Answered By: Priscilla Coulter Last Updated: Jul 29, 2022 Views: 252758
Essentially, peer review is an academic term for quality control. Each article published in a peer-reviewed journal was closely examined by a panel of reviewers who are experts on the article's topic (that is, the author’s professional peers…hence the term peer review). The reviewers assess the author’s proper use of research methods, the significance of the paper’s contribution to the existing literature, and check on the authors’ works on the topic in any discussions or mentions in citations. Papers published in these journals are expert-approved…and the most authoritative sources of information for college-level research papers. Show Articles from popular publications, on the other hand (like magazines, newspapers, or many sites on the Internet), are published with minimal editing (for spelling and grammar, perhaps; but, typically not for factual accuracy or intellectual integrity). While interesting to read, these articles aren’t sufficient to support research at an academic level. But, with so many articles out there, how do you know which are peer-reviewed?
See also: What does "scholarly" mean? Is it the same as "peer-reviewed?" credible
Need a visual? Watch this quick video from the North Carolina State University Libraries: Media
Related Topics
There are many demands on a researcher’s time today and so it is a legitimate question to ask why some of that precious time should be spent reviewing someone else’s work. How does being a reviewer help you in your career? Here are some top ways that you can benefit. Keep up with the latest thinking: As a reviewer you get an early view of the exciting new research being done in your field. Not only that, peer review gives you a role in helping to evaluate and improve this new work. Improve your own writing: Carefully reviewing articles written by other researchers can give you an insight into how you can make your own better. Unlike when you are reading articles as part of your research, the process of reviewing encourages you to think critically about what makes an article good (or not so good). This could be related to writing style, presentation, or the clarity of explanations. Boost your career: While a lot of reviewing is anonymous, there are schemes to recognize the important contribution of reviewers. You can also include your reviewing work on your resumé. Your work as a reviewer will be of interest to appointment or promotion committees who are looking for evidence of service to the profession. Become part of a journal’s community: Many journals act as the center of a network of researchers who are in conversation about key themes and developments in the field. Becoming a reviewer is a great way to get involved with that group. This can give you the opportunity to build new connections for future collaborations. Being a regular reviewer may also be the first step to becoming a member of the journal’s editorial board. Your research community needs youOf course, being a reviewer is not just about the benefits it can bring you. The Taylor & Francis peer review survey found that these are the top 3 reasons why researchers choose to review: Playing a part as a member of the academic community: Peer review is the bedrock of academic publishing. The work of reviewers is essential in helping every piece of research to become as good as it can be. By being a reviewer, you will play a vital part in advancing the research area that you care about. Reciprocating the benefit: Researchers regularly talk about the benefits to their own work from being reviewed by others. Gratitude to the reviewers who have improved your work is a great motivation to make one’s own contribution of service to the community. Enjoying being able to help improve papers: Reviewing is often anonymous, with only the editor knowing the important contribution you’ve made. However, many reviewers attest that it is work that makes them feel good, knowing that they have been able to support a fellow researcher. Page 2
Chief Editor ROGER A. PIELKE Colorado State University
William M. FRANK
Editors ROBERT MADDOX ERL/NOAA, Boulder, CO
KEVIN E. TRENBERTH NCAR, Boulder, CO
