My NOAA Technical Memorandum NWS HYDRO-30 ^OF Co< ^TES O* METEOR BURST COMMUNICATION SYSTEM --ALASKA WINTER FIELD TEST PROGRAM Office of Hydrology- Silver Spring, Md. March 1976 \ noaa NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION National Weather Service NOAA TECHNICAL MEMORANDA National Weather Service, Office of Hydrology Series The Office of Hydrology (HYDRO) of the National Weather Service (NWS) develops procedures for making river and water supply forecasts, analyzes hydrometeorological data for planning and design criteria for other agencies, and conducts pertinent research and development. NOAA Technical Memoranda in the NWS HYDRO series facilitate prompt distribution of scientific and technical material by staff members, cooperators, and contractors. Information presented in this series may be preliminary in nature and may be published formally elsewhere at a later date. Publication 1 is in the former series, Weather Bureau Technical Notes (TN) ; publications 2 to 11 are in the former ser- ies, ESSA Technical Memoranda, Weather Bureau Technical Memoranda (WBTM). Beginning with 12, publica- tions are now part of the series, NOAA Technical Memoranda, MS. Publications listed below are available from the National Technical Information Service, U.S. Depart- ment of Commerce, Sills Bldg., 5285 Port Royal Road, Springfield, Va. 22151. Price: $3.00 paper copy; $1.45 microfiche. Order by accession number shown in parentheses at end of each entry. TN 44 HYDRO 1 WBTM HYDRO 2 WBTM HYDRO 3 WBTM HYDRO 4 WBTM HYDRO 5 WBTM HYDRO 6 WBTM HYDRO 7 WBTM HYDRO 8 WBTM HYDRO 9 WBTM HYDRO 10 WBTM HYDRO 11 NWS HYDRO 12 NWS HYDRO 13 NWS HYDRO 14 Weather Bureau Technical Notes Infrared Radiation from Air to Underlying Surface. Vance A. Myers, May 1966. 664) ESSA Technical Memoranda (PB-170- Annotated Bibliography of ESSA Publications of Hydrological Interest. J. L. H. Paulhus, February 1967. (Superseded by WBTM HYDRO 8) The Pole of Persistence, Instability, and Moisture in the Intense Rainstorms in Eastern Colorado, June 14-17, 1965. F. K. Schwarz, February 1967. (PB-174-609) Elements of River Forecasting. Marshall M. Richards and Joseph A. Strahl, October 1967. (Superseded by WBTM HYDRO 9) Meteorological Estimation of Extreme Precipitation for Spillway Design Floods. Vance A. Myers, October 1967. (PB-177-687) Annotated Bibliography of ESSA Publications of Hydrometeorological Interest. J. L. H. Paulhus, November 1967. (Superseded by WBTM HYDRO 8) Meteorology of Major Storms in Western Colorado and Eastern Utah. January 1968. (PB-177-491) Robert L. Weaver, Annotated Bibliography of ESSA Publications of Hydrometeorological Interest. J. L. H. Paulhus, August 1968. (PB- 179- 855) Elements of River Forecasting (Revised). Marshall M. Richards and Joseph A. Strahl, March 1969. (PB-185-969) Flood Warning Benefit Evaluation - Susquehanna River Basin (Urban Residences), J. Day, March 1970. (PB-190-984) Harold Joint Probability Method of Tide Frequency Analysis Applied to Atlantic City and Long Beach Island, N.J. Vance A. Myers, April 1970. (PP-192-745) NOAA Technical Memoranda I Direct Search Optimization in Mathematical Modeling and a Watershed Model Application. John C. Monro, April 1971. (COM-71-00616) Time Distribution of Precipitation in 4- to 10-Day Storms — Ohio River "Basin. John F. Miller and Ralph H. Frederick, May 1972. (COM-72-11139) National Weather Service River Forecast System Forecast Procedures. COM-73-10517) December, 1972. (Continued on inside back cover) NOAA Technical Memorandum NWS HYDRO-30 METEOR BURST COMMUNICATION SYSTEM --ALASKA WINTER FIELD TEST PROGRAM Henry S. Santeford Alaskan River Forecast Center, Anchorage, Alaska Office of Hydrology- Silver Spring, Md. March 1976 8 o a a) a UNITED STATES DEPARTMENT OF COMMERCE Elliot I. Richardson, Secretary NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION Robert M White Administrator National Weather Service George P Cressman Director PREFACE The Winter Test Program of the Boeing Meteor Burst Telemetry System was supported by five Federal Agencies under Contract No. 03-5-027-004 between the National Weather Service (contract agency) and the Boeing Aerospace Company. The results of this Program as prepared by Boeing in fulfillment of the contract are presented in the report entitled Meteor Burst Telemetry Winter Test Program - Test Report, Boeing Aerospace Company Document No. D182-10423-1. In addition to the above mentioned report, the Alaskan River Forecast Center, National Weather Service (NWS), performed its own analysis of the system using the raw data obtained during the test program. This analysis was separate from that performed by the Boeing Company. The results of the NWS analysis and the conclusions drawn from the findings are those of the Nation- al Weather Service and do not necessarily represent the feelings of the other agencies participating in the Test Program. The use of manufacturers' iden- tification on equipment used is merely for easy reference and no endorsement nor criticism is implied or intended. In some instances, discrepancies will be noted between the findings presented in the report prepared by the Boeing Company and those presented in this re- port. However, it should be noted that although the basic raw data used in both analyses were the same, the method of reduction, the primary statistic, and the methods of analysis were completely different. Thus, resulting discrepancies could be anticipated. n ACKNOWLEDGEMENT The National Weather Service is grateful for the support and cooperation of the numerous agencies and organizations which assisted with the Test Program upon which this report is based. These organizations include: Alyeska Pipeline Service Company BP Alaska Oil Company Bureau of Land Management Corps of Engineers Federal Aviation Administration Soil Conservation Service U.S. Army Cold Regions Research and Engineering Lab. U.S. Forest Service, Institute of Northern Forestry U.S. Geological Survey Obviously, agencies are composed of people, and it is really these field personnel and administrators we have to thank. Equally obvious, as a quick review of the report will show, they are too numerous to mention individu- ally. It is the work of each and all of them that made this Program as productive as it was. Stuart G. Bigler Director, Alaska Region m TABLE OF CONTENTS PAGE ABSTRACT .1 INTRODUCTION 2 OBJECTIVES 4 TEST PROCEDURES 5 Base Station Operation 10 DATA PROCESSING 12 ANALYSIS OF THE DATA 20 Fairbanks Station 20 Environmental Station 31 AMOS Interfacing 38 Remote Unit Antenna Test 38 Other Stations 38 Aurora Effects 42 GENERAL DISCUSSION 46 SUMMARY AND CONCLUSIONS 50 IV METEOR BURST COMMUNICATION SYSTEM ALASKA WINTER FIELD TEST PROGRAM Henry S. Santeford Alaskan River Forecast Center Anchorage, Alaska ABSTRACT. The Boeing Meteor Burst Communication System was tested in Alaska for a thirty day period beginning February 14, 1975. During the testing program the base station was located in Anchorage and re- mote field units were operated from Fairbanks, the Caribou-Poker Creeks Research Watershed (located approximately 35 miles north of Fairbanks), McGrath, Bethel, Kotzebue, Prudhoe Bay, Dietrich Camp, and Delta Junc- tion with the first two stations serving as the primary source of the data. A total of approximately 40,000 messages were received by the base station from the various remote units. The data were classified according to remote unit identification, date, time, and condition of the message. An analysis of variance was per- formed on the hourly groupings of data for the various stations. These comparisons indicated that: 1) no significant difference could be identified between the data from the beginning and end of the test period, i.e., the anticipated seasonal variation was not apparent; 2) a diurnal variation in the mean waiting time between consecutive good messages occurred such that a minimum value was observed between 09:00 and 10:00 AST and averaged 2.0 minutes; 3) the maximum mean waiting time between con- secutive good messages occurred during the time interval between 17:00 and 18:00 AST and averaged 7.0 minutes; 4) the anticipated sinusoidal variation in the diurnal distribution of the waiting times between con- secutive messages was not observed; 5) no conclusive statements can be made concerning the effects of distance on the distribution of waiting times between consecutive messages. In addition to the studies involving an analysis of variance, the data indicates that: 1) the aurora boreal is may have either an enhancing or a detrimental effect on the operation of the system; 2) the detrimental effects of the aurora result in a multipath communication which scrambles the message, such conditions were observed during approximately 11 percent of the hourly time periods; 3) the data sample obtained during the hourly periods when the adverse aurora effects were present represents nearly half of the total test data; 4) during aurora events, it took nearly 9 transmissions by the remote unit to get one good message through to the base station; 5) during non-aurora periods it took less than 2 attempts to get one good message through to the base station; 6) the system operated properly at temperatures below the specified the manufacturer's limit of -30°C; 7) no major malfunctions of the research equipment were encountered during the test program. INTRODUCTION In February-March, 1975, the National Weather Service, Alaska Region, participated with several agencies in a winter test program of the Boeing Company's Meteor Burst Communication System. The test program was supported by the National Weather Service (NWS); Bureau of Land Management (BLM); Federal Aviation Administration (FAA); U.S. Army Corps of Engineers (COE); Soil Conservation Service (SCS); and the U.S. Geological Survey (USGS). The NWS, Alaskan River Forecast Center served as the contract agent for the project and supervised the field operations and the general testing program. The Boeing Meteor Burst Communication System (MBCS) utilizes meteor trails to reflect or reradiate VHF radio signals between the base station and a number of remote stations used for data acquisition or message communication points. During the operations, the base station transmits a coded message. If a meteor trail is in the proper location, the radio signal will be reflected to a remote station located at some distance from the base station. If the remote station has the address that the base station is trying to contact, then the remote station is switched from a receive to a transmit mode and the data or message from the remote is sent back to the base station along the same communication path. The entire process of the base station probing a remote unit, the remote interpretating the signal, switching to a transmit mode and transmitting the message takes place in approximately 20-60 milli- seconds, depending on message length. If the reflected signal should be received by a remote station which is of a different address or code from the one that the base station is trying to call, or if the reflected signal returns to the earth at a point where no station exists, then nothing happens and the base station keeps sending out its probing message until such time that the required response is obtained. Since there are only a limited number of locations where a meteor trail can be positioned such that it will provide the proper reflecting surface for any given remote unit, and further, since the occurrence of a meteor in one of these required locations is a random event, the. communication between base station and any one remote station is of an intermittent nature reflected by the laws of probability. Because of the earth's rotation about its axis and its elliptical path around the sun, the number of meteors which are potentially available for use as re- flecting surfaces vary with the time of day and the season of the year. The diurnal variation is approximately sinusoidal with a high in the early morning and a low in the late afternoon. The seasonal variation is also approximately sinusoidal with a high in July-August and a low in January-February. With these combined variations, one would expect to find the lowest communication rate in late afternoon during January-February and the highest rate in the early morning hours during July-August. 