 2009 precipitation event is the only one in this study where hail was reported with the storm. Using data gathered from the CHILL radar a plot of HDR, ZDR, and dBZ versus time shows that between 22:02 UTC and 22:26 UTC HDR had values greater than 5 and ZDR had values close to zero (Figure 14). Both of these measurements together strongly indicate that hail was present during this part of the storm. Figure 15 shows that from 22:04 UTC to 22:13 UTC the KDEN ASOS accumulation rate was less than the rate from 22:14 UTC to 22:22 UTC. During this time that the rain rate was lower the radar reflectivity is around 65 dBZ so the lower rain accumulation rate was not caused by less intense precipitation. There is a possibility that the smaller accumulation rate was caused by a mixture of rain and hail because during this time the HDR was measuring values in the high 30s dB to around 40 dB. ZDR values were only a few tenths of a decibel below zero which indicates the hydrometeor was symmetrical. A mixture of rain and hail could partially block the opening to a tipping bucket gauge decreasing the accumulated amount. The line plot of the CHILL radar rainfall rate estimation versus gauge accumulation shows that the radar estimation over predicted the amount of precipitation that would fall after 22:23 UTC (Figure 16). This over prediction could be the result of the hail decreasing the amount of accumulated precipitation in the gauge. Strong horizontal winds do not appear to be a factor for this case because recorded wind speed was very light. Figure 17 shows both the radar estimated accumulation and the actual gauge accumulation


versus the recorded wind speed, both in five minute intervals. This plots show that the highest recorded wind speed during this time is only 4.12 m/s (9.21 mph) which is not a high source of error from precipitation missing the tipping bucket. The correlation plot of the radar estimated accumulation and the gauge accumulation has a R value of 0.9888. This value is less than that of the July 30-31 event however there is still a strong correlation between the radar estimation and the actual gauge accumulation.
An analysis of the plot for the June 26, 2009 event shows that for the KDEN ASOS the CHILL radar rain estimation is larger and starts sooner then what the gauge recorded.
Precipitation accumulation begins on the CHILL radar estimation as early as 21:14 UTC when
2
the gauge started recorded precipitation measurements at 21:24 UTC (Figure 18). The R correlation value between the radar estimation and the gauge measurements is 0.7167 because the radar estimation accumulation total was 10.70 mm while the gauge only measured a total of 9.40 mm (Table 3). For the DIA2 plot the figure indicates the CHILL radar estimation and gauge measurements are in better agreement over the DIA2 station. This station has the highest correlation for this precipitation event with an R value of 0.9899. The KFTG data is plagued by noise for the gauge accumulation (Figure 19). Precipitation from the cluster of storms moving through the area only brushes over the KFTG gauge as shown in figure 20. As a result this station has a very low correlation of only 0.1238 for the R value. When plotting the CHILL rainfall rate estimation for each station (Figure 21) DIA2 has the largest amount of predicted precipitation while KDEN ASOS shows the second most. KFTG shows very little accumulated precipitation which is expected since the radar images showed the storm just passed this station. The reason that the KFTG station data is strongly affected by noise could be because of strong surface winds generated from the nearby thunderstorm. Strong winds are shown in Figure 22 to


have been reported at the KFTG station around 21:22 UTC and these winds peaked at about 25 m/s or 56 mph. The KDEN ASOS station reported much lighter winds and had no noticeable noise in the data. However the structure of the storm when it passes over the ASOS station looks like a squall line which usually is associated with brief strong winds. The five minute data plot might be spaced too far apart and some important wind speed data missed.
For June 7, 2009 Table 4 shows the R2 correlation values for the KDEN ASOS station, DIA2 station, and KFTG station. The KDEN ASOS has an R2 correlation value of 0.8885. Figure 23 shows that on the plot of the CHILL radar rainfall estimation versus Gauge Accumulation have different accumulations at the end of the time period. The CHILL radar estimated that at the end of the time period the precipitation accumulation would be 22.82 mm when the ASOS gauge only recorded 4.57 mm of precipitation. This wide accumulation difference causes a lower R correlation. DIA2 has an R value 0.9845 because this station also has two different accumulation totals but the difference is not as great as for the ASOS station. The radar estimation accumulation total was 14.11 mm and the gauge accumulation was 5.46 mm. Station KFTG has a smaller difference between the accumulation totals for the radar estimation and precipitation gauge with the R value being 0.9908. Examining the wind speed versus gauge accumulation plot for KFTG shows that the strongest winds for this station occurred between 20:12 UTC and 20:26 UTC (Figure 24). During this time the gauge accumulation continued to increase at a steady rate, so there does not appear to be any loss of precipitation due to strong horizontal winds for this event. DIA2 wind speed was taken at a height of 10 meters above ground. While the winds speed versus gauge accumulation does not give noticeable indication that there is any loss in precipitation from wind, turbulence near the surface could alter the amount of precipitation entering the gauge. Wind measurement at 10m is


taken high enough that turbulence near the surface of the DIA2 gauge is unknown. A 5 minute interval plot of the KDEN ASOS shows that around 19:55 UTC the winds speed increases to 4.12 m/s (9.22 mph) while the gauge accumulation stays the same for about 10 minutes. This slight increase in horizontal wind speed might be enough of a change to induce error in the tipping bucket. Wind error could be the reason for the accumulation differences between the radar and gauge amounts for the KDEN ASOS and DIA2 stations. When comparing the CHILL rainfall estimation amounts between the three stations (Figure 25) the plots look similar with the ASOS gauge reporting 22.82 mm, the DIA2 gauge reporting 14.11 mm, and the KFTG reporting 12.25 mm for their rainfall estimation totals at the end of the time period.
Table 5 shows the R correlation value for the KDEN ASOS station on August 6, 2010 is 0.8291 and the correlation value for the DIA1 station is 0.9619. For the ASOS station the CHILL estimated accumulation is greater than the actual gauge measurements while for the DIA1 station the radar estimation and actual gauge measurements are closer. The line plot for DIA1 even has a time period from 22:54 to 23:03 UTC where the recorded gauge accumulation is greater than the radar estimated accumulation (Figure 26). This does not happen in other plots to the extent seen in this one and could be caused from radar beam overshooting the falling precipitation. DIA1 wind speed versus precipitation accumulation plot shows that in this case the rate of precipitation accumulation could be dependent on the wind speed (Figure 27). During 22:52 UTC to 23:00 UTC when there is a slight increase in the precipitation accumulation rate the wind speed continues to increase. When as the wind speed peaks between 23:01 and 23:05 UTC the precipitation accumulation rate is steady with almost no additional accumulation during this time. Then as the wind speed decrease the precipitation accumulation rate begins to increase again. The increased wind speeds could reduce the amount of precipitation accumulation at