Technical Editors Copy Editors Editorial Secretary RICK LYONS OLIVE DENNISTON ANN C. GAYNOR Mary L. MCMAHON HAROLD A. NAGEL LISA A. HIGGINS LESLIE BURGESS Associate Editors LLOYD J. SHAPIRO NCAR, Boulder, CO Simpson Weather Assoc., Inc. University of Washington AOML/NOAA, Miami, FL KENNETH C. CRAWFORD Charlottesville, VA RICHARD D. ROSEN JOANNE SIMPSON NOAA/NWS Forecast Office ROLAND A. MADDEN Atmospheric and EnvironmentalGoddard Space Flight Center Oklahoma City, OK NCAR, Boulder, CO Research, Inc., Cambridge, MA Greenbelt, MD ROGER DALEY PAUL W. MIELKE, JR. GRAEME L. STEPHENS THOMAS SCHLATTER NCAR, Boulder, CO Colorado State University Colorado State University NOAA/ERL, Boulder, CO THOMAS T. WARNER NOAA/ERL, Boulder, CO NOAA/PMEL, Seattle, WA RODERICK A. SCOFIELD Pennsylvania State University NOAA/NESDIS, Washington, DC EUGENE M. RASMUSSON EDWARD J. ZIPSER Climate Analysis Center NCAR, Boulder, CO Washington, DC The MONTHLY WEATHER REVIEW, a medium for the publication of research on meteorological topics, is published monthly by the American Meteorological Society. The editorial policy is to encourage papers in which the emphasis of the research is related to weather observations, analysis and forecasting. Papers in more specific applied areas (e.g., environmental health, air pollution meteorology, and agricultural and forest meteorology) in which applications of meteorology are critical to an interdisciplinary approach to environmental problems, should be directed to the JOURNAL OF CLIMATE AND APPLIED METEOROLOGY. Basic atmospheric research is published in the JOURNAL OF THE ATMOSPHERIC SCIENCES. A detailed discussion of the topics of research appropriate for publication within the AMS journals is presented in the August 1975 issue of the BULLETIN OF THE AMERICAN METEOROLOGICAL SOCIETY (also the July 1975 issue of MONTHLY WEATHER REVIEW) and the 1983 AMS Authors' Guide. INFORMATION FOR CONTRIBUTORS Manuscripts for the MONTHLY WEATHER REVIEW and all cor- Law (Title 17 USC, as revised by P.L. 94-553) in order for AMS to respondence should be sent to Dr. Roger A. Pielke, Chief Editor, % have valid rights to continue its wide dissemination of research results Mrs. Ann C. Gaynor, Editorial Assistant, Box 500, Afton, Virginia and other scientific information. Editorial action on a manuscript 22920. Authors may be members or nonmembers of the American that is not accompanied by the completed copyright transfer form, Meteorological Society and may be of any nationality, but only signed by all authors, will be delayed until the form is received. manuscripts in the English language can be accepted. When a manuscript is submitted the author must inform the Chief Editor if it has Three departments (Articles, Notes and Correspondence, and Picbeen published previously in any language, and he must similarly ture(s) of the Month) and one regular feature (Seasonal Climate inform the Chief Editor promptly if he submits it for publication Summary and/or Annual Summary) are maintained in the MONTHLY elsewhere before its disposition by the MONTHLY WEATHER REVIEW. WEATHER REVIEW. Correspondence is intended for informal com ment and discussion of published papers and will be published at Each manuscript should be accompanied by a statement trans the discretion of the Chief Editor. A Note may be a comment on ferring copyright from the authors (or other holder of the copyright) the application of theory, a brief description of a short development to the American Meteorological Society. The appropriate form for project, or a report on significant preliminary results of a longer the transfer of the copyright to AMS is printed near the end of the investigation. January issue of AMS journals; copies of the form are also available from the Chief Editor's office or the Society. The written, signed Authors will receive copies of page proof for correction. Authors' transfer of the copyright is required under the new U.S. Copyright institutions are requested to pay a publication charge, which, if hon Continued on Cover 3 © Copyright 1985, American Meteorological Society (AMS). Permission to use figures, tables and brief excerpts from this journal in scientific and educational works is hereby granted provided that the source is acknowledged. Any use of material in this journal that is determined to be "fair use" under Section 107 or that satisfies the conditions specified in Section 108 of the U.S. Copyright Law (17 USC, as revised by P.L. 94-553) does