1/ 1/ For further information on the operation of the Boeing MBCS, the reader is referred to: Leader 3 R.E., Meteor Burst Communication in Advanced Concepts and Techniques in the Study of Snow and Ice Resources } compiled by (Continued) With these known variations in communication rates, the participating agencies wanted to test the system during the low period of operation and to determine if the system was capable of meeting the various communica- tion requirements. The system had previously been tested during July-August, 1974 (the high point of the communication period) and demonstrated that it could more than adequately meet the communication requirements of the partic- ipating agencies during this period. 2/ There were also a number of other items such as environmental effects, remote antenna configurations, and others, which were of interest to the participat- ing agencies. Thus, the following test program was established. (Continued) H.S. Santeford and J.L. Smithy National Academy of Sciences , Washington, D.C.j 1974. For a detailed discussion of the propagation phenomenon 3 see: Sugar 3 G.R. 3 Meteor Burst Signal Distributions, NBS Report 7224 , National Bureau of Standards, Boulder, Colorado, January 30, 1962. Sugar, G.R., Radio Propagation by Reflection from Meteor Trails, Proceedings IEEE, 52:2, February, 1964, 116-136. Proceedings of the IRE, 45:12, December, 1957, 1642-1740. (Special series of 12 papers and 2 letters to the editor) 2/ Santeford, H.S., Meteor Burst Telemetry in Alaska, unpublished Field Test Report, NOAA/NWS, Anchorage, Alaska, September, 1974. OBJECTIVES The primary objective of the winter test program was to evaluate the Boeing MBCS under winter conditions in Alaska and to determine if the system was capable of meeting the communication needs of the various participating agencies. To meet these objectives, the following secondary objectives were established: 1. To evaluate the season variation in communication rates between the July-August period and the February-March period. 2. To determine the effects of the Alaskan winter climate on the operation of the system. 3. To investigate the communication rate from various locations which were not studied during the summer test period program. 4. To determine the adaptability of the system to various NWS equipment such as the AMOS (Automatic Meteorological Observation System). 5. To investigate the effects of various antennas on the operation of the remote stations. TEST PROCEDURES The general procedure consisted of establishing a base station at the BLM offices at Campbell Air Strip, Anchorage, with remote stations located at various sites throughout the state. The base station was operated on a continuous basis. A remote unit was located at the SCS offices in Fairbanks utilizing the same equipment and site location as used during the summer test program. Thus, this station provided the comparison between summer and winter test data. The station was interfaced with two NWS DARDC units which generated a 7-word message with each word containing four digits. Here, a fixed message was used in order that a comparison could be made between what was sent and what was received. This station was operated for the full period of the test program and is subsequently referred to as the "Fairbanks Station. ,; A second remote unit was installed shed (approximately 35 miles north mental Test Station. The station c sensors for monitoring the ambient electronics, the charge on each of recording each time the transceiver sixth data word called "status" was operation of the individual sensors properly, a message word consisting base station. on the Caribou-Poker Creeks Research Water- of Fairbanks) and operated as the Environ- onsisted of a Boeing Hydro -Met unit with air temperature, the temperature of the the two battery units, and a counter for was switched into the transmit mode. A also provided which was a check on the i.e., if the sensors were performing of six "0" would be printed out at the The original test procedure called for the electronics associated with the Environmental Station to be housed in an uninsulated shelter with the batter- ies buried in the snow pack. However, at the time of installation tempera- tures in the Fairbanks area were dropping to the -40°F range and thus exceeded the specified operational temperatures for the equipment (-22°F or -30°C). A change in procedures was thus made. The electronics package and the bat- teries were both buried in the snow pack in such a way that they were in direct contact with the relatively warm ground and covered with approximately 18 inches of snow for insulation. Figure 1 shows the initial installation. Here, only the antenna and the ambient air temperature sensor were fully exposed to the natural environment. Shortly after installation, an operational problem was noted with this station. Upon investigation it was found that a battery lead was disconnected. However, to fully test the equipment and to assure that the loose battery lead was the sole source of the problem, the entire unit was removed from the snow pack and transported to the main field station on the Research Watershed for investiga- tion. Following the testing it was decided to place the unit outside adjacent to the main field station instead of returning it to the snow pack. Thus if further operational difficulties should be encountered, the unit would be much more accessible for examination. There was, however, one major difficulty with the main field station site. Namely, a severe stream icing (aufeis) was occurring in the vicinity and consequently it was not possible to bury the batteries in the snow pack as no "snow pack" existed at the site. It was e o c <0 to 4-> 00 C ■r* (O •r- •r— C en therefore decided to leave the batteries inside the field station. If the unit operated as anticipated, and further, if the weather warmed sufficient- ly, the batteries were, at some later date, to be placed outside adjacent to the electronics unit. The revised site location for this installation is shown in Figure 2. The project proposal called for the Environmental Station to be operated on battery power for the full test period. If the power drain became too great, the batteries were to be removed and temporarily placed on a charger. By equipping the station with sensors on the batteries, and also a counter on the number of transmissions that were made, it would be possible to adequately determine the power drain of the system under actual field conditions. This procedure was originally followed. However, when the station was removed from the snow pack (day 5 of testing), a battery charger was attached to the batteries and a second charger which is an integral part of the Hydro-Met system was also connected to "line power." The system operated in this condi- tion for a period of approximately 2-1/2 weeks. When it was learned that the system was operating on line power, the Watershed technician was instructed to remove the chargers from the system. In doing so, a short circuit was made and a 28-volt current was placed across the 12-volt system. This caused serious damage to the Hydro-Met portion of the system, but the meteor burst portion was unaffected. With the Hydro-Met sensors not operative, the meteor burst message would not get updated and an indicated error was included in the transmitted message. Thus it was not possible to obtain the data necessary to perform an analysis on the power consumption of the unit. However, as will be seen later, the station did provide more information than was originally anticipated. In addition to these two remote stations which were operated as "permanent" installations, a portable communication unit (PCU) was operated from a number of locations for short time periods. Here, the object of the testing was to determine both the difference in communication rates between the summer and winter periods and the communication rate from sites which were potentially different from those used in the summer test program. The PCU was tested from Dietrich Camp and Prudhoe Bay which were stations used during the summer program and also from McGrath, Bethel, and Kotzebue. At each site the unit was installed and operated for a period of from 4 to 18 hours depending on travel arrangements. Both the fixed message generator consisting of eight 4-digit words and the 16 alphanumeric keyboard input were used at all locations Figure 3 shows the geographic location of these various test sites. The PCU was also used as the remote transceiver for interfacing to the AMOS located at the Big Delta Flight Service Station. Under normal operations, the AMOS is interrogated once an hour through a Service A teletype link. During the two days of testing with the MBCS, the AMOS was removed from the Service A link, and updated on a manual basis. The updated message was then transmitted via the meteor burst to the base station in Anchorage. A hard copy of the AMOS output was also provided on a local teletype circuit at the Big Delta Flight Service Station, FAA. During the first day of testing from the AMOS interface system, a standard dipole antenna (approximately 10 feet long) was used. On the second day of CO <0 c 0) E c o J- > c c o (O u o 0) > CM i- 3 E s- O) o i- CO 0) -a (L) CO to 4-> •f— c =J o •^ 3 «/» c 11 DATA PROCESSING The output from the test system was through a Model 33 Teletype with the only output being a hard copy teletype printout. With nearly 40,000 messages re- ceived at the base station during the 30-day test period, it was necessary to manually reduce the data into a computer compatible format for processing and analysis. In Figure 5 a typical section of the original output is shown. Here, it can be seen that the Fairbanks Station (31) and the PCU both had the capability of transmitting a fixed character message. With such messages, the form of the message and not the message itself is of primary importance. Thus, in the data processing a coding system was devised such that each message was assigned a code which placed it in a given category. If the message was good it was assigned a code of 1. Under normal operations such messages are the only ones that would be processed by the mini -computer associated with the base station. These messages can be identified in Figure 5 as those which do not have an asterisk (*) following the message. The bad messages (those followed by an asterisk) were subdivided into three categories each readily distinguishable by its characteristic format. The first classification of bad messages are those caused by meteor trail die-off. With such messages, the first portion of the message is good, and then at some point the signal becomes weak and the message becomes scrambled. Such mes- sages were given a code of 2. This classification is the normal type of 'bad message. " The second of these