DIAL When comparing the CHILL radar rainfall estimations plots of the ASOS and DIA1 stations the rainfall estimations are very similar for most of the line plot (Figure 28). However between 22:53 thru 23:10 UTC the DIA1 estimated rainfall is greater than the ASOS estimated rainfall. This could possibly be because the returned radar reflectivity over DIA1 was more intense than the returned reflectivity over the ASOS station.
Conclusion
Accurate rainfall accumulation estimation using the CHILL radar could have implications that can be used on warning radars to make rainfall estimation more accurate. Accurate rainfall forecasting is very important for flooding forecasting especially in urban environments. Often to test the radar rainfall estimation accuracy a network of rain gauges is used to compare the results of the gauge accumulation versus the predicted radar amount.
As a final verification all the radar estimation accumulations and all the gauge accumulations for every event were plotted with the R correlation value (Figure 29). The R value is 0.8304 which indicates that the overall CHILL radar rainfall estimated accumulation is strongly correlated with the gauge accumulations. This plot demonstrates a high level of agreement between the radar accumulations and gauge accumulations. This type of comparison is the more traditional way of comparing radar estimated and gauge accumulations. The single event scatter plots demonstrate a high level of agreement between the radar and gauge accumulations at minute by minute intervals. These comparisons can be used to demonstrate how the radar estimated accumulation behaves during the lifecycle of the precipitation event.


The only exceptions that lowers the agreement between the radar and gauge accumulations is when there is hail contamination that affects the gauge, data affected by noise, and high surface winds.
Possible future activities that could follow this project are seeing if any changes could be applied to the CHILL radar that would help make the rainfall estimation more accurate. Comparing one precipitation event with multiple radar equations to see which rainfall accumulation matches the recorded gauge accumulation the most. See if the radar accumulations are well correlated for cold season events.
These single point observations are used to test the accuracy of the rainfall rate estimations used by the CHILL radar. The CHILL radar is used primarily for research but techniques perfected at this radar are implemented on other weather radars nationwide that are used for weather forecasting and public warnings. Accurate rainfall accumulations are very important for flood forecasting especially in urban environments. Radar coverage might extend into an area that lacks a rain gauge network, providing the only way to predict how much precipitation if falling on an area and what actions should be taken.


EdSthAve
Front Range Airport
KFTG
Figure 1
irordaU
Figure 2


Figure 3
radar sweep at time 2
whole whole whole
min #1 min #2 min #3
Figure 4


DIA1 CHILL Radar versus Gauge Accum. for July
30-31, 2010
Time (UTC)
DIA1 chill rainfall est. (mm) DIA1 Gauge Accum. (mm)
Figure 5
KFTG CHILL Radar versus Gauge Accum. for July
30-31, 2010
(N(N(N(N(N(N(N(NOOOOOOOOOOOO Time (UTC)
KFTG CHILL rainfall est. (mm) KFTG Gauge Accum. (mm)
Figure 6


Figure 7


DIA1 CHILL Radar versus Gauge Accum. for July
30-31, 2010
Gauge Accumulation (mm)
Radar versus Gauge plots
----Linear (Radar versus Gauge
plots)
Figure 8
KFTG CHILL Radar versus Gauge Accum. for July
30-31, 2010
Radar versus Gauge Plots
----Linear (Radar versus Gauge
Plots)
Figure 9


KDEN ASOS CHILL Radar versus Gauge Accum. for
July 30-31, 2010
Radar versus Gauge Plots
----Linear (Radar versus Gauge
Plots)
Figure 10
CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for July 30-31, 2010
DIAl chill rainfall est. (mm)
DIA2 CHILL radar rain rate (mm)
KFTG CHILL rainfall est. (mm)
ASOS CHILL radar rainfall est. (mm)
Figure 11


DIA1 Wind Speed versus Precipitation Accum. for
July 30-31, 2010
---- WSpd (m/s)
MaxWSpd (m/s)
Gauge Accum. X10 (mm)
Figure 12
KFTG Wind Speed versus Precipitation Accum. for
July 30-31, 2010
. 20 | 18
M 16
<
a 14
12
| 10
I 8
a>
> 6 T3
ai a 2
1 0 §
i r
Ik

i o fliw
I 1 \
l*L 1 f 1 1 1 1
vRwffjs I 1 I la > An

nr
' ^
o rH cn O rH O) O rH O) O rH O) o rH O) o rH O) o rH
o o q O o LT) O O IS) O o IS) o o IS) o o IS) o o
cn d fN O ro rH 66 fN cn LH ro o d rH LH fN
o rH fN nq nq qt q IS) O T1 rH rsj ro q o o T1 rsj
ro ro ro ro ro ro ro ro o o o o o o o o rH rH rH rH
r\] r\] r\j r\j r\] r\j r\j CNj
Time (UTC)
----WSpd (m/s)
MaxWSpd (m/s)
Gauge Accum. X10 (mm)
Figure 13


Hail Identification using CHILL radar for June 23,
2009
00 m 00 m 00 ro 00 ro 00 ro 00 ro 00 ro 00 ro 00 ro 00
m ,=3- i_n lo O o T1 tl rsj rsj ro ro '=3- '=3- i_n i_n O o
T1 rH rH T1 T1 rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj ro ro
rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj
^HDR>5
Radar Reflectivity (dBZ)
Reflectivity Depolarization Ratio (ZDR)
Hail Differential Reflectivity (HDR)
Figure 14
June 23 ASOS Hail vs KDEN ASOS Gauge Accum.
Radar Reflectivity (dBZ)
Reflectivity Depolarization Ratio (ZDR)
Hail Differential Reflectivity (HDR)
ASOS Gauge Accum. (mm)
Figure 15


KDEN ASOS Radar versus Gauge Accum. for June
23, 2009
UTC
ASOS CHILL radar rainfall est. (mm)
ASOS Gauge Accum. (mm)
Figure 16


KDEN ASOS 5 Minute Intervals for Radar and Gauge Accum. versus Wind Speed for June 23,
2009
o lo o i_n O i_n O i_n O i_n O i_n O i_n O i_n O i_n O
'=3- '=3- lo i_n O O rH T1 rsj rsj ro m '=3- LO i_n O O tl
rH rH T1 T1 rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj ro ro ro
rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj rsj
ASOS Gauge Accum. (mm) Wind Speed (m/s)
Figure 17
Figure 18




Reflectivity Plot Type : dBZ
Radar: CSU-CHILL Elevation : 1.19 Azimuth : 259.56 Date : Fri 26 Jun 09 Time : 21:28:36 UTC Gates : 1019x150m
Marker: Elevation : 0,0 Azimuth : 0.0
Range : 0.0km Height: 0,0km (O.Okft) Value : N/A
Figure 20