not require the Society's permission. Items that do not bear their own separate copyright notices either are in the public domain or are U.S. Government works for which copyright protection is not available. Authorization to photocopy items bearing individual AMS copyright notices at the bottom of their first page is granted by AMS for libraries and other users registered with the Copyright Clearance Center (CCC) Transactional Reporting Service, provided that the copies are for internal or personal use, or for the internal or personal use of specific clients, and further provided that the base fee of $4.25 per copy plus $0.25 per page is paid directly to CCC, 21 Congress St., Salem, MA 01970, and that the following fee code for this journal is reported with the payment to CCC: 00270644/85 $4.25 + 0.25. Republication, systematic reproduction, and other uses of any material in this journal, unless exempted by the above statements, require written permission or license from the AMS. The MONTHLY WEATHER REVIEW (ISSN 0027-0644) is published monthly by the American Meteorological Society at 45 Beacon Street, Boston, MA 02108. Second class postage paid at Boston, MA, and additional mailing offices. The subscription price is $30.00 per year to members of the Society. The price of single issues is $7.50 for members. The subscription price to nonmembers is available on request; single issues are $15.00 each. All subscriptions outside the United States carry an additional charge of $15.00 per year for postage and handling. Subscription orders, claims for missing numbers, changes of address, and other business correspondence should be addressed to the American Meteorological Society, 45 Beacon Street, Boston, MA 02108 (617-227-2425). Page 3
Table 1. Years of Warm and Cold Events. a: immediately was higher in Yr-, than in Yro in both MJJ and ASO, follows a Yearo of a WE; b: immediately precedes a Yearo of a WE. whereas in 14 of the Warm Events in MJJ and 16 in ASO it was more than one 3-month standard deviation Warm Events Cold Events higher in Yro. This should be viewed against the fact 1864 that we are dealing with a spot value and that the SO 1868 anomalies do not necessarily appear in exactly the same 1877 position in each event. 1880 The differences between two consecutive years in 1884 1886 1888 1889a MJJ for the 97 years which were not WEs are shown 1891 1892a in Fig. 3c. Of the 97, the pressure in the first year was 1896 higher than that in the second in 45 instances and lower 1899 b1898 in 49 instances, and in three the pressure was the same 1902 1904 61903 1911 1906 The signal in Fig. 1a thus seems reasonably well es1913 1908 tablished over southern Australia, although one should 1918 1916 not expect it to appear in the same place in all WEs. 1923 1920 In the latest event of 1982 the signal was very clear as 1925 61924 the SLP at Adelaide in JJA of 1981 was almost 6 mb 1930 61931a 1932 61938 below normal, and in the following year 5 mb above 1939 1942 normal. 1951 1949 The longest series of observations in the center of 1953 1954a the area in Fig. la where the SLP is lower in Yro than 1957 1963 61964a in Yr-, is that of Rapa Island. The value at the island 1965 1966a exceeds the 99% confidence level (Fig. 1b). Figure 4a 1969 1970a shows time series of the MJJ pressure at Rapa and 1972 1973a Adelaide, starting in the year when observations began 1976 1978 at Rapa. All the eight Yearso at Rapa lie in a valley of 1982 the curve, from 2 to 4 mb lower than the previous year, and thus corroborate the results of the deficient grid point values in Fig. 1a. as 3-month running means. This difference is appre- The pressure in seven of the eight WEs was higher ciably larger than that between Year, and the long term than in the previous year at Adelaide (Fig. 4a), but mean, which also holds true for pressure and temper- there are peaks-notably 1959 and 1979—which are ature over other areas of Australasia and the South not associated with WEs, just as there are valleys in Pacific Ocean. During the southern winter and spring Rapa's curve which are not in a WE. The SLP differ(Fig. 2) the pressure is on the average 2–2.5 mb higher in Year, than in Year-1. In May-June-July (MJJ), when the difference in Fig. 1a is about 5 mb at Adelaide, mb it is close to 2 mb for the 28-year means in Fig. 2. The ADELAIDE 