categories are the bad messages which contain a single bit error in one of the data words other than the last word. Such errors were presumed to be caused by the Department of Commerce, Space Disturbance Monitor- ing Station's radar which is operated in the Anchorage area on a frequency of 49.6 MH Z . The closeness of the operating frequencies of the two systems (MBCS - transmits on 46.6 MH Z , receives on 49.73 MH Z , and the radar transceives on 49.6 MH Z ) coupled with the sharpness of the radar pulse caused its transmission spectrum to spread, giving a strong signal in the base station's receive band. At times, the two systems interfered in such a way that the single bit errors were recorded in the messages received by the MBCS. By changing frequency, this type of error can be eliminated and thus it was classified as a separate category using a code 3. If, however, such an error was encountered in the last word of the message, it is possible that such an error could be either a radar error or an error resulting from the meteor trail die-off. Thus, if such an error (i.e., single bit change in one digit) was encountered in the last word of a message, the message was assumed to be a general bad message and was given a code of 2. Such a distinction does have the tendency to make the classification slightly on the conservative side. The last type of bad message is one that is typical of operations in the northern latitudes, yet according to Boeing, is uncommon for operations at lower latitudes. In Figure 5 it can be seen that some of the messages have a 12 Time Sta ID Message 0132 0132 0133 0133 0133 0133 134 135 136 136 0137 0137 0137 138 0140 0141 0142 0143 144 0146 0146 0146 0146 0147 0147 0147 0147 0147 0147 147 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 7777 7777 7777 ?68S? 7777 7777 7777 7777 7777 7777 7777 7777 7777 7777 7777 9888? 7777 7777 7777 7777 • O • • > f J > 77?7 77 18 01 1?? 7777 7779? 7 7 67? 7319? 7538? 7776 7777 7777 7777 8848? 7777 7777 7777 7777 7777 7777 7777 7777 7777 7777 7777 803? 77; 8? 7777 8883? 1 166 988>? 8S8'S? 8386 8338? 8383? 7774? 7777 7777 7777 SSS? 77?6 3880 7?86 8837 8017 4<44? 777 7 7777 7 77 7 27; 4? 7777 7777 7777 7777 7777 7767 7777 7777 7777 7777 7777 8083 7777 777? 7?77 800; 5>27 057 5 888< 7777 7177 7?9? 3S>6 7 337 7777 4444 2222 2222 2222 8329? 2222? 2222 2222 2222 2222 2222 2222 2222 2222 2222 2222 =7< = ? 2222 2222? > = = = ? :323 2>;; 222 = 2226 = ?<4 2222 2222 3>=2? 2223 2222 2222 2222 2222 2222 2222 2222 2222 2222 2222 < = 95? 2222 022=? 2224? 2602 5:30 :== = ? 42:3 2222 0121 2222 2222 2222 <5: 9? 2223 2222 2222 2222 2222 0<1>? 2222 2222 2222 2222 220: 57? = ? 2222 349 2 :4?0 = 322 = = ? = ? 1322 o « • • C. • ) 3 5226? 2: :9 = = ? = 9 = = ? <16;? 2222 >>>6 * & r^ *1Z. * * * * * * * * * 147 0147 0147 147 0147 0147 0147 0147 147 321* 3567 3104- 2106- 1702 411 44 1 421 3467- 00008* 31 7798? 4;S0 7777 7777 24<=? 3222 2222 * 31 7777 7506 7778? 0777 22<;? :5?5 <522 * 315 >220? ?>1= >>?> >>>6 444:? 444:? 4442? * 31 8338? <677 ?777 7777 >>==? 3>==? ====? * 31 7770? 8333? 8678 8883 ====? ====? 7322 * 31 7?86 >677 7?83? 8888 ====? ==32 ::>=? * 31 777?? 8>77 7777 7779 ===3 22:=? ====? * 31 7777 7777 7777 7788 ==== 2324? 2222 * Figure 5 Typical section of base station teletype printout, i.e. raw data. 13 format which in no way resembles the fixed message format associated with the good messages of code 1. Such messages are the result of a multipath communi- cation occurring during aurora events. Such messages were given a classifica- tion of 4. During the aurora periods it was possible that so many messages could be received by the base station that the buffer capacity of the mini -computer was exceeded. Under such circumstances the individual messages were lost and a buffer or computer overflow was indicated in the output. During such circum- stances a data entry of code 7 "Computer Overflow" was used. In addition to the message codes, two operational codes were used in processing the data. These included a code 5 for off times and a code 6 for dummy mes- sages. The "Dummy" resulted from the specific input format that was used and was not processed in any of the analyses. Utilizing the above coding system and the time of each message as indicated by the base station printout, the remote station ID and the date, each message could be classified into a computer compatible format. Once the grouping was performed on a station and date basis, the individual messages were entered as a 5 digit word - the first 2 digits representing the hour, the next 2 digits the minute, and the last digit the code representing the form of the message. For example, a good message (code 1) received at 10:15 a.m. was recorded as 10151 The messages from the Environmental Station and the AMOS were actual real-time messages where the data itself had meaning. Thus here the actual message, as well as the time and code, were recorded and processed in the analysis. How- ever, a slightly different coding was used. Good messages were still recorded as a code 1, and off times were recorded as a code 5. All bad messages, whether from meteor trail die-off, radar, or aurora were classified as a code 2 "Bad Message." The individual minutes during which computer overflows were recorded were also given a code 2 yet the message entry was recorded as "Computer Overflow." With the Environmental Station, two cases were found where the message contain- ed an error yet the base station computer did not detect a parity error in the received message and thus the message was labeled by the base station as a "good message." Both of these messages were classified in the data processing scheme as a code 3 and are indicated in the output summaries with an "E" where the error was detected. The messages from the Environmental Station had a number of parity checks built into the system which were outputted on the teletype when parity errors were detected. An asterisk (*) following the station ID indicates meteor burst parity, a vertical arrow (t) following the station ID indicates vertical parity in the message, and a horizontal arrow («-) following any of the individual data words indicates horizontal parity in that particular word. A modified system was used to enter these same indicators into the data processing scheme. Here, an asterisk (*) was used for indicating horizontal parity and is printed out with each bad message in the data summaries. Since the occurrence of only meteor burst or only vertical parity was extremely rare, no distinction in the output format was made for such errors. They have, however, been indicated in 14 the initial input data and can be identified if someone should so desire. The occurrence of meteor burst parity and/or vertical parity is indicated in the output summary as a bad message. Figure 6 shows a typical message with parity errors as recorded by the base station teletype. Once the raw data had been reduced to a computer compatible format and check- ed for accuracy, a preliminary analysis was performed on a station by station basis. Since the occurrence of a suitable meteor trail capable of providing a communication link between the base station and any given remote station is a random event influenced in part by the time of day, season of the year, distance between the base and remote stations, and the system parameters such as antenna design and orientation, base station output, etc., it becomes neces- sary to perform a statistical analysis of the data and from this analysis infer what can be expected within the laws of probability concerning the oper- ation of an actual operational system. In its analyses of the data, Boeing Corporation has chosen to use the number of good messages per hour as the controlling statistic. However, it is possible and actually quite common to find a number of meteors grouped together such that there will be a number of good communications in a wery short time interval separated by a much longer period when there are no communications. Thus, in the analyses presented here, the primary statistic used in the comparison is the waiting time between consecutive messages. If the analysis were performed on only good messages, then the waiting time would be the time between consecutive good messages. If the analysis were performed on good + radar messages (Note: If the frequency had been changed, the radar messages would have been good messages.) then the waiting time would be the time between consecutive good messages, or good and radar messages, and so on. Analyses of the distribution of waiting times were made for each station, grouping the data on an hourly basis. Thus, for any station there is a prelim- inary analysis of the waiting times for the first hour of the first day, then the second hour of the first day, and so on throughout the entire test period. An example of one such output is shown in Figure 7. During certain test periods, waiting times in excess of one hour were encount- ered. Such cases involve a somewhat different situation and may cause mis- interpretation in the analysis. Consider the case when a good message is received at 0140 and the next good message is received at 0255. The waiting time between consecutive good messages is 75 minutes and was recorded as occurring in hour 3 of the day (between 2 and 3 a.m.) which corresponds to the time grouping of the latter message. Thus, it is possible to have waiting times in excess of 60 minutes recorded for any given hour. Because of computer limitations, a maximum waiting time used in the analysis was set at 119 minutes or 1 hour and 59 minutes. If the waiting time exceeded this limiting value, it was printed out with the preliminary analysis but was not included in the computations on the distribution of waiting times. The base station was designed to print out the time of each message, but only to the minute. Thus, it is possible to have two or more messages with the same indicated time. For such cases, a waiting time of zero was used in the anal- ysis. Also, for all other possible waiting times, integer numbers (or full minutes) were used as it was not possible to determine the fractional portion of a minute. 15 Good Message 0019 376 STATUS 000000 TEMP IN -00C08 TEMP OUT -GCC06 EATTERY1 01353 EATTERY2 03925 TX COUNT 01705 Bad Mess age 0019 376*» STATUS *- TEMP IN 0008 7 TEMP OUT 0008 5 EATTERY1 00890- EATTERY2 02408 TX COUNT 00932 * Meteor Burst Parity T Vertical Parity - Horizontal Parity Figure 6 Examples of Environmental Station data showing good message and bad message with various types of parity as indicated by the base station monitor. 16 FAIRBANKS STATION DAY 1* HOUR 23 TO 2* NUMBER OF GOUD MESSAGES = 1* NUMBER OF BAD MESSAGES - 11 NUMBER OF RAOAR MESSAGES = 5 NUMBER OF AURORA MESSAGES = 6 NUMBER UF MINUTES WITH COMPUTER OVERFLOW • TUTAL NUMBER OF MESSAGES RECEIVED EXCLUDING PERIODS KITH COMPUTER OVERFLOW * 36. RATIO BETWEEN COOU/TOTAL NUMBER OF MESSAGES RECEIVED * 1*/ 36 » 38.88 PER CENT RATIO BETWEEN GOOD ♦ RADAR/TOTAL NUMBER OF MESSAGES RECEIVED ■ 19./ 36. • 52.77 PER CENT DISTRIBUTION OF WAITING TIMES FOR GOOD MESSAGES WAITINI, TIMES IN MINUTES 0123*56789 10 11 12 NO. MESS. 2313100110011 PER CENT D1ST. 1*.2 21. « 7.1 21.4 7.1 0.0 0.0 7.1 7.1 0.0 0.0 7.1 7.1 CUM. PER CENT 1*.2 35.7 42.8 6*. 2 71.* 71.* 71.* 78.5 85.7 85.7 85.7 92.8 99.9 MEAN WAITING TIME = *.0 MINUTES STANDARD DEVIATION • 3.96 DISTRIBUTION OF WAITING TIMES FOR GOOD • RADAR MESSAGES WAITING TIMES IN MINUTES 0123*56789 NO. MESS. 3*2*121101 PER CENT DIST. 15.7 21.0 10.5 21.0 5.2 10.5 5.2 5.2 0.0 5.2 CUM. PER CENT 15.7 36.8 *7.3 68.* 73.6 8*. 2 89.* 9*. 