CHILL Rainfall Estimation Amounts Between Rain
Gauge Locations for June 26
OOOOOl*1*1*1*1
PWUIMUPUIUIMUPWUIMUJPUIUI
UTC
ASOS CHILL Rainfall Est. DIA2 CHILL Rainfall Est. KFTG CHILL Rainfall Est.
Figure 21


KFTG Wind Speed versus Precipitation Accum. for
June 26, 2009
Winds Speed (m/s)
Max Wind Speed (m/s) Gauge Accum. X10 (mm)
Figure 22


KDEN ASOS CHILL Rainfall Estimation versus
Gauge Accum. for June 7, 2009
E
£_
C
o
+*
(D
<
C
o
+*
(D
[q.
'u
0J
L_
CL
CDCDCDCDCDOOOOOOOOO^H^H^H^H
HHHHHrv|r\JfM(NfM(Nr\JfMr\jrv|fMfMfM
UTC
ASOS CHILL radar rainfall est. (mm)
ASOS Gauge Accum. (mm)
Figure 23
KFTG Wind Speed versus Precipitation Accum. for
June 7, 2009
111111111n 11 rrn111111 ii111i in in 111iiii ii in 11i ii 11 n i ri11 r111 in 11i riii h i 11
CM 00
r\j m m (X) (X) (X) (X) (X) (X)
O r\i oo
LD LD O O
O O O
O to r\] 00 r\j r\j m m
o o o
O tD fM 00 LO LD o O O O tH
HHHHHHrMfMrM(N(N(NfMrMfM(N(N(NrMfM(N
UTC
WSpd (m/s)
MaxWSpd (m/s)
KFTG Gauge Accum. (mm)
Figure 24


CHILL Rainfall Estimation Amounts Between Rain
Gauge Locations for June 7
UTC
ASOS CHILL rainfall est. DIA2 CHILL rainfall est. KFTG CHILL rainfall est.
Figure 25
DIA1 CHILL Rainfall Estimation versus Gauge Accum. for August 6, 2010
UTC
DIA1 CHILL radar rainfall est. (mm)
DIA1 Gauge Accum. (mm)
Figure 26


DIA1 Wind Speed versus Precipitation Accum. for
August 6, 2010
u
u
<
Q.
u
0J
L_
CL
~O
ai
ai
a.
18
16
14
12
10
8
6
4
2
0

aa/Ma

1 \ sJy a
2sJrfy* \
. J rv# VV A
ssa
i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i
CD
HO
CN
fN
rsj
rsj
CD
HO
co
rsj
rsj
CD
HO
o
ro
rsj
CD
i_n
O ro CD CD
ro
rsj
----WSpd (m/s)
MaxWSpd (m/s)
DIA1 Gauge Accum. (mm)
UTC
Figure 27
CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for August 6, 2010
rvirMfMfMfMfMfMfMfMmmmmmmmmm
(N(N(N(N(N(N(N(N(N(N(N(N(N(N(N(N(N(N
UTC
DIA1 CHILL rainfall est. AS OS CHILL rainfall est.
Figure 28


Every Gauge Total versus Every Radar Estimated
Total
Gauge versus Radar Totals
----Linear (Gauge versus Radar
Totals)
Figure 29


Full Text

PAGE 1

Comparing CSU CHILL Polari metric Radar Rain Rate Estimations Versus Precipitation Gauge Accumulations for Thunderstorm Precipitation by Kendell LaRoche An undergraduate thesis submitted in partial completion of the M etropolitan State University of D enver Honors Program December 2011 Dr. Richard Wagner Dr. Keah Schuenemann Dr. Megan Hughes Zarzo Primary Advisor Second Reader Honors Program Director

PAGE 2

Kendell LaRoche Comparing CSU CHILL Polarimetric Radar Rain R ate Estimations Versus Precipitation Gauge Accumulations for Thunderstorm Precipitation Introduction Weather radar has come a long way over the years, but an important characteristic of radar is its ability to estimation rainfall rate and its advantages to rain gauges. Radar has the ability to detect multiple precipitation events over a wide area and the use of a single radar versus a number of rain gauges are just two of these advantages. However the rain fall rate observed by a radar is an estimation and should be compared with actual rain gauge data for verification of the radar estimates For this pr oject precipitation accumulations from Federal Aviation Administration Automated Surface Observing Station (ASOS) tipping bucket gauges and the National Center for Atmospheric Research (NCAR) wire strain gauges are compared minute by minute with rain fall rate and precipitation type data from the CHILL weather research radar The CHILL radar began as a National Science Foundation (NSF) funded joint project from the University of Chicago and the Illinois State Water Survey, (hence the name CHILL). The diff erent precipitation events examined in this paper occurred near Denver International Airport (DIA) in northeast Colorado during the summer months (June, July, August) of 2009 and 2010. Precipitation accumulation was examined for the KDEN ASOS, NCAR DIA1, NCAR DIA2 and NCAR KFTG gauges located near the vicinity of DIA (Figure 1) The se single point observations are used to test the accuracy of the rain fall rate estimations used by the CHILL radar. The CHILL radar is used primarily for research but techniq ues perfected at this radar are implemented on other weather radars nationwide that are used for weather forecasting and public storm warnings.

PAGE 3

The CHILL radar was originally located in Champaign, IL when it was first constructed in 1970 to conduct variou s field projects. The Illinois State Water Survey was the site of the first radar observation of a tornadic hook echo in 1953. The group that made this observation would later go on to develop the CHILL program and first radar. After reviews from differ ent universities applying to have the CHILL program moved to their campus the National Science Foundation moved the CHILL facility to its current location at Colorado State University in 1990 The CHILL radar is currently funded by NSF and Colorado State University, and is hosted by the Departments of Atmospheric Science and the D epartment of Electrical and Computer Engineering ( Colorado State University, 2009: CHILL History. Available at [ http://www.chill.colostate.edu/w/CHILL_history ] ). ASOS station is a surface weather observing station designed to support weather forecasting and aviation operations. The information these stations gather include : sky conditions with cloud height and sky cover visibility present weather inf ormation including fog and h aze atmospheric pressure (sea level p ressure and altimeter settings) Temperature dew point temperature wind speed and direction precipitation accumulation significant remarks