2.4 latter value is equal to nearly one standard deviation 2.2 of the mean pressure in MJJ (crosses in Fig. 2), and is 2.0 several times larger than the standard deviation of 28 1.8 year averages (dots in Fig. 2; see vL for a definition of a 1.6 this quantity). The differences in SLP in Fig. 2 may 1.4 also be judged in the following way: For the total record 1.2 at Adelaide the average difference between consecutive 1.0 means of three months is zero with a standard deviation a YRO-MEANS 0.8 ranging from about 1 mb in summer to about 3 mb in winter. The differences between Yro and Yr-, in Fig. 0.4 2 are from one-half of this standard deviation in the 0.2 early part of the year to one standard deviation in O spring. Note also that the difference between Yro and -0.2 the long term mean in Fig. 2 is half of that between J F M A M J J A S O N D Yro and Yr-, which, of course, means that on the average Yr-, deviates as much from the mean as does FIG. 2. Upper curve: The three-month running mean difference in Yro, but with the opposite sign. Page 4
that in Yro over the area north of 45°S in the Pacific The difference in sea surface temperature, Yro minus Fig. 7a. Objective analysis of the mean difference (°C) in sea surface Yr-1, is given in Fig. 7 for various 3-month periods: temperature, Yr, minus Yr... There were always insufficient data within the triangle. December-January-February. DJF, MJJ; ASO, called January, June, and September in the following. In January (Fig. 7a) the pattern of the SST differences is unsystematic and the differences are comparatively small. By June, however, the pattern is The Warm Event of 1982 (not included in Fig. 7) conformed to this pattern: As demonstrated in Fig. 6 well organized and in our area of interest the SST is higher in Yr-1 (Fig. 7b). This higher water temperature of vL, the trough had not appeared in the 3-monthly in Yr-, continues into the southern spring (Fig. 7c). It mean as late as July of 1981 and the meridional anomshould be noted that a test of the statistical significance June-July of 1981 (Fig. Se). As a result the SST in this aly wind in the area was therefore northerly in Mayof Figs. 7b, c gave very small areas where the confidence level reached 95%, possibly owing to the noisy nature area (Fig. 8) was as much as 2°C higher than in 1982 (the SST data in this figure were kindly supplied by R. of the data. It is more important, however, that the Reynolds of the Climate Analysis Center/NOAA). changes in wind (pressure) and SST are physically con What is said here about the SST applies equally well sistent. to the air temperature: Rarotonga (21°S, 160°W) lies in the area of largest SST difference between Yro and Yr-1, and as its record begins in 1907 it is possible to EQ sample 17 Warm Events by means of its air tempera ture. The lower curve in Fig. 9 shows that the mean 20°S temperature in winter and spring is higher in Yr-, than in Yro, and that this difference begins in fall at the time 1891 the trough fails to amplify in Y1-1. The higher air tem40° S perature in Yr-1 is not due to a few large values as one can see in Fig. 10: In this time series of the temperature EQ difference between consecutive years in July-August September positive values denote that the temperature 20° S was higher in the second year, and conversely for the negative values. It is clear that the temperature was 1925 higher in Yr-, than in Yro in 16 of the 17 Warm Events. 40° s As with the SLP, we note that there are many in160° E 180° 160° W 140° 120° 100° 80° stances when the temperature was higher in the first year (Fig. 10) which were not associated with a Warm EQ Event. Three of the twenty such instances were con tinuations of Warm Events, such as 1958. If we dis20° S regard those three, there are as many instances when the temperature was higher in the first year which did 1951 not become Warm Events as there are instances which 40°S became Warm Events. This indicates that even if we know what this stage of the development of a Warm EQ Page 5