7 9*. 7 99.9 MEAN WAITING TIME » 2.9 MINUTES STANDARO DEVIATION » 2.52 DISTRIBUTION OF WAITING TIMES FOR ALL MESSAGES WAITING TIMES IN MINUTES 12 3*567 NO. MESS. 12 8 9 3 1 1 2 PER CENI U1ST. 33.3 22.2 25.0 8.3 2.7 0.0 2.7 5.5 CUM. PER CENT 33.3 55.5 80.5 88.8 91.6 91.6 9*.* 99.9 MEAN WAITING TIME - 1.6 MINUTES STANDARD OEVIATION ■ 1.88 Figure 7 Example of computer output for the first step in the analytical procedure used with the Fairbanks data. 17 Following the individual hourly analysis for each station, an analysis of variance was performed on the data for each hourly grouping. Here, the be- tween group variance was compared to the within group variance in testing the null hypothesis - there is no difference between the means of the two groups. If at some specified confidence level the null hypothesis has to be accepted, then it can be inferred that the two groups are samples from the same popula- tion and that there is no apparent difference between the two. If, however, at the specified confidence level the null hypothesis can not be accepted, then it must be interpreted that there is a difference between the two samples and that they are not from the same population. Examples of such comparisons included: a) a comparing of the waiting times for the first hour of the first day for station 1 with the 15th hour of the day for day one and station 1; b) comparing the waiting times for the first hour of the first day for sta- tion 1 with the first hour of the last day for station 1; c) comparing the first hour of the first day for station 1 with the first hour of the first day for station 2; and so on. If the null hypothesis is accepted on the first example test, then there is no apparent diurnal variation. If it is accepted on the second example test, then there is no apparent seasonal variation. And if it is accepted on the third example test, then there is no apparent differ- ence between stations. If, in this last example, the stations are at differ- ent distances from the base station, and the null hypothesis was accepted, then it can be further concluded that there is no apparent distance effect. This comparative process was performed for all possible combinations. For someone unfamiliar with statistical analyses, the above may be sometimes confusing and misinterpreted. When the test is made and the null hypothesis accepted, it can be inferred that at a given confidence level it may be stat- ed that there is no apparent difference between the two sample means. Such a condition may result from the two samples actually being from the same population, i.e., there is no difference, or it may result from the spread of the data within either or both groups being so great that the within group variance masks the between group variance to such a point that the between group variance is no longer recognizable. In such a case the two samples may actually be different yet the spread of the data is so great that the difference can not be satisfactorily identified. If the sample means are normally distributed, the probability of saying there is no difference when there actually is a difference is given by: 1 - confidence level. If a confidence level of 95% is used, then on the average, 5% of the time one would conclude that there is no difference between the groups when in actual- ity there is. The purpose for making the various comparisons between hourly groups of data was a means of determining where the predominant differences occurred as well as a means of reducing the overall number of data samples. Consider for a moment a hypothetical case in which there were 5 stations, each operating 24 hours a day for 30 days. Presuming that there is a diurnal effect, a seasonal effect, and a station effect, then one would be forced to look at 3600 data groups (24/day for 30 days for 5 stations). If each hourly group contained an average of 20 messages/hour then the statistics for each of the individual 3600 groups would be rather weak. If, however, no difference can be recogni- zed between the individual hourly groups for the various days, then the hourly groups for the 30 days can be treated as one sample and the number of 18 samples is thus reduced by a factor of 30 to 120 samples each having an average of 600 data entries per group. If there is no apparent diurnal variation as well as no seasonal variation then the number of samples is reduced to 5, one for each station, having an average number of 14,400 data entries per group. Note the statistics associated with an analysis of this size sample are far more reliable than those performed on a sample size of 20. To go one step further, if it can also be shown that there is no apparent difference between stations, then the entire record of 72,000 messages could be analyzed as one overall group having one mean value and one set of statistics. Once the various analyses of variance were performed and the various data groupings performed, each resulting group was analyzed using the original data inputs. In addition to the above mentioned analyses on the waiting times between consecutive messages, analyses were also performed on the distribution of the various types of messages, (i.e., percentage of good, bad, radar, and aurora), and how these percentages change with the time of day and the season or day of testing. By using the data from the Environmental Station it was also possible to determine the percentage of messages received compared to the number sent by the remote unit. Such comparisons become extremely important in designing power sources for battery operated units. The Environmental Station also provided information on the operating character- istics of the equipment during conditions of extreme cold. 19 ANALYSIS OF THE DATA Fairbanks Station The data from the Fairbanks Station was processed as described above. When an analysis of variance was performed on the consecutive hourly periods for any given day, it was found that in almost all cases the null hypothesis was accepted (95% confidence level), that is, there was no significant difference. Yet, when the data from the early morning hours were compared to the late morning, the afternoon, or the evening hourly data, there was a distinct difference. Thus, it was concluded that the hourly time period was a reason- able grouping for the data and all subsequent analyses were performed on this basis. The hourly data for each day was then compared on a daily basis. Here, the comparisons showed that there was no significant difference between consecu- tive days of testing. The hourly data was then grouped on a weekly basis and the hourly data for the first week compared with the hourly data from the second, third, and fourth weeks of the test period. The procedure was then repeated for all remaining combinations of comparisons such as the hourly data from the second week to the hourly data from the third week and so on. In almost all cases the null hypothesis had to be accepted at the 95% confidence level and in most cases, the acceptance could be made at the 99% confidence level. Thus, it may be concluded that for the winter test period (February 12 through March 14, 1975) there was no apparent seasonal difference in the data and that the entire data set can be treated as being from one population in regard to seasonal variations. The acceptance of the null hypothesis concerning the seasonal variation in the data is significant for several reasons. First, the comparisons were made including all of the hourly periods when the base station was in operation. The periods when interference from the aurora was present, the periods when the base station antenna was rotated away from the Fairbanks station for test- ing from Bethel and McGrath (for the Bethel testing, the antenna was rotated 105°), and all other periods which may have had an adverse influence on the communication rate were included in the comparisons, yet no significant difference could be found when the hourly data were compared on a day to day basis. The second major conclusion that may be drawn from the acceptance of the null hypothesis concerning the seasonal variations is one that is more of inference than actual fact. Boeing has stated that a seasonal variation in the communi- cation rate exists such that a low period occurs in late January - early February and that the maximum occurs in late July - early August. Furthermore the difference between maximum and minimum communication rates expressed in terms of messages per hour is approximately a 4 to 1 ratio. Using the Boeing figures and the sinusoidal relationship as proposed, it can be seen that by mid-March the communication rate should be 1-1/2 times that which occurred in early February. Since this was not observed, it may be inferred that: a) the low point of the curve occurred sometime between mid- to late February 20 and not January-early February as proposed, or b) the relationship for the northern latitudes is more of a step function and not truly a sine curve. However, with the available data, it is not possible to determine the actual relationship. The only conclusion that can be drawn from the data is that there is no apparent seasonal effect in the distribution of waiting times between the beginning and end of the test period. There is, however, a difference between the summer and winter test data. Since the seasonal effect is nonsignificant for the test period, the hourly data for each day of the test period may be grouped into 24 samples, one for each hour of the day. The distribution of waiting times for each of the 24 hourly periods was then determined. From these distributions it was then possible to determine a number of statistics for each of the hourly groups. Two such statistics are shown in Figure 8. There, the mean waiting time and the 90 percentile (90% of the messages were received in the indicated time or less) are plotted versus the time of day. From these curves it can be seen that the mean waiting time varied between 2.0 minutes for the period between 10-11 a.m., and 7.0 minutes for the period between 5-6 p.m. The numbers listed along the top margin of the curve are the number of messages for each hourly group used in defining the individual frequency distributions from which these statistics were drawn. In earlier discussions it was noted that the diurnal variation in message rate expressed in terms of messages per hour was approximately a sine curve with the maximum occurring in the early morning and the minimum occurring in the late afternoon or early evening. If mean waiting time is used as the controlling statistic instead of messages per hour, the diurnal relationship should again be a sine curve but with the positions of the maximum and mini- mum values interchanged. When the data from the summer test program was plot- ted in such a manner, the expected sinusoidal relationship was obtained. However, from Figure 8 it can be seen that the resulting function for the diurnal variation for the winter test data is not sinusoidal. Furthermore, the occurrence of the daily minimum waiting time occurs in the late morning (9-11 a.m.) and not in the early morning (5-6 a.m.) as had been expected. This change in functional relationship and timing of the minimum waiting time is a result of the increased influence of the auroral periods. When the auroral periods are encountered, the waiting time between consecutive good messages was increased. When the aurora and non-aurora periods are grouped together, they have a moderating effect on each other which tends to increase the mean waiting time and alter the expected functional relationship. Since adverse auroral activity was newer encountered during the late morning period, the expected low waiting times were experienced as anticipated. For each of the individual hourly groups there is a frequency distribution curve which defines the percentage of the messages that were received within a given waiting time or less. Two such curves are shown in Figure 9. There, the frequency distributions for the hour with the minimum mean waiting time and the hour with the maximum mean waiting time are shown. Similar curves exist for the remaining 22 hourly groups. For the hourly period with the minimum mean waiting time it can be seen that 35% of the messages were received with waiting times of less than one minute, 60% were received in one minute or less, and 97.2% were received in 10 minutes or less. Similar figures can be seen for the hour with the largest mean waiting time. 21 » « n to I— t o CO « r»» « ii C4 o e in J— f« •A •o CI n « c« c> n n ■o ■© 4-> C 0) o s. CO n 04 « to a* o> CO to cu o s- QJ -Q E \ o o O CM t 1 r cm 1 1 p «3- o CO ■*-> to tO c n JO S- > i- fO O Q t>- «4- to O O) E - 00 _ «* en 10 5 c o c S- 3 CO C CD U $- •r— +■> A3 E <_> 20 - »*-■-*• « »' «^» • • • • Hour with largest mean waitinq time 17:00 - 18:00 159 messages Hour with smallest mean waiting time 09:00 - 10:00 466 messages -j— 20 — r 40 -7— 60 Waiting Time in Minutes Figure 9 Examples of cumulative frequencv distributions for GOOD messages from Fairbanks station. 