PAGE 4

This information is updated every minute year round. These stations are oft en located near airports including DIA ( National Weather Service Office of Meteorology, 1999: Automated Surface Observing System. Available online at [ http://www.nws.noaa.gov/ost/asostech.html ] ). Literature Review A brief over view of radar is as follows. R adar system s use a transmitter to generate a high frequency signal that will bounce off of objects away from the radar. The radar antenna sends out the generated signal and receives the reflected echo. Then the signal goes to a receiver where all the ca lculations are made and data can be interpreted. Finally the data is display ed so the observer can understand with the radar has detected (Rinehart 2008) One of the more common radar display s is radar reflectivity (Z) which is measured in decibels (dBZ ). The intensity of the precipitation is proportional to the strength of the returned radar beam signal These signals are displayed using a color coordinated system where light blue indicates weak returned signal and very light precipitation (mist and l ight drizzle) up to dark red or purple for a strongly returned radar signal indicating heavy precipitation (heavy rain and hail <2 inches). What makes the CHILL radar unique from other forecasting radars is its dual polarization capability The CHILL can send out a signal in the horizontal and vertical plane simultaneously. The horizontal orientation is usually displayed as a positive number and the vertical orientation is displayed as a negative number. This radar signal orientation techni que is called polarization. A radar that emits polarized signals in both the horizontal and vertical position is called a dual polarization radar. Raindrops falling from a cloud are not symmetrical circles or even the classic teardrop shape. These falli ng drops flatten horizontally due to surface tension and falling

PAGE 5

terminal velocity which causes a larger horizontally returned radar signal then a vertical signal (Figure 2 ) (Rinehart 2008 ). F or a dual polarization radar signals passing through an area of large raindrops the horizontal signal lag s behind the vertical signal. T he difference in the phase shift between the horizontal and vertical return signals is called phidp, and this quantity will change in the portions of the radar beam path that contain a high rain rate. Specific differential phase or Kdp is the range derivative of phidp and is an important variable in radar rainfall estimation equation s The larger the phidp phase shift the larger the Kdp magnitude (Wang and Chandrasekar 2009). Polariz ation can also indicate if there is any hail present in the sampling area using different polarization parameters Hail from a storm could cause errors when trying to measure rainfall with a rain gauge. A hailstone usually tumbles as it falls and is seen by the radar as a more symmetrically shaped object than a raindrop. The horizontal signal is usually equal to or very close to the vertical signal. One parameter for hail detection in a storm is reflectivity depolarization ratio ( ZDR ) which is used to determine if the precipitation is symmetrical or no t. ZDR is measured in decibels. A measurement of 0 decibels means the precipitation is nearly spherical, positive for horizontally oriented precipitation, and negative for vertically oriented precipitation. ZDR is defined as where Zh is the horizontally oriented radar signal and Zv is the vertically oriented radar signal. ZDR can be used to indicate which storms contain spherical precipitation that might be hail an d could cause errors with rainfall measurements (Rinehart 2008 ). However ZDR alone is not a good indicator of hail because not every kind of spherical object observed by ZDR is hail. Very small drops are nearly spherical and could be mistakenly used as h ail. Hail Differential Reflectivity or HDR is another parameter that is used along with ZDR to give a better indication of hail HDR is defined as

PAGE 6

where HDR is in dB. Radar observations of HDR greater than 0 dB indicate the presenc e of hail. Larger HDR values increases the certainty that the radar reflectivity is not from raindrops but hailstones (Aydin et al. 1986). Previous studies by Aydin et al (1995) using the CHILL radar have shown that collected gauge precipitation measurements were within 10% agreement of the radar rainfall estimation measurement for storms that contain hail. Some of the data that was examined for this project contain ed hail signat ures and were compared to other data to see if there is a noticeable discrepancy. Brandes and Wilson (1979) noted that a predominate source of error that affects the radar rainfall estimation is the changes that affect precipitation as it falls from the cloud base This shift in precipitation that causes the estimated rainfall error relates to the rain rate history such as storm size, storm intensity, duration, and storm speed and direction. Because of this shift the estimated rainfall amount can be dif ferent than the actual gauge measurement. Because rain gauges have been around longer and thus studied longer they are usually the measurement that is assumed to be the correct amount. A radar beam will increases with height as it moves away from the ra dar source because of the curvature of the earth. The farther away from the source the more the radar beam will ascend and pass through a higher part of the atmosphere. Looking at a precipitation that is high in the atmosphere could affect the rainfall e stimation rather than looking at the surface All the precipitation gauges used for this study were not far enough from the radar site that radar beam height will cause significant errors to radar rainfall estimation One problem that often affects gauge measurements is the turbulence created by strong horizontal winds flowing over the top of the gauge Horizontal winds can cause rain to fall at an angle, reducing the amount that falls in the collection bucket. Strong winds could also affect p recipitatio n falling from a storm by changing the location where the precipitation hits the ground. Brandes and

PAGE 7

Wilson estimated that the amount of precipitation missing the gauge in strong thunderstorm winds (10 35 m/s) could be as much as 20 40 %. The possibility that the gauge data is missing precipitation due to strong winds is accounted for as the accumulated gauge data is plotted alongside the mean wind speed. Areas of stronger wind speeds that have less recorded precipitation in the rain gauges could indicate missing precipitation The CHILL radar uses a variety of techniques and algorithms to minimize rainfall errors. Using techniques that utilize the shape of the precipitation rather than the intensity give more accurate rainfall measurements. These techn iques depend on Kdp, ZDR and Zh. These measurements perform well in storms that have heavy precipitation and a high rain rate. The CHILL radar uses a program called CSU HIDRO ( Hydrometeor Identification based Rain rate Optimization ) that is used to guid e rainfall estimations based on precipitation identification. This program has different algorithms that are used for storms that contain hail and other ice particles with storms that do not. This reduces errors from previous algorithms (CSU ICE algorith m) where small, spherical drops were seen by the radar as spherical ice particles causing rainfall errors for light rain. The algorithm logic is shown in Figure 3 This algorithm is based on three precipitation parameters: Liquid precipitation Ice precipi tat ion Mix precipitation The numbers at the bottom of Figure 3 indicate which rainfall rate equation the radar would use after the algorithm logic was completed. The rain rate equations are 1. R(K dp Z DR )=90.8(K dp ) 0.93 10 ( 0.169ZDR)