location during the initial 12-month period, and an It is generally desirable to compare the accuracy of additional 208 experimental forecasts for each location the forecasts of interest-for example, the SUB forewere formulated during the five-month period in 1982. casts—with the accuracy of forecasts produced by some The objective QPFs, as well as the subjective and ob- "standard of reference." This comparison is usually jective PoP forecasts, have also been obtained for this accomplished by computing a skill score (or index of same set of occasions. Observations of precipitation relative accuracy) defined in terms of the basic measure amounts in the corresponding 12-hour periods at these of accuracy. The primary skill scores used here are inlocations have been used to verify these two types of dices of relative accuracy in which the standard of refprobability forecasts. erence consists solely of climatological probabilities of the respective events. Such climatological probabilities 3. Methods of evaluation can be based either on the long-term historical relative frequencies of the events or on the sample relative frePrecipitation occurrence is a dichotomous variable, quencies of these events during the experimental peand the Brier score (Brier, 1950) is a suitable measure riod. Skill scores are computed here for both “definiof the accuracy of probability forecasts for such a vari tions" of climatology. In this regard, if we denote skill able. Precipitation amount, on the other hand, is an scores based on long-term and sample climatological ordinal variable (i.e., its values are defined on an ordinal probabilities by SSC and SSsc, respectively, then scale), and a measure of the accuracy of probability forecasts for such a variable should take its ordinal na SSC 1 – (RPS/RPSC), (4) ture into account. Evaluation measures that possess and SSsc = 1 - (RPS/RPSSC), – (5) ranked probability score (RPS) (Epstein, 1969; Murphy, 1971) is a measure of accuracy that is sensitive to where RPS is the average accuracy of (for example) distance (see Staël von Holstein, 1970). Thus, the RPS the SUB forecasts and RPSC (RPSSC) is the average is the basic evaluation measure used in this paper. accuracy of the long-term (sample) climatological For a sample of K probability forecasts for a variable probabilities. The values of SSc and SSsc range from consisting of N = 6 MECE events, the average RPS minus infinity to one, with the latter being the best possible score (RPS = 0). Positive (negative) values of can be expressed as follows: SSC and SSsc indicate that the accuracy of the SUB forecasts is better (worse) than the accuracy of the reRPS = (1/K) E (Rnk – Dnk)?, 1E (1) spective climatological probabilities." As noted in Section 1, we are also interested in comin which paring the SUB and OBJ forecasts. It is convenient to make this comparison by computing a skill score in (2) which the MOS system plays the role of the standard SSOBJ, then (3) SSOR = 1 – (RPS/RPSOR), (6) where RPSOBJ is the average accuracy of the objective where I'mk denotes the forecast probability of the mth forecasts produced by the MOS system. The range of event (or, in this case, the mth category of precipitation SSOR in (6) is the same as the range of SSc and SSC amount) on the kth occasion (Tmk > 0, Em imk = 1) in (4) and (5), respectively, and SSos is positive (neg ) and dmk = 1 if the mth event occurs on the kth occasion ative) when the accuracy of the SUB forecasts is greater and dmk = 0 otherwise (m = 1, ...,6; k = 1, ...,K). (less) than the accuracy of the OBJ forecasts. Since Rnk and Dink represent components of cumulative In addition to accuracy and skill, other attributes of forecasts and observations, respectively, RPS can be probabilistic forecasts are of interest. Two such attriviewed as the mean square error of such forecasts (see butes are reliability and resolution. Reliability refers to Murphy, 1971). The RPS also is a strictly proper scor- the degree of correspondence between forecast probing rule (Murphy, 1969). Finally, the values of RPS, abilities and observed relative frequencies for specific as defined in (1), range from zero (best possible score) (sets of) probability values, whereas resolution relates to five (worst possible score). to the forecaster's (or forecast system's) ability to identify subsamples of forecasts for which the observed rel
Let Rink denote the probability that the observed precipitation on the kth occasion exceeds the amounts associated with the nth category. Then, Rink = 1 – Rnk (n = 1, ...,6), and it can be seen that RPS in (1) is a measure of the accuracy of the forecasts in their original form.