23 Also shown in Figure 9 is the maximum waiting time between consecutive good messages for the given hourly periods. Thus, it can be seen that for the hourly period with the minimum mean waiting time there was one period when the waiting time between consecutive good messages was 26 minutes even though, on the average, 97.2% of the messages were received within 10 minutes of one another. When all of the hourly periods are considered it was found that the maximum waiting time between consecutive good messages was 107 minutes. This occurred on the morning of the third day of testing when an aurora event prevented the transmission of good messages by saturating the system with a large number of multipath communications, (205 messages were received in the intervening time interval ). In earlier discussions, it was mentioned that in the Anchorage area the Depart- ment of Commerce operates a radar system for monitoring the upper atmosphere for radio propagation. It was also noted that due to the closeness of the operating frequency of the radar and of the MBCS the two systems would, at times, interfere with one another. Since by changing frequencies of the MBCS it would be possible to eliminate this source of error, an analysis similar to that discussed above was performed on the good + radar messages. The relationships from this analysis are similar to those discussed above and the similar graphical relationships are shown in Figures 10 & 11, respectively. A similar analysis was also performed on all of the messages received at the base station. The graphical representations of these are shown in Figures 12 and 13. When Figures 8 and 12 are compared, it will be noted that when all of the messages are considered, the waiting time between messages is greatly reduced and gives some indication of the number of meteor trails or other suitable reflecting surfaces that are potentially available for use. It is also noteworthy that during the test period there were nearly 20,000 messages received from the Fairbanks station. An analysis of the distribution of these messages into the four classifica- tions is of interest in assessing the operation of the system and in the design of power sources for remote units. Figure 14 shows the hourly distri- bution of the percentages of messages that were contained in the good, radar, and aurora classifications. Here, it can be seen that on the average 41.5% of the messages were good, 5.3% contained radar errors, 33.4% contained aurora errors, and 19.8% contained typical meteor burst type errors. Since it is possible through the appropriate engineering design and station oper- ation to eliminate the errors caused by the radar and also those caused by the aurora, a more realistic evaluation of the anticipated station perform- ance can be obtained by considering just the good and meteor burst type error messages. When this is done, the graph of Figure 15 is obtained. Here, a plot of the hourly ratio between good and good plus meteor burst type bad messages is shown. It will be noted that when this is done, the percentage of good messages now varies from 60 to 82% of the total. It should be noted that the relationship shown in Figure 15 is not the same as might be expected at lower latitudes where the occurrence of the aurora interference is not a significant problem. Since during the non-daylight hours, the occurrence of the aurora reduced the number of good messages, the relationship presented in Figure 15 is unduly biased downward. Also, 24 CT> t— I CO re K n CM O m O CM tO * \ c o TO ■M V) V) J* C re -O s- re o to CD C7> fO CO to 0) Q n ■o •o cu ■o K CO Q >> O CM (O O o o E en •r- CU en c - co re 5 c o cu re V) re c s_ Q O CU CD 100 H 80- c o s_ Q_ c CD 3 a i— 3 E 3 O 60- 40- 20- Hour with largest mean waiting time 17:00 - 18:00 174 messages Hour with smallest mean waiting time 09:00 - 10:00 520 messages "T - 20 ~T — 40 60 Waiting Time in Minutes Figure 11 Examples of cumulative frequency distributions for GOOD + RADAR messages from Fairbanks station. 26 r cm ID C\J tO •> K «3 o o o o n ■o CO o •o CD 4-> C CD O s_ CD c o. CD o s: cr. X • to a; 01 fO to QJ OJ E 3 0> O CO <£> _ C\J to -*: c E O S- (A CD cr> «o to to QJ E O Q to OJ «4- E O ••- (1) E - 00 1 r 1 1—1 1 1 CO 1— t 1 1 1 00 1 «3- CD C TO 5 -M 03 •r— s- > (T3 Q CM CD s- en 27 100 i 80 - g 60 o s- GL O c 0) 40 - ITS E 3 O 20 Hour with largest mean waitinq time 17:00 - 18:00 209 messages Hour with smallest mean waitinq time 1:00 - 2:00 1435 messages 20 40 Waiting Time in Minutes 60 Figure 13 Examples of cumulative frequency distributions for ALL messages from Fairbanks station. 28 \ -•" +o \ *s- CM TO CD r^ > CM •r- CTi cd o S- CD to t— 1 CD CO Q-l^ >> +-> "O to o tO 3 o t— t O CD • •f - Vi s_ •• ^3- rtJ to • > CD CO oro CD «a JZ to II -l-> to CD CO «4- E O CO >> »t- tn CD O O 3 jQ S- .Q E 3 CD •i— 3 "O »- to c •r— • «o -o o •« CD •1- fc« -C 4J CO ■»-> +j en CO c tO t— t V) II c .^ o c o •r" JD «>f fd $- m •r" •r— 1_ (O "O fO U- ra > CD ^~ -C •* *3- fO -M ^S c ro $- E • 3 O Lf) •r— S- Q «*- II CD i- 3 cn O CO o to o o «3- CM 29 C\J c CM CO C\J fO •i— s- > i— c 03 O C -r- l- 4-> •i- +-» Q to LT) CD J- o CO o -si- ^uaojad ui (ava + aoo9)/aoo9 10 the occurrence of the radar errors is such that a large percentage of these messages would normally be good messages if the radar interference had not been present. Thus, this also tends to bias the entire curve downward. When these additional factors are considered it would appear that under normal operations, 80 to 90% of the messages received at the base station would be good messages. For an operational system, the base station is normally programmed to process only the good messages. When messages with parity errors (bad messages) are encountered, the system would merely continue to probe the remote station until such time that a good message was received. Therefore, the percentage of good messages is of primary importance in designing the power source for battery operated units and has only minimal effect on the operation of the base station or remote stations operated on line power. Since only the good messages would normally be processed by the base station, the occurrence of bad messages would have no effect on the subsequent processing of the raw data once the message had been accepted by the base station. Environmental Station The Environmental Station was operated as a real-time data acquisition site and therefore, several of the analyses performed on the data from the other stations transmitting a fixed message pattern were not possible. Also, the scheduling sequence, or the time when the station was called by the base station, was changed throughout the test period. Thus, it was necessary to use a slightly modified procedure throughout the analysis of these data. As mentioned earlier, the data were coded into two main categories, good and bad messages. Although the aurora messages were generally distinguishable from the general meteor burst type of bad message, no attempt was made to classify them into a separate category. During the first 15 days of the test period, the Environmental Station was operated on a continuous probing schedule. For this time period, the same type of analyses was performed on the data as was performed on the data from the Fairbanks Station. However, since the message itself contained actual data, the format of the preliminary output was slightly modified. Figure 16 shows a typical example of one such hourly grouping of the data. The first portion of the printout contains the time of each message (good and bad) plus the actual content of the message. The first word is the status or check on the sensors; the second is the outside temperature in °F; the third is the temperature in °F within the electronics unit; the fourth is the voltage of one of the batteries (times 100 volts); the fifth is the voltage of the other battery (times 100 volts); and the last is the number of times since the beginning of the test period that the transmitter has sent a message. (Note: The TX counter would recycle itself at a value of 2047, and the outputted value is the actual value as sent by the remote station. In comparing data from the beginning to the end of the test period, the number of cycles must be included in the computations.) The electronics associated with the Hydro-Met interfacing between the sensors and the meteor burst transceiver were set in such a way that an updated 31 ENVIRONMENTAL STATION - CARIBOU-POKER CMfKS RESEARCH WATERSHED DA It <; 2i 7b TIME MESSAGE GOOD a i 13 18 1367 3927 41 8 4 13 18 1367 3927 41 8 5 13 18 1367 3927 43 811 816 1* 18 1363 3927 45 81a J I* 18 1363 3927 45 817 1* 18 1363 3927 45 825 14 18 1365 3927 48 827 1* 18 1367 3927 50 B29 14 19 1367 3927 50 833 15 21 1367 3927 52 834 835 15 21 1367 3927 52 8*3 16 21 1368 3927 55 8*7 16 22 1368 3927 57 856 856 17 22 1368 3927 58 d57 17 22 1368 3927 58 OAY 10 HOUR 8T0 9 18 1367 3927 15 21 1367 3935 2046« 17 22 1368 3904 t>l NUMBER UF GOOD MESSAGES » 15 NUMBER JF BAO MESSAGES = 3 MTAL NUMBER OF MESSAGES RECEIVED EXCLUDING PERIODS WITH COMPUTER OVERFLOW « NUMBER JF REMOTE TRANSMISSIONS * 20 18. RATIO bLTWEEN GOOD/TOTAL NUMBER OF MESSAGtS RECEIVED * 15/ 18 ' 83.33 PER CENT RATIU BLTWEEN GOOD/TOTAL NUMBER OF MESSAGES SENT * 15/ 20 = 75.00 PER CENT KATIU BcThEEN TOTAL NUMBER OF MESSAGES RECEIVED/TOTAL NUMBER OF MESSAGES SENT > 18./ 20 > DISTRIBUTION OF WAITING TIMES FOR GOOD MESSAGES WAITING TIMES IN MINUTES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 13312000210100000000 0000000000000000001 6.6 20.0 20.0 6.6 13.3 0.0 0.0 0.0 13.3 6.6 0.0 6.