PAGE 8

2. R(K dp )=40.5(K dp ) 0.85 3. R(Z h Z DR )=6.7x10 3(Z h ) 0.927 10 ( 0.343ZDR) 4. R(Z h )=0.0170(Z h ) 0.7143 After the most appropriate precipitation parameter is selected a proper equations can be employed to give the most accurate rainfall estimation possible for this radar (Cifelli et al 2010 ). The two different types of rain gauges used to collect precipitation were tipping bucket and weighing bucket rain gauges. The KDEN ASOS station uses the tipping bucket gauge which is a precipitation recording device where precipitation falls/melts through a funne l into a pair of identical buckets that are horizontally balanced. When a predetermined amount of water is collected and tips the bucket over the spilled water amount and rate are recorded Then the other bucket moves into position under the funnel for the next movement This accumulation and tipping process is repeated throughout the precipitation collection event (Gilckman 2000 ). The Denver International Airport NCAR stations DIA1, DIA2 and KFTG use a wire strain weighing bucket. This precipitat ion recording device is a bucket being weighted by a vibrating wire. The frequency of the empty bucket is known and as the buckets weight increases from precipitation the frequency of the vibrating wire changes. This change in frequency can be used to de termine the amount of frozen or liquid precipitation collected (GEONOR Inc., 1998: T 200B Series Precipitation Gauge. [Available online at http://www.geonor.com/precipitation_gauge.html ] ). Tokay et al (2009) studied the performance between the tipping bu cket and weighing bucket for rain events along the East coast. Their study showed that weighing bucket gauges have high er performance for both daily and monthly rainfall collections, then the tipping buckets Whereas the tipping buckets are not as depend able for daily rainfall observation but are

PAGE 9

reliable for monthly collections. Tokay et al concluded that since maintenance and data retrieval were the same for both types of rain gauges, the weighing bucket is superior for accurate precipitation observat ion This was something that was taken into account in our study to see if the weighing bucket gauges are more a ccurate then the tipping bucket Methodology Both the gauge data and the CHILL radar data came in text format that was lo aded into Microsoft E xcel 2007. For each precipitation event the precipitation gauge time intervals that most matched up with the CHILL radar rainfall estimation were selected and any invalid data was changed to 0. The NCAR wire strain gauges record the precipitation amount value after one minute. As precipitation accumulated in the bucket the recorded amount increased. The NCAR gauges still had precipitation recorded from previous rain event so the starting precipitation amount was subtracted so the gauge readings at 0 inc hes. Since the CHILL radar measurements were in millimeters per hour the NCAR gauge units were changed from inches to millimeters (mm). The KDEN ASOS data come from the Nation al Climatic Data Center records and was recorded at individual one minute inter vals. To develop a precipitation accumulation curve versus time this data was conve rted from inches to mm and the data summarized over time to give an accumulation curve that could be compared to the NCAR data. Even with the versatility of the CHILL radar, the time it would take to make area sweeps and gather data over the rain gauges varied between 1 to 5 minutes. The rain gauges gathered data at one minute intervals which meant the CHILL data was not usually i n exa ct time agreement with the gauge data Because of the unevenness of the time intervals a time

PAGE 10

interpolation procedure was used on the CHILL data. This interpolation was constructed by first indentifying all the whole minute time values between each pair of successive radar sweeps over the rain gauge locations. Then using the two successive radar observations the fractional time location of each whole minute was calculated. These time fractions were then used to interpolate the observed radar data v alues to whole number values. The interpolation would output the CHILL data in one minute time values so a more accurate comparison could be made between the radar and gauge data ( Figure 4 ). This interpolated data units were mm/hr. To convert the CHILL r ainfall unites to mm the equation used was Like the KDEN ASOS, the CHILL data was recorded at individual one minute intervals, and the data was summarized to give a comparable accumulation curve. The data was then plotted on line graphs and scatter plots. The line graphs were plotted to show the visual comparison between the variables. Scatter plots were made to show a mathematical comparison between the precipitation gauge accumulation and the C HILL radar rainfall accumulation estimation. Microsoft Excel uses the equation w here and to display the R 2 correlation value This value was plotted on the scatter plot to display the correlation coeff icient between the gauge and radar estimation accumulation. R 2 value s closer to one indicate a good correlation between the variables while values closer to zero indicate the values are not in good correlation. Radar images were used from the VCHILL radar database. VCHILL can be found on the CHILL web homepage.

PAGE 11

Results July 30 31 R 2 Wind speed average (m/s) Max Speed Average (m/s) Remarks DIA1 0.9940 9.62 11.36 KFTG 0.9933 5.49 7.01 KDEN ASOS 0.9961 DIA2 Data Corrupted June 23, 2009 R 2 Wind speed average (m/s) Remarks KDEN ASOS 0.9888 2.73 Hail present June 26, 2009 R 2 Average Wind Speed (m/s) Average Max wind speed (m/s) Remarks KDEN ASOS 0.7167 2.70 DIA2 0.9899 KFTG 0.1238 8.22 10.74 Noise in data June 7, 2009 R 2 Average wind speed (m/s) Average 10m wind speed (m/s) Average max wind speed (m/s) remarks KDEN ASOS 0.8885 2.37 DIA2 0.9845 7.13 KFTG 0.9908 6.72 8.41 August 6, 2010 R 2 Average wind Average max wind remarks

PAGE 12

speed (m/s) speed (m/s) KDEN ASOS 0.8291 DIA1 0.9619 7.07 8.53 The following presents a summary of the results of the gathered data for each rain event. For each precipitation event the R 2 correlation along with wind speed and event remarks are displayed in Tables 1 5. All the figures that show plots are displayed at the end of the paper. For the July 30 31, 2010 precipitation event the gauge accumulation and CHILL radar estimation acc u mulation for each station look to be in reasonable agreement with each other. The line plot for the DIA1 station has the actual gauge accumulation greater than the CHILL estimation accumulation for most of the plot (Figure 5) For s tations KFTG and KDEN ASOS the CHILL estimation accumulation is slightly higher than the actual gauge measurement (Figure 6, 7) The scatter plots show a linear trend line that intersects the majority of the data for all 3 gauge stations (Figures 8, 9, 10) Table 1 shows t hat the correlation for all three gauge stations are strongly correlated with the R 2 correlation value for DIA1 being 0.9940, KFTG being 0.9933, and KDEN ASOS being 0.9961. To see how the CHILL rainfall es timation accumulations compare Figure 11 shows the precipitation accumulation estimation for these stations on one plot. The KDEN ASOS looks to be within reasonable agreement with the DIA1 and DIA2 station estimations. A plot of wind speeds versus actual gauge accumulation s to see if there is evidence that strong horizontal winds caused the gauges to record less precipitation than what actually fell. For DIA1 the 3m wind speed (m/s) and maximum wind speed (m/s) were plotted versus the recoded gauge accumulation amount times 10 The gauge accumulation is multiplied by 10 to move the accumulation values higher on the y axis so they can be more easily compared to the wind speeds. DIA1 the gauge accumulation starts to record precipitation as the peak wind speed occurs at the s tation around 23:31 UTC to 23:34 UTC (Figure 12 ) As the wind speed starts to decrease around 23:35 the accumulation has a steep increase as the gauge continues to