* The relative merits of using sample climatological probabilities or long-term historical climatological probabilities as a standard of reference have been discussed in some detail by Murphy (1973). Page 6
ative frequencies differ from the overall observed rel- ities and the corresponding cumulative observed relative frequencies for the sample of forecasts as a whole. ative frequencies, where the weights are simply the Thus, in order to study the reliability and resolution subsample sizes (e.g., see Murphy and Daan, 1984). of a sample of forecasts, it is necessary to divide the REL equals zero (i.e., the reliability is perfect) if and sample into subsamples each of which is associated only if the respective relative frequencies and probawith a distinct set of forecast probabilities. Moreover, bilities are equal for all subsamples. The values of REL two approaches are possible-a two-event (N = 2) ap- range from zero to five in a six-event situation. proach in which (in effect) the probabilities associated The third term on the right-hand side of (7) is a with each event are treated separately and a multiple- measure of the resolution of the cumulative forecasts. event (N > 2) approach in which the N probabilities Specifically, this term is the average weighted squared are treated simultaneously. Both of these approaches difference between the subsample cumulative observed are employed in this paper. relative frequencies and the overall cumulative ob We use two different "tools” to investigate the reli- served relative frequencies for the sample as a whole “ ability and resolution of the SUB and OBJ forecasts: (e.g., see Murphy and Daan, 1984). The term RES (a) an attributes diagram and (b) a partition of RPS. equals zero (i.e., the forecasts are completely unreAttributes diagrams (Hsu and Murphy, 1985) are em- solved) if and only if the subsample and sample relative , ployed here within the two-event approach, in which frequencies are equal for all subsamples. The values of the number of possible subsamples equals the number RES range from zero to RPSsc, with the latter repreof distinct probability values (e.g., 11 subsamples when senting the best possible score (note that this term is the permissible probability values are 0.0(0.1)1.0). In preceded by a minus sign). this approach, such a diagram consists of a plot of ob- The existence of the partition of RPS in (7) also served relative frequency against forecast probability provides an alternative—and quite useful-expression for specific probability values associated with a partic- for the skill score SSSc. By substituting for RPS from ular event. The empirical curve describing the rela- (7) into (5), the latter becomes tionship between these two quantities is then compared with lines (in the diagram) representing perfect reli SSsc = (RES – REL)/RPSSC. (10) ability, no resolution, no skill, etc. Thus, the attributes Thus, the skill of a sample of probability forecasts with diagram is an extension of the more familiar reliability respect to the sample climatological probabilities is diagram (e.g., see Murphy and Winkler, 1977; Murphy positive (negative) if the magnitude of the RES term and Daan, 1985). is greater (less) than the magnitude of the REL term. Under the assumption that the sample of forecasts 4. Some results can be divided into subsamples, each of which is associated with a distinct set of forecast probabilities, it This section has a threefold objective: 1) to present is possible to partition (or decompose) RPS in (1) into some results related to the quality of the SUB forecasts; three terms as follows: 2) to investigate the effects of feedback and experience on the quality of these forecasts; and 3) to compare the RPS = RPSsc + REL - RES (7) quality of the SUB and OBJ forecasts. To accomplish (e.g., see Murphy and Daan, 1985). In this paper the this objective, attention will be focused on specific segpartition of RPS in (7) is employed within a multiple- ments of the 17-month experimental period in the re a event approach (see Section 4). The first term on the spective subsections. Sample sizes are generally not sufright-hand side of (7) is simply the average accuracyficient to allow stratification of the results by time of of the cumulative sample climatological probabilities. day (i.e., day/night) or by forecaster. However, some It may be of interest here to note that results are stratified by location (i.e., NWS office) or by category of precipitation amount. A discussion of the results appears in Section 5. E D (1 – Dn), (8) a. Quality of the SUB forecasts in which In this subsection attention is focused on the quality Dn = (1/K) E Edmk (9) of the SUB forecasts over the 12-month period from February 