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 O.O 6.6 6.6 26.6 46.6 53.3 66.6 66.6 66.6 66.6 79.9 86.6 86.6 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.) CUM. Ptx bNT 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 93.3 99.9 MEAN HAITI JG TIME = 6.2 MINUTES STANDARD DEVIATION • 9.41 NO. MESS. NO. MESS. l-cd lcNI t 1ST. PER CENT ulST. CUM. PER LENT DISTRIBUTION OF WAITING TIMES FOR ALL MESSAGES WAITING TIMES IN MINUTES 1 2 10 11 12 13 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 NO. MESS. NO. MESS. PER CtNT uISl. PER LENT LIST. 38 1 252121102100000000 000000000000000000 11.1 27.7 11.1 5.5 11.1 5.5 5.5 0.0 11.1 5.5 0.0 0.0 0.0 0.0 0.0 d.O 0.0 0.0 0.0 0.0 o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o o.o 5.5 CUM. PER LENT 11.1 38.8 50;0 55.5 66.6 72.2 77.7 77.7 88.8 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 I CUM. PER ctNT 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 94.4 99.9 MEAN WAITING TIME ■ 5.2 MINUTES STANDARD DEVIATION > 8.66 Figure 16 Example of computer output for the first step in the analytical procedure used with the Environmental Station data. 32 message would be provided to the meteor burst system on a 3 minute interval. With the system interrogating on a continuous basis, it was possible to receive several messages at the base station within the 3 minute time interval between updates. When this occurred, the same message would be recorded for all receptions which occurred during the time between updates. In most cases this caused no problem. However, with the counter on the number of transmissions this operation could cause a misinterpretation of the outputted data in that it was possible to show more indicated receptions than transmissions during any given hourly period. The discrepancy caused by the delay in updating the message would then be applied to the statistics for the next hourly period. An analysis of variance was performed on the hourly data from the Environment- al Station for the period when it was operated on a continuous probe basis. Here, as with the Fairbanks Station, no significant difference could be found between the hourly data from the beginning of the test period and that at the end of the period. Thus it may be inferred that there is no apparent season- al variation in the data and that they may be treated as one sample. However, as with the Fairbanks Station, an analysis of variance on a diurnal basis indicated that there was a difference in the hourly data. A graphical representation of this diurnal variation is shown in Figure 17. Also shown in Figure 17 is the diurnal variation of the mean waiting times for the Fairbanks Station for the same time period. It will be noted that the two relationships are not the same. This conclusion is also supported by an analysis of variance using the corresponding hourly data from each station. The distinct difference between the communication rates from the Environment- al Station and the Fairbanks Station, both located at approximately the same distance from the base station and in the same general direction, could presumably be attributed to a number of factors including: a) a difference in site locations; b) the different addressing schemes used on the two stations [the Fairbanks Station used a 5 bit address and the Environmental Station used a 7 bit address]; c) some difference in the internal elect- ronics of the two stations; or d) some unknown factor. As mentioned earlier, during the third week of testing, the field personnel accidently short circuited the 28 volt power source across the 12 volt electrical system and destroyed much of the electronics associated with the Hydro-Met unit. However, the meteor burst unit was unaffected and continued to operate. With the Hydro-Met inoperative, the communication rate from the Environmental Station measured in messages per hour nearly doubled and at least equalled the rate from the Fairbanks Station. However, with the Hydro-Met unit not operating, it was not possible to readily distinguish good from bad messages. A comparison of the total number of messages received from the Environmental Station with the total number of messages received at the Fairbanks Station indicates that there was no significant difference between the two stations. It should be noted that this comparison was made using only a limited amount of data in comparison to what was used in the other portions of the testing, and thus, the small sample size (3 days data) may have adversely influenced the indicated results. If it is assumed that the small sample was representative of the actual conditions, then it may be concluded that the Hydro-Met unit did interfere with the meteor burst system and that the two research units were not compatible. 33 *!- C\J tation. rbanks station. O t/> •!- CM Environmental & 12 for the Fa VO E i— 1 OD messages fro Figures 8, 10, >> o c CM •r - E - U. _1 >-* h- Q O t— i CO ca CO s: o o h- t— o«=t Ul CO cc 10 +J CO >> >> ro CO TJ co H T ro CO CO co ro CO CL CL >> s- J* -t-> CO >> c >> c c >> CO C ro co c s- c S- ro c cl ro -a E rO E Q. E re E CL E E rO <0 E CL JO TO •r- •^ CO •1- ro O CO to ro CO O ro CO O 10 CO RJ ro ro 4-> CO Lu CO c c CO CO CL 01 co c co c ££ CO CO CO >> E O >> CO co CO CO c E CO 13 4-> •1- JZ c E +J E -!-> rO E -a CO T3 rO •r" sz •^ c JO 4-> •r" 1— T3 3 rC +-> (O S- -l-> +J CO CO S- 3 3 •^ CM CO +■> > CL. r— CO 1 — CO c CO c rO 1 CO s_ A3 re ro 1 1 Lc. Jf 1 p— 4- co c > > CM •r— CM •r- r— CM O CO T- CO +-> -0 ■i-> -a c tO JO -O a; -0 •O ro CO -a ro CO ro -a TO -M CD O +-> CO E co co +J +J CO 4J +J jz CO CO (O C O +-> •1- 4J 4-> CO c +-> CO c +J +J ■m -a •r~ ro C O +-» CO CO a E S- CO (U CO CO •1— CO CO •r- CO CO CO CO E O +-> Q. CO Q- CL CO S- Q. co S- CO CL Cl E -(-> E CO CO £= ro CO co c ro ro CO CO CO CO CJJ CO J* E -^ E ^ C e c ro >> ro >> ro -Q JO JO S- T! i- T S- •^ 3 •r- 3 •r- ro (O ro u_ JZ U. -C Lu -^ CO CO 2 +-> CO 1— CO 1 c ro CO CO j^ E c ro >> JO S- s- •r— 3 ro o JC CO (O CO co -^. c ro JO S- rO s_ 3 O J= CO •O CO O E •i— ro S- CO CO Cl ro CO S- _^ O c s- ro 3 JO ro S- 1 ■^ c ro O CO ro CO $- S O s_ ■M 3 CO ro 1 c ro CO CO -^ E C co ro >>"0 JO i— O J- S- •!- •r- 3 $- (O O CO U_ J= Cl -t-> ro S- o CO CO CO CO 3 JO CO N -»-> 5 >> >> ro ro « CO CO CO JZ 3 a CO CO JO •p" O CO s_ JZ J= N 4-> "TO" TO +-> CO 3 3 O •t— i. S- :*£ Q Cl. Q_ 41 Aurora Effects The previous portions of this report clearly indicate that the aurora boreal is had a pronounced effect on the operation of the Test Program. In the report by Boeing, it was noted that the adverse aurora effects were present in approximately 11 percent of the hourly time periods. Being a rather small portion of the total time, Boeing chose to exclude these time periods from their analysis of the data. If time were the only consideration, we could concur with such a decision. However, during the 56 hourly periods when adverse aurora effects were observed in the data from the Fairbanks Station, nearly half of the total number of messages from that station were received. The potential impact of these effects on an opera- tional system is such that in the design and operation of a system for use in the high latitudes special consideration must be given to the effects of the aurora even though it may only affect the operations approximately 11% of the time. The operation of the meteor burst system is based on the premise that the meteor trails located in the 60 to 100km region of the atmosphere act as reflecting surfaces by which a radio signal is reflected or reradiated back toward the earth. The system cannot distinguish between a meteor trail, an aircraft, a satellite, or even the aurora boreal is as the reflecting surface. With the meteor trail, the reflecting surface can be envisioned as being a small flat surface encompassing virtually a point in the sky. With such conditions, there is one path that the transmitted signal can follow in completing its travel between base station and remote unit. With the aurora, the reflecting surface can take several different forms: 1) it can be too weak to serve as a reflecting surface; 2) it can be of moder- ate strength such that it provides a good reflecting surface which encom- passed more than just a point in the sky; 3) it can be of moderate strength yet an irregular surface analogous to a broken mirror lying in the sand where each piece reflects the incident light into a different direction; 4) or it can be of strong intensity (either smooth or irregular) such that its strength is great in comparison to the incident radio signal. Each of these categories has a distinct effect on the operation of the meteor burst system and worth discussing on an individual basis. With the first case, i.e., low level, the presence of the aurora has virtually no effect on the operation of the meteor burst system. However, as the intensity of the aurora increases, a point is reached at which the ionized layer resulting from the aurora is sufficient to be used as a reflecting surface. If the reflecting surface is smooth, there is one small area which will be so positioned that it can reflect the transmitted signal between the base station and any given remote unit. A second small area could act similarly for a different remote unit, and so on. With this condition, the communication between the remote and base stations is enhanced by the presence of a suitable reflective surface. Such an aurora condition can be detected visually and generally exists when the sky or a portion thereof is illuminated with a steady glowing condition. When an oscilloscope is used to monitor the detected HF signal from an antenna, this category of aurora appears as a relatively smooth trace at an elevated intensity. 42 If, however, the aurora is of a moderate intensity, yet is an irregular surface, then like the broken mirror lying in the sand, the signal is reflected from a number of sources. With the meteor burst system there is sufficient time differential between the signals received via the various multipaths that the true signal becomes scrambled and nonrecognizable. Such a message was classified as an "aurora message" in the reduction of the data. To the visual observer, this category of aurora interference results from both the moving curtain and variable intensity displays. When viewed on an oscilloscope, this type of aurora produces a trace at an elevated level which is highly irregular (i.e., has a lot of "grass"). The last major category of aurora interference is that in which the aurora is of such an intensity that its own emitted radiation is large in comparison to the meteor burst signal. In such situations, the meteor burst signal is lost in the overall radiation signal. To the visual observer this category is recognizable as the very intense displays, either stationary or moving. On the oscilloscope they appear as an elevated intensity, generally with a highly irregular trace (i.e., a lot of "grass"). Both the moderate intensity irregular surface and the high intensity aurora can, and do, cause adverse effects to a meteor burst operation. With the moderate intensity irregular surface type, the remote unit receives a signal from the base station requesting a transmission. However, due to the multipath condition, the message received at the base station has parity errors and thus is interpreted by the base station as a "bad" message. The station continues to probe the remote until a favorable response is received. During the process, the remote unit may send several hundred messages per hour before a good message is received at the base station. This excessive activity on the part of the remote unit can cause a severe power drain on systems operating on a self contained power source, and thus must be consid- ered in the design of such stations. A second problem that can occur during a period with moderate intensity irregular surface aurora is associated with the probing signal from the base station. Here, the multipath works in reverse. When the base station sends out its coded message the multipath condition results in a false adddress being received at the remote unit. Thus, even though the remote unit was not being called by