PAGE 13

take measurements. Because the storm was moving northwest to southeast the accumulation occ urred at KFTG at a later time than the DIA and KDEN gauges. For the KFTG gauge the greatest increase in precipitation occurred during most of the highest wind speeds between 00 :13 through 00:22 UTC July 31. As the wind speed decreased after 00:21 UTC and the rate of accumulated precipitation decreases (Figure 13 ) The June 23, 2009 precipitation event is the only one in this study where hail was reported with the storm. Using data gathered from the CHILL radar a plot of HDR, ZDR and dBZ versus time show s that between 22:02 UTC and 22:26 UTC HDR had values greater than 5 and ZDR had values close to zero (Figure 14 ). Both of these measurements together strongly indicate that hail was present during this part of the storm. Figure 15 shows that from 22:04 UT C to 22:13 UTC the KDEN ASOS accumulation rate was less than the rate from 22:14 UTC to 22:22 UTC. During this time that the rain rate was lower the radar reflectivity is around 65 dBZ so the lower rain accumulation rate was not caused by less intense pre cipitation. There is a possibility that the smaller accumulation rate was caused by a mixture of rain and hail because ZDR values were only a few tenths of a decibel below zero which indicates the hydrometeor was symmetrical. A mixture of rain and hail could partially block the opening to a tipping bucket gauge decreasing the accumulated amount. The line plot of the CHILL radar rainfall rate estimation versus gauge accumu lation shows that the radar estimation over predicted the amount of precipitation that would fall after 22:23 UTC (Figure 16) This over prediction could be the result of the hail decreasing the amount of accumulated precipitation in the gauge. Strong horizontal winds do not appear to be a factor for this case because recorded wind speed was very light. Figur e 17 shows both the radar estimated accumulation and the actual gauge accumulation

PAGE 14

versus the recorded wind speed, both in five minute intervals. This plots show that the highest recorded wind speed during this time is only 4.12 m/s (9.21 mph) which is not a high source of error from precipitation missing the tipping bucket. The correlation plot of the radar estimated accumulation and the gauge ac cumulation has a R 2 value of 0.9888 This value is less than that of the July 30 31 event however there is still a strong correlation between the radar estimation and the actual gauge accumulation. An analysis of the plot for the June 26, 2009 event sh ows that for the KDEN ASOS the CHILL radar rain estimation is larger and starts sooner then what the gauge recorded. Precipitation accumulation begins on the CHILL radar estimation as early as 21:14 UTC when the gauge started recorded precipitation measur ements at 21:24 UTC (Figure 18 ) The R 2 correlation value between the radar estimation and the gauge measurements is 0.7167 because the radar estimation accumulation total was 10.70 mm while the gauge only measured a total of 9.40 mm (Table 3) For the D IA2 plot the figure indicates the CHILL radar estimation and gauge measurements are in better agreement over the DIA2 station. This station has the highest correlation for this precipitation event with an R 2 value of 0.9899 The KFTG data is p lagued by noise for the gauge accumulation (Figure 19 ). Precipitation from the cluster of storms moving through the area only brushes over the KF TG gauge as shown in figu re 20 As a result this station has a very low correlation of only 0.1238 for the R 2 value When plotting the CHILL rainfall rate estim ation for each station (Figure 21 ) DIA2 has the largest amount of predicted precipitation while KDEN ASOS shows the second most. KFTG shows very little accumulated precipitation which is expec ted since the radar images showed the storm just passed this station. The reason that the KFTG station data is strongly affected by noise could be because of strong surface winds generated from the nearby thunderstorm. Strong winds are shown in Figure 22 to

PAGE 15

have been reported at the KFTG station around 21:22 UTC and these winds peaked at about 25 m/s or 56 mph. The KDEN ASOS station reported much lighter winds and had no noticeable noise in the data. However the structure of the storm when it passes over the ASOS station looks like a squall line which usually is associated with brief strong winds. The five minute data plot might be spaced too far apart and some important wind speed data missed. For June 7, 2009 Table 4 shows the R 2 correlation values fo r the KDEN ASOS station, DIA2 station, and KFTG station. The KDEN ASOS has an R 2 correl ation value of 0.8885. Figure 23 shows that on the plot of the CHILL radar rainfall estimation versus Gauge Accumulation have different accumulations at the end of the time period. The CHILL radar estimated that at the end of the time period the precipitation accumulation would be 22.82 mm when the ASOS gauge only recorded 4.57 mm of precipitation. This wide accumulation difference causes a lower R 2 correlation. DIA2 has an R 2 value 0.9845 because this station also has two different accumulation totals but the difference is not as great as for the ASOS station. The radar estimation accumulation total was 14.11 mm and the gauge accumulation was 5.46 mm. Station KFTG h as a smaller difference between the accumulation totals for the radar estimation and precipitation gauge with the R 2 value being 0.9908. Examining the wind speed versus gauge accumulation plot for KFTG shows that the strongest winds for this station occur red between 20:12 UTC and 20:26 UTC (Figure 24) During this time the gauge accumulation continued to increase at a steady rate, so there does not appear to be any loss of precipitation due to strong horizontal winds for this event. DIA2 wind speed was taken at a height of 10 meters above ground. While the winds speed versus gauge accumulation does not give noticeable indication that there is any loss in precipitation from wind, turbulence near the surface could alter the amount of precipitation entering the gauge Wind measurement at 10m is

PAGE 16

taken high enough that turbulence near the surface of the DIA2 gauge is unknown. A 5 minute interval plot of the KDEN ASOS shows that around 19:55 UTC the winds speed increases to 4.12 m/s (9.22 mph) while the gauge accumulat ion stays the same for about 10 minutes This slight increase in horizontal wind speed might be enough of a change to induce error in the tipping bucket. Wind error could be the reason for the accumulation differences between the radar and gau ge amounts for the KDEN ASOS and DIA2 stations When comparing the CHILL rainfall estimation amounts between the three stations (Figure 25 ) the plots look similar with the ASOS gauge reporting 22.82 mm, the DIA2 gauge reporting 14.11 mm, and the KFTG reporting 12.25 mm for their rainfall estimation total s at the end of the time period. Table 5 shows the R 2 correlation value for the KDEN ASOS station on August 6, 2010 is 0.8291 and the correlation value for the DIA1 station is 0.9619. For the ASOS station the CHILL estimated accumulation is greater tha n the actual gauge measurements while for the DIA1 station the radar estimation and actual gauge measurements are closer. The line plot for DIA1 even has a time period from 22:54 to 23:03 UTC where the recorded gauge accumulation is greater than the radar estimated accumulation (Figure 26) This does not happen in other plots to the extent seen in this one and could be caused from radar beam overshooting the falling precipitation. DIA1 wind speed versus precipitation accumulation plot shows that in this case the rat e of precipitation accumulation could be dependent on the wind speed (Figure 27) During 22:52 UTC to 23:00 UTC when there is a slight increase in the precipitation accumulation rate the wind speed continues to increase. When as the wind speed peaks between 23:01 an d 23:05 UTC the precipitation accumulation rate is steady with almost no additional accumulation during this time. Then as the wind speed decrease the precipitation accumulation rate begins to increase again. The increased wind speeds could reduce the am ount of precipitation accumulation at