1981 through January 1982. Thus, we are concerned here with the forecasters' performance prior (n = 1, ..., 6). Thus, RPSsc depends solely on the to their receipt of formal feedback at the end of the (cumulative) sample climatological probabilities; that first year of the experiment. The results of a comparison is, it is not a function of the forecast probabilities. of the forecasters' performance in the initial and final The second term on the rhs of (7) is a measure of five months of the 17-month experimental period will the reliability of the cumulative forecasts. Specifically, be reported in Section 4b. this term is the average weighted squared difference The overall reliability, or bias in-the-large, of the between the subsample cumulative forecast probabil- SUB (and OBJ) forecasts is indicated in Table 1. A Page 7
titative precipitation forecasts (QPFs) on an experi- traditionally expressed forecasts for most events in catmental basis. These QPFs (when considered in con- egorical terms, and categorical forecasting frequently ( junction with concurrent forecasts of the probability involves explicit or implicit guidelines that encourage of measurable precipitation) specify the likelihoods of overforecasting. Another possible explanation of the occurrence of six categories of precipitation amount in overforecasting is the so-called “value-induced” bias, a 12-hour period, and they were made twice a day for which arises when the impact of events on users of the four locations in Texas during a 17-month period in forecasts unduly influences the probabilities that fore1981-82. In the evaluation of the experimental fore- casters assign to the events. This factor presumably casts, attention was focused on three issues: 1) the basic plays a particularly important role in situations inquality of the subjective probabilistic QPFs; 2) the ef- volving forecasts of potentially dangerous or damaging fects of feedback and experience on the quality of these events (e.g., large amounts of precipitation). Of course, , , forecasts; and 3) the relative performance of subjective lack of experience in probability forecasting on the part (SUB) and objective (OBJ) probabilistic QPFs. In this of the forecasters could also contribute to such oversection we discuss the results presented in Section 4 forecasting, and the NWS forecasters involved in this with respect to these three issues—the discussion con- experiment had no experience in probabilistic quansists primarily of efforts to provide interpretations of titative precipitation forecasting prior to January 1981. and/or explanations for the results and to compare In February 1982 the forecasters were provided with these results with the results of other probability fore- feedback regarding their performance—and that of the casting studies (where appropriate). MOS system-during the first year of the experiment. Evaluation of the experimental probabilistic QPFs Thus, an important issue in this paper is whether or during the initial 12-month period (prior to the pro- not the feedback (or the experience gained by the forevision of any formal feedback) reveals that the forecasts casters during the initial 12-month period) led to subare quite reliable for categories involving smaller pre- sequent improvements in the forecasts. Comparison cipitation amounts and that overall they possess ap- of the SUB forecasts in the initial and final five-month preciable skill (see Fig. 1a-b and Table 2). This result periods reveals a modest improvement in reliability, a should not be surprising in view of the fact that the moderate deterioration in resolution, and a definite forecasters have had considerable experience in for- decrease in skill (see Tables 3-5 and Figs. 3 and 4). In mulating probability forecasts of measurable precipi- fact, this pattern of results holds for each of the four tation and that precipitation amounts equal to or locations as well as for all locations combined. Thus, greater than 0.635 cm occurred rather infrequently in the results of the experiment appear to differ markedly this period. On the other hand, considerable overfore- from the results of the Zierikzee experiment (Murphy casting is evident for probabilities associated with cat and Daan, 1984), in which substantial improvements , egories involving larger precipitation amounts, and the in the quality of subjective probability forecasts in the skill of these probability forecasts is clearly marginal second year were attributed to the provision of detailed (see Figs. 1c-f and 2). Moreover, the range of proba- feedback to the forecasters at the end of the first year. bility values used in the forecasts decreases markedly Several factors may have contributed to this relaas the category number (i.e., precipitation amount) in- tively negative result concerning the effects of feedback creases. These results are consistent with the results on the quality of the SUB forecasts. First and foremost, of other experiments in which forecasters have subjec- it is widely recognized that forecast accuracy is inversely tively assessed probabilities for weather events involv- related—and skill is directly related—to the relative ing a wide range of climatological probabilities (e.g., frequency of occurrence of the event in question. In Daan and Murphy, 1982; Murphy and Winkler, 1982). this regard, it should be noted that the sample relative Several possible explanations have been offered for frequencies of all five categories associated with meathe overforecasting which frequently occurs in such surable precipitation amounts decreased between the experiments. For example, it has been suggested that initial five months and the final five months (see Table this overforecasting may be a “carry-over" effect from 1). Thus, the decrease in sample relative frequencies ” categorical forecasting.” In this regard, forecasters have may have confounded, at least in part, the beneficial effects of the feedback. This argument is supported by the fact that substantial reductions in resolution and 6 It might be argued that this relationship results from the fact that skill between these two periods also occurred for the the OBJ forecasts are based on regression models for which the re OBJ forecasts (see Table 3). Moreover, the skill score duction in variance (i.e., performance) generally decreases as the climatological probability of the event decreases. Of course, this argu SSOBJ indicates that the quality of the SUB forecasts ment assumes that the SUB forecasts are strongly influenced by the OBJ forecasts. Page 8
vided to the forecasters. Obviously, however, substan- itation forecasting experiment. Corvallis, Oregon State Univertial improvements in quality are still possible, especially sity, Department of Atmospheric Sciences, M.S. thesis, 108 pp. -, and A. H. Murphy, 1985: Geometrical interpretation of some in the reliability of the probabilities assigned to cate attributes of probability forecasts: two-event situation. Int. J. gories involving larger precipitation amounts. In this Forecasting, 1, in press. regard, it would be desirable to design and conduct a Hughes, L. A., 1980: Probability forecasting-reasons, procedures, more extensive probabilistic quantitative precipitation problems. Silver Spring, MD, NOAA, National Weather Service, forecasting experiment. This experiment should be Tech. Memo. NWS FCST 24, 84 pp. (Available from National Weather Service, 8060 13th Street, Silver Spring, MD 20910.) conducted over a time period of sufficient duration to Krzysztofowicz, R., 1983: Why should a forecaster and a decision ensure adequate sample sizes, and it should include maker use Bayes' theorem. Water Resour. Res., 19, 327-336. the provision of feedback to forecasters at appropriate Lowry, D. A., and H. R. Glahn, 1976: An operational model for intervals. Such an experiment would also provide an forecasting probability of precipitation—PEATMOS POP. Mon. Wea. Rev., 104, 221-232. opportunity to make detailed comparisons of the rel Murphy, A. H., 1969: On the ranked probability score. J. Appl. Meative performance of forecasters and the output of nu teor., 8, 988-989. merical and/or numerical-statistical models. The ex- --, 1970: The ranked probability score and the probability score: periment, together with the analysis of its results, would a comparison. Mon. Wea. Rev., 98, 917-924. appear to be an essential component of any serious 1971: A note on the ranked probability score. J. Appl. Meteor., 10, 155-156. program designed to improve the quality of operational 1973: A sample skill score for probability forecasts. Mon. Wea. forecasts of quantitative precipitation events. Rev., 102, 48-55. 1977: The value of climatological, categorical and probabilistic Acknowledgments. We would like to express our forecasts in the cost-loss ratio situation. Mon. Wea. Rev., 105, appreciation to the forecasters at the WSFO in San 803-816. Antonio, Texas, for their cooperation and participation and H. Daan, 1984: Impacts of feedback and experience on the quality of subjective probability forecasts: comparison of in the probabilistic quantitative precipitation forecast results from the first and second years of the Zierikzee experiing experiment. The assistance of Messrs. Gifford F. ment. Mon. Wea. Rev., 112, 413-423. Ely, Jr. (Deputy MIC, San Antonio WSFO) and Daniel and 1985: Forecast evaluation. Probability, Statistics, L. Smith (Chief, Scientific Services Division, NWS and