the base station, it received a message which it inter- preted as its address and responded. Here again, the main problem is with the excessive power drain on self contained units. With the high intensity aurora interference, there are again two major types of problems. First, when the intensity reaches some threshold value, all communication is interrupted. With this condition, the base station can not recognize a fixed pattern in the incoming messages and thus merely interprets them as no message. The second type of problem with the high intensity aurora interference, is that the remote unit can again receive a signal which it interprets as its address and attempts to send a message. The above discussion is somewhat of an over-simplification of the actual problems encountered in the field. An examination of the base station output 43 data and visual observation in the field show that the condition of the aurora, and thus its interference pattern, is a continuously changing phenomenon that may not be affecting all remote units similarly at a given moment. The data clearly indicates that periods of multipath conditions were of a variable time duration, and often were interrupted by abnormally high communication rates of good messages and/or periods of no communi- cation. Several conditions noted in the data are of significance in attempting to ascertain the full impact of the aurora interference and also in consider- ing potential methods for overcoming it. On a number of occasions, multi- path aurora conditions were noted in the data from the Fairbanks Station but were absent in the messages from the Environmental Station. However, at the same time, the Environmental Station either did not respond or responded very infrequently (in comparison to the large number of multipath communications from the Fairbanks Station). When multipath aurora inter- ference was observed from the Big Delta Station, a similar situation was observed from the Fairbanks Station. Once the Hydro-Met unit associated with the Environmental Station was made inoperative, the Environmental Station and the Fairbanks Station, both responded similarly to multipath aurora interference. When multipath aurora interference was observed at the Fairbanks Station, no interference was observed in the transmissions from Prudhoe Bay. These and other similar observations of irregularities in the data have raised a number of questions. Although answers to many of these questions are currently not available, several questions are listed here in that they give further insight into some of the factors that must be considered in design- ing a protective mechanism to overcome the aurora interference. 1. Is it possible that on repeated occurrences a multipath aurora condition could exist in such a way that it would interfere with the Fairbanks Station, yet not affect the Environmental Station located some 35 miles away? 2. Is it possible that the difference in address codes (5 bit at the Fairbanks Station and 7 bits at the Environmental Station) combined with the Hydro-Met interference desensitized the Environmental Station to such a point that it was not "hearing" the weak multipath signals calling for a transmission which were obviously heard by the Fairbanks Station? 3. When the remote unit was being tested from Prudhoe Bay, a period of approximately 2 hours was encountered when there were a large number of transmissions from the remote unit. Unfortunately, during the same time, Fairbanks was experiencing a multipath condition which resulted in a jamming of the paper on the teletype output. Thus the base station data was lost for this period. In light of the conditions observed with the Environmental Station, was this activity at Prudhoe Bay the result of "false triggering" caused by aurora interference or, due to the consider- able difference in distance, was it using the aurora as a suitable reflec- tive surface? Unfortunately, there is no way of telling what actually was happening. 44 From the above, it should be apparent that during aurora events, some mechanism or procedure must be available whereby the effects of the multi- path communication can be filtered out of the system. Boeing has suggested that one alternative is to merely shut down the base station. This would work. However, it would also mean that there would be a considerable time period when the entire system was not operative merely because aurora effects were present at a few of the remote units. It should also be apparent that the removal of a particular remote unit from the base station polling sequence will not insure that an affected station will not respond. This was clearly shown with the Environmental Station which did respond during aurora events even though it was not being polled. Therefore, any protective device which may be conceived for removing the adverse interference effect of the multipath aurora must be concerned with the remote receiver. Here, it is conceivable that the required protection can be provided by: 1) a more complicated addressing scheme; 2) a multi- stage sequential addressing; or 3) a simple switch which when a preset number of transmissions are made in a given time window, the remote unit is locked into a standby mode for a predetermined time interval. 45 GENERAL DISCUSSION The experience gained from the Summer and Winter Test Programs revealed a number of points concerning the operation of the Boeing Meteor Burst Communication System which are of significent value in designing an operational system and which are not yet recorded in any of the previous technical papers concerning meteor burst systems. The conclusions presented in this section are the result of trial -error, cause-effect, and pure deductive reasoning. The scientific studies necessary to support or dis- approve these conclusions have, as yet, not been performed. However, it is felt that the field observations are sufficient to support these conclusions. On several occasions during both the Summer and Winter Test Programs it was found that when the antenna for the remote unit was placed on a support mast 25-30 feet high and anchored adjacent to a building with a metal roof, the communication rate was seriously hampered — 1/4 to 1/10 the expected value. However, when the mast was moved as little as 8-10 feet from the building, the anticipated communication rate was obtained. This condition occurred even though the separation between metal roof and antenna was theoretically great enough to eliminate the observed interference. Thus, it would appear that when a remote unit is to be housed in a building with a metal roof it is better to provide a free standing antenna mast a short distance from the building than to use one anchored to the building such that the antenna is positioned directly above the metal roof. The antenna used with the base station was a double yagi having a beam width of approximately 50 degrees. If the antenna was pointed due north, it would illuminate an area of sky approximately 25° on either side of north, or from azimuth 335° through 0° to 25°. The area of sky from approximately 25° through 180° to 335° was either not illuminated or illuminated with a greatly reduced signal. When such antennas are used in an operational system it can be seen that in order to communicate with two remote stations separated by 120°, one of two base station antenna configurations must be used. Either the base station must be equipped with a rotor such that the antenna can be moved from one azimuth to the other, or two separate antennas must be used. With the rotor system, communication with the first station will be lost when the antenna is moved into position to communicate with the second remote unit. With the double antenna system, the base station output power must be doubled in order to maintain the same communication efficiency. If 1000 watts of out- put power was being supplied to one single antenna, then to keep the same efficiency with a two antenna system 1000 watts of output power must still be supplied to each of the two antennas for a total of 2000 watts for the station If it is further assumed that the axis of both of the antennas are oriented directly toward each of the remote units, i.e., 120° apart, then there would still be an area 70° wide between the two units which was not being illumin- ated or illuminated with a greatly reduced signal. From the above example it can be seen that if it is desirable for a base station to have the capability of communication with remote units spaced at 360° about the station, then either the rotor system or 8 fixed base antennas 46 each having a beam width of 50° would have to be used. In the latter case the total base station output power would have to be (from our example) 8000 watts or 1000 watts per antenna to maintain the same relative power per antenna. Needless to say, the increased station capacity as well as the necessary switching and other support mechanism will greatly increase the cost. As an alternative, the rotor system also has an additional cost over a single fixed antenna system, has the increased operational mechanisms for synchronizing the polling sequence with the movement of the rotor, and has the distinct disadvantage of sequentially losing communication with the various remote units as the axis of the antenna moves around the circle. The experience gained during both the Summer and Winter Test Programs indicates that the actual problems associated with the orientation of the base station antenna are not nearly as great as might first be envisioned. In the Summer Test, it was shown that the communication rate from a station located 45° from the axis of the antenna was statistically the same as a station located along the axis of the antenna. Thus the effective beam width was approximately 90° and not the specified 50°. During the Winter Test, it was shown that when the antenna was orientated toward Fairbanks, the reception from Kotzebue, located 55° from the axis of the antenna, was statistically the same as that from Fairbanks. If the distance effect was truly present, then the reception from Kotzebue should have been greater than that from Fairbanks and the orientation of the antenna had a countering effect and reduced the rate to a level statistically the same as that from Fairbanks. A third example which is less significant yet worth mentioning was that encountered when the base station antenna was oriented toward Bethel and reception was received from Fairbanks, 105° removed from the axis of the antenna. Although the total number of messages received from the Fairbanks Station during this portion of the testing was significantly reduced, the waiting time between consecutive good messages was statistically the same as when the antenna was oriented toward the Fairbanks Station. This leads us to believe that the unit power per antenna as used in the test (600-1000 watts) is not mandatory. These limited examples suggest that the "effective beam width" of the base station antenna is much greater than the specified beam width. This con- clusion is also supported by a theoretical consideration of the operation of the system. The basic premise by which the system operates implies that the incident radio signal is reflected or reradiated from the meteor trail in such a way that a change in the vertical path of the signal occurs. Thus, it is equally reasonable to assume that a change in the horizontal path of the signal can, and does, also occur. This change in the horizontal path of the signal results in the effective beam width of the antenna being signifi- cantly different from the specified or illuminating beam width of the antenna. Although limited data is currently available, it would appear that the full 360° around a base station could adequately be served by 4 antennas spaced at 90° intervals all powered by a single 4000 watt transceiver. If the remote units are strategically located, it is possible that as few as three antennas 47 could serve the entire 360° without an appreciable loss in efficiency. With such an arrangement, the cost of the additional fixed-base antennas would not be significantly different from that of a rotor system or possibly less, and would be more than offset by the capability of continuous potential communication with any of the remote units. In considering the design of a base station, either location or antenna con- figuration, it is apparent that the distance effect as observed during the Summer Test Program and as recorded by Boeing, is of minor consideration. Since under most operations, the same requirements would be applied to the near stations as are applied to the more distant stations, the increase in communication rate with distance is of little significance in that the design must be based on the worst operating conditions, i.e., the remote unit located 150 to 300 miles from the base station. Thus, if a given design will satisfy the nearby remote units, it may be possible that a decrease in antenna efficiency for a more remote unit would be offset by an increased efficiency resulting from the increased distance. The last point which needs consideration at this time involves the polling sequence of the base station. During the entire Summer Test and for most of the Winter Test a polling sequence was used whereby a common call was trans- mitted by the base station. With this type of operation the base station sends out a signal which can basically be interpreted as "anyone who hears me, respond." For a small remote network such a polling sequence should cause no problem. However, when the number of remote units is increased to several hundred, this type of polling can potentially cause serious diffi- culty. Under such conditions it is possible that several remote units will receive a message to report at the same instant resulting in a number of communications being received at the base station at the same time. The net result is a multipath communication which the base station interprets as a scrambled message. Unfortunately, no one has yet tested a system consisting of several hundred remote units and thus the seriousness of such potential problems is unknown. However, some insight into the magnitude of the problem can be obtained from the results of the Winter Test. If the waiting time distribution function as presented in Figure 9 for the hour with the minimum mean waiting time is extrapolated to 100 milliseconds (0.1 second) it can be seen that approximately 2% of the messages would be received in that waiting time or less. If there are two remote units which are statistically inde- pendent yet which have the same distribution of waiting times (the case shown in the Winter Test for all stations except the Environmental Station) then the probability of having both units respond within the same time window is approximately 0.04%. When the system is expanded to three remote units and the figures cited above are used, it can be seen that the probability of units A & B responding at the same time is 0.04%, the probability of units A & C responding at the same time is 0.04%, and the probability of units B & C responding at the same time is 0.04%. Thus the probability of having a multipath communication is increased to approximately 0.12%. As the number of stations increases, the probability of having a multipath response from more than one station also increases (approximately 4% for 100 remote units). Thus the polling sequence 48 used in this testing program, i.e., "anyone who hears me, respond" cannot be effectively used in an operational system involving more than just a very few remote units. As an alternative, a sequential polling by individual remote unit address or possibly a sequential polling by small groups of remote units, will have to be used on systems containing a large number of remote units. By so doing, the statistics associated with the operation of such a system may be different from those presented here. However, since a large system has never been tested, the applicability of the results present- ed here is unknown. It is realized that the extrapolation of the data as presented above is a questionable operation and further, that the conclusions drawn from such an extrapolation often are misleading and incorrect. However, with the amount of information currently available on meteor burst systems it is felt that such an analysis does give further insight into what might be expected as well as highlighting another item which must be considered in the design and operation of a large system. 49 SUMMARY AND CONCLUSIONS The Winter Test Program of the Boeing Meteor Burst Communication System consisted of a thirty day period during which a base station located in Anchorage was operated on a continuous basis. For the full test period a remote unit located at the SCS offices in Fairbanks was also in operation. A second remote unit located at the Caribou-Poker Creeks Research Watershed located approximately 35 miles north of Fairbanks was operated for the first three weeks of the test program. In addition to these "permanent" stations, a remote unit was tested for short periods of time from Dietrich Camp, Prudhoe Bay, McGrath, Bethel, Kotzebue, and Big Delta. A total of approx- imately 40,000 messages were received by the base station and form the basis for the analyses presented in this report. The data from the Fairbanks Station indicated that a mean waiting time be- tween consecutive good messages varied from a low of approximately 2.0 minutes for the period between 10-11 a.m. to a maximum of approximately 7.0 minutes for the period 5-6 p.m. The anticipated sinusoidal diurnal variation in waiting times was not encountered. Rather, a diurnal relationship was defined whereby a gradual decay function existed for the period of approx- imately 10 p.m. to 11 a.m. at which time a gradual ascent with a maximum at approximately 6 p.m. occurred. The ascent was followed by a rapid decline until approximately 10 p.m. at which time the function would repeat itself. An analysis of variance was performed on the data and indicated that for the thirty day test period there was no apparent seasonal variation (95% confi- dence level) even though such an effect was anticipated from previous studies. The data concerning the effect of distance on the communication rate was inconclusive but it appears that there was no distance effect as had been suggested by previous studies. The data from the Environmental Station indicated that the system operated at temperatures of at least -28°F with no apparent malfunction. The minimum specified operating temperature as supplied by the manufacturer is -22°F. The Environmental Station was equipped with a counter for recording the number of times the remote transceiver was switched to the transmit mode. From these data it was determined that for non-aurora periods, it took an average of 2.2 attempts to get one good message through to the base station. When aurora periods were included in these computations, it took an average of nearly 9 attempts to get one good message through to the base station. Thus, it can be seen that for operations in the high latitudes, the frequent occurrence of the aurora can have a pronounced effect on the operation of the system. The effects of the aurora are highly complex and can both enhance and hinder the operation of the system. Since the aurora can result in a serious power drain on remote units powered by a self-contained power source, and further, since the aurora can affect one station while another station a short dis- tance away is unaffected, it is imperative that some protective device be 50 provided for the remote units associated with an operational system. Since the aurora effects are "remote station selective" the protection device should be with the remote unit and not with the base station. In this way, the maximum use of the system can be made in that remote units unaffected by the aurora disturbance can still be operated either normally or at an en- hanced level while the adversely affected stations are temporarily shut down, On several occasions during the test program, the Boeing Hydro-Met Unit did not provide a proper update-message to the meteor burst system. However, in an equal number of cases (4) the NWS DARDC also failed to provide a proper up-date message. Except for minor difficulties with the research equipment as discussed periodically throughout the text, no significant malfunctions of the equipment were encountered. A test was performed using an inverted "V" antenna with the remote unit in place of the standard dipole. It was found that at the 95% confidence level there was no apparent difference in the communication rate regardless of the antenna used. The results of the Winter Test indicate that a meteor burst system properly designed and operated could meet many of the envisioned telemetry needs for the acquisition of environmental data in Alaska. ■it U.S. GOVERNMENT PRINTING OFFICE: 1976- 210-801/367 51 (Continued from inside front cover) NWS HYDRO 15 Time Distribution of Precipitation in 4- to 10-Day Storms — Arkansas-Canadian River Ba- sins. Ralph H. Frederick, June 1973. (COM-73-11169) NWS HYDRO 16 A Dynamic Model of Stage-Discharge Relations Affected by Changing Discharge. D. L. Fread, December 1973. (COH-74-10818) NWS HYDRO 17 National Weather Service River Forecast System — Snow Accumulation and Ablation Model. Eric A. Anderson, November 1973. (COM-74-10728) NWS HYDRO 18 Numerical Properties of Implicit Four-Point Finite Difference Equations of Unsteady Flow. D. L. Fread, March 1974. NWS HYDRO 19 Storm Tide Frequency Analysis for the Coast of Georgia. Francis P. Ho, September 1974. (COM-74-11746/AS) NWS HYDRO 20 Storm Tide Frequency for the Gulf Coast of Florida From Cape San Bias to St. Petersburg Beach. Francis P. Ho and Robert J. Tracey, April 1975. (COM-75-10901/AS) NWS HYDRO 21 Storm Tide Frequency Analysis for the Coast of North Carolina, South of Cape Lookout. Francis P. Ho and Robert J. Tracey, May 1975. (COM-75-11000/AS) NWS HYDRO 22 Annotated Bibliography of NOAA Publications of Hydrometeoro logical Interest. John F. Miller, May 1975. NWS HYDRO 23 Storm Tide Frequency Analysis for the Coast of Puerto Rico. Francis P. Ho, "ay 1975. (COM- 11 001 /AS) NWS HYDRO 24 The Flood of April 1974 in Southern Mississippi and Southeastern Louisiana. Edwin H. Chin, August 1975. NWS HYDRO 25 The Use of a Multizone Hydrologic Model With Distributed Rainfall and Distributed Par- ameters in the National Weather Service River Forecast System. David J. Morris, August 1975. NWS HYDRO 26 Moisture Source for Three Extreme Local Rainfalls in the Southern Intermountain Region. E. Marshall Hansen, October 1975. NWS HYDRO 27 Storm Tide Frequency Analysis for the Coast of North Carolina, North of Cape Lookout. Francis P. Ho and Robert J. Tracey. November 1975. NWS HYDRO 28 Flood Damage Reduction Potential of River Forecast Services in the Connecticut River Basin. Harold J. Day and Kwang K. Lee, February 1976. NWS HYDRO 29 Water Available for Runoff for 4- to 15-Days Duration in the Snake River Basin in Idaho. Ralph H. Frederick and Robert J. Tracey, June 1976. 'iiiir ADD007Som 3 qt NOAA--S/T 76-2105