PAGE 17

DIA1. When comparing the CHILL radar rainfall estimations plots of the ASOS and DIA1 stations the rainfall estimations are very similar for most of the line plot (Figure 28) However between 22:53 thru 23:10 UTC the DIA1 estimat ed rainfall is greater than the ASOS estimated rainfall. This could possibly be because the returned radar reflectivity over DIA1 was more intense than the returned reflectivity over the ASOS station Conclusion Accurate rainfall accumulation estimatio n using the CHILL radar could have implications that can be used on warning radars to make rainfall estimation more accurate. Accurate rainfall forecasting is very important for flooding forecasting especially in urban environments. Often to test the rad ar rainfall estimation accuracy a network of rain gauges is used to compare the results of the gauge accumulation versus the predicted radar amount. As a final verification all the radar estimation accumulations and all the gauge accumulations for every event were plotted with the R 2 correlation value (Figure 29 ) The R 2 value is 0.8304 which indicates that the overall CHILL radar rainfall estimated accumulation is strongly correlated with the gauge accumulations. This plot demonstrates a high level of agreement between the radar accumulations and gauge accumulations. This type of comparison is the more traditional way of comparing radar estimated and gauge accumulations. The single event scatter plots demonstrate a high level of agreement between the radar and gauge accumulations at minute by minute intervals. These comparisons can be used to demonstrate how the radar estimated accumulation behaves during the lifecycle of the precipitation event.

PAGE 18

The only exceptions that lowers the agreement betwee n the radar and gauge accumulations is when there is hail contamination that affects the gauge, data affected by noise, and high surface winds. Possible future activities that could follow this project are seeing if any changes could be applied to the CHI LL radar that would help make the rainfall estimation more accurate. Comparing one precipitation event with multiple radar equations to see which rainfall accumulation matches the recorded gauge accumulation the most. See if the radar accumulations are w ell correlated for cold season events. These single point observations are used to test the accuracy of the rainfall rate estimations used by the CHILL radar. The CHILL radar is used primarily for research but techniques perfected at this radar are implemented on other weather radars nationwid e that are used for weather forecasting and public warnings. Accurate rainfall accumulations are very important for flood forecasting especially in urban environments. Radar coverage might extend into an area that lacks a rain gauge network, providing th e only way to predict how much precipitation if falling on an area and what actions should be taken.

PAGE 19

Figure 1 Figure 2

PAGE 20

Figure 3 Figure 4

PAGE 21

Figure 5 Figure 6 0.00 5.00 10.00 15.00 20.00 25.00 2309 2312 2315 2318 2321 2324 2327 2330 2333 2336 2339 2342 2345 2348 2351 2354 2357 Precipitation Accumulation (mm) Time (UTC) DIA1 CHILL Radar versus Gauge Accum. for July 30 31, 2010 DIA1 chill rainfall est. (mm) DIA1 Gauge Accum. (mm) 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 2309 2316 2323 2330 2337 2344 2351 2358 0005 0012 0019 0026 0033 0040 0047 0054 0101 0108 0115 0122 Precipitation Accumulation (mm) Time (UTC) KFTG CHILL Radar versus Gauge Accum. for July 30 31, 2010 KFTG CHILL rainfall est. (mm) KFTG Gauge Accum. (mm)

PAGE 22

Figure 7 0.00 5.00 10.00 15.00 20.00 25.00 30.00 2309 2317 2325 2333 2341 2349 2357 0005 0013 0021 0029 0037 0045 0053 0101 0109 0117 Precipitation Accumulation (mm) Time (UTC) KDEN ASOS CHILL Radar versus Gauge Accum. for July 30 31, 2010 ASOS CHILL radar rainfall est. (mm) ASOS Gauge Accu. (mm)

PAGE 23

Figure 8 Figure 9 R = 0.994 0.00 5.00 10.00 15.00 20.00 25.00 0.00 5.00 10.00 15.00 20.00 25.00 CHILL Radar Accumulation Est. (mm) Gauge Accumulation (mm) DIA1 CHILL Radar versus Gauge Accum. for July 30 31, 2010 Radar versus Gauge plots Linear (Radar versus Gauge plots) R = 0.9933 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 0 2 4 6 8 CHILL Radar Accumulation Est. (mm) Gauge Accumulation (mm) KFTG CHILL Radar versus Gauge Accum. for July 30 31, 2010 Radar versus Gauge Plots Linear (Radar versus Gauge Plots)

PAGE 24

Figure 10 Figure 11 R = 0.9961 0.00 5.00 10.00 15.00 20.00 25.00 30.00 0 5 10 15 20 25 30 CHILL Radar Accumulation Est. (mm) Gauge Accumulation (mm) KDEN ASOS CHILL Radar versus Gauge Accum. for July 30 31, 2010 Radar versus Gauge Plots Linear (Radar versus Gauge Plots) 0.00 5.00 10.00 15.00 20.00 25.00 2309 2312 2315 2318 2321 2324 2327 2330 2333 2336 2339 2342 2345 2348 2351 2354 2357 Estimated Rainfall Amount (mm) UTC CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for July 30 31, 2010 DIA1 chill rainfall est. (mm) DIA2 CHILL radar rain rate (mm) KFTG CHILL rainfall est. (mm) ASOS CHILL radar rainfall est. (mm)

PAGE 25

F igure 12 Figure 13 0 2 4 6 8 10 12 14 16 18 20 23:09:00 23:12:00 23:15:00 23:18:00 23:21:00 23:24:00 23:27:00 23:30:00 23:33:00 23:36:00 23:39:00 23:42:00 23:45:00 23:48:00 23:51:00 23:54:00 23:57:00 Wind speed versus Precip Accum. TIme (UTC) DIA1 Wind Speed versus Precipitation Accum. for July 30 31, 2010 WSpd (m/s) MaxWSpd (m/s) Gauge Accum. X10 (mm) 0 2 4 6 8 10 12 14 16 18 20 23:09:00 23:16:01 23:22:59 23:30:00 23:37:01 23:43:59 23:51:00 23:58:01 0:04:59 0:12:00 0:19:01 0:25:59 0:33:00 0:40:01 0:46:59 0:54:00 1:01:01 1:07:59 1:15:00 1:22:01 Wind speed versus Precip Accum. TIme (UTC) KFTG Wind Speed versus Precipitation Accum. for July 30 31, 2010 WSpd (m/s) MaxWSpd (m/s) Gauge Accum. X10 (mm)

PAGE 26

Figure 14 Figure 15 20 10 0 10 20 30 40 50 60 70 2138 2143 2148 2153 2158 2203 2208 2213 2218 2223 2228 2233 2238 2243 2248 2253 2258 2303 2308 Hail Identification using CHILL radar for June 23, 2009 HDR > 5 Radar Reflectivity (dBZ) Reflectivity Depolarization Ratio (ZDR) Hail Differential Reflectivity (HDR) 20 10 0 10 20 30 40 50 60 70 2138 2143 2148 2153 2158 2203 2208 2213 2218 2223 2228 2233 2238 2243 2248 2253 2258 2303 2308 UTC June 23 ASOS Hail vs KDEN ASOS Gauge Accum. Radar Reflectivity (dBZ) Reflectivity Depolarization Ratio (ZDR) Hail Differential Reflectivity (HDR) ASOS Gauge Accum. (mm)

PAGE 27

Figure 16 0 10 20 30 40 50 60 2138 2143 2148 2153 2158 2203 2208 2213 2218 2223 2228 2233 2238 2243 2248 2253 2258 2303 2308 Precipitation Accumulation (mm) UTC KDEN ASOS Radar versus Gauge Accum. for June 23, 2009 ASOS CHILL radar rainfall est. (mm) ASOS Gauge Accum. (mm)

PAGE 28

Figure 17 Figure 18 0 5 10 15 20 25 30 35 40 2140 2145 2150 2155 2200 2205 2210 2215 2220 2225 2230 2235 2240 2245 2250 2255 2300 2305 2310 KDEN ASOS 5 Minute Intervals for Radar and Gauge Accum. versus Wind Speed for June 23, 2009 ASOS Gauge Accum. (mm) Wind Speed (m/s) 0 1 2 3 4 5 6 7 8 9 10 11 2100 2102 2104 2106 2108 2110 2112 2114 2116 2118 2120 2122 2124 2126 2128 2130 2132 2134 Precipitation Accumulation (mm) UTC KDEN ASOS CHILL Radar versus Gauge Accum. for June 26, 2009 ASOS CHILL radar rainfall est. (mm) ASOS Guage Accum. (mm)

PAGE 29

Figure 19 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 2100 2102 2104 2106 2108 2110 2112 2114 2116 2118 2120 2122 2124 2126 2128 2130 2132 2134 KFTG CHILL radar rainfall est. versus Gauge Accum. for June 26, 2009 KFTG CHILL radar rainfall est. (mm) KFTG Gauge Accum. (mm)

PAGE 30

Figure 20

PAGE 31

Figure 21 0 2 4 6 8 10 12 14 16 18 2101 2103 2105 2107 2109 2111 2113 2115 2117 2119 2121 2123 2125 2127 2129 2131 2133 2135 Estimated Rainfall Amount (mm) UTC CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for June 26 ASOS CHILL Rainfall Est. DIA2 CHILL Rainfall Est. KFTG CHILL Rainfall Est.

PAGE 32

Figure 22 0 5 10 15 20 25 30 35 21:00:00 21:01:59 21:04:01 21:06:00 21:07:59 21:10:01 21:12:00 21:13:59 21:16:01 21:18:00 21:19:59 21:22:01 21:24:00 21:25:59 21:28:01 21:30:00 21:31:59 21:34:01 Wind speed versus Precip Accum. UTC KFTG Wind Speed versus Precipitation Accum. for June 26, 2009 Winds Speed (m/s) Max Wind Speed (m/s) Gauge Accum. X10 (mm)

PAGE 33

Figure 23 Figure 24 0.00 5.00 10.00 15.00 20.00 25.00 1926 1933 1940 1947 1954 2001 2008 2015 2022 2029 2036 2043 2050 2057 2104 2111 2118 2125 Precipitation Accumulation (mm) UTC KDEN ASOS CHILL Rainfall Estimation versus Gauge Accum. for June 7, 2009 ASOS CHILL radar rainfall est. (mm) ASOS Gauge Accum. (mm) 0 2 4 6 8 10 12 14 16 18 20 1926 1932 1938 1944 1950 1956 2002 2008 2014 2020 2026 2032 2038 2044 2050 2056 2102 2108 2114 2120 2126 Wind speed versus Precip Accum. UTC KFTG Wind Speed versus Precipitation Accum. for June 7, 2009 WSpd (m/s) MaxWSpd (m/s) KFTG Gauge Accum. (mm)

PAGE 34

Figure 25 Figure 26 0.00 5.00 10.00 15.00 20.00 25.00 1926 1932 1938 1944 1950 1956 2002 2008 2014 2020 2026 2032 2038 2044 2050 2056 2102 2108 2114 2120 2126 Estimated Rainfall Amount (mm) UTC CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for June 7 ASOS CHILL rainfall est. DIA2 CHILL rainfall est. KFTG CHILL rainfall est. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 2223 2227 2231 2235 2239 2243 2247 2251 2255 2259 2303 2307 2311 2315 2319 2323 Precipitation Accumulation (mm) UTC DIA1 CHILL Rainfall Estimation versus Gauge Accum. for August 6, 2010 DIA1 CHILL radar rainfall est. (mm) DIA1 Gauge Accum. (mm)

PAGE 35

Figure 27 Figure 28 0 2 4 6 8 10 12 14 16 18 22:22:59 22:25:59 22:28:59 22:31:59 22:34:59 22:37:59 22:40:59 22:43:59 22:46:59 22:49:59 22:52:59 22:55:59 22:58:59 23:01:59 23:04:59 23:07:59 23:10:59 23:13:59 23:16:59 23:19:59 23:22:59 23:25:59 Wind speed versus Precip Accum. UTC DIA1 Wind Speed versus Precipitation Accum. for August 6, 2010 WSpd (m/s) MaxWSpd (m/s) DIA1 Gauge Accum. (mm) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 2234 2237 2240 2243 2246 2249 2252 2255 2258 2301 2304 2307 2310 2313 2316 2319 2322 2325 Estimated Rainfall Amount (mm) UTC CHILL Rainfall Estimation Amounts Between Rain Gauge Locations for August 6, 2010 DIA1 CHILL rainfall est. ASOS CHILL rainfall est.

PAGE 36

Figure 29 R = 0.8304 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Radar Estimated Total (mm) Gauge Total (mm) Every Gauge Total versus Every Radar Estimated Total Gauge versus Radar Totals Linear (Gauge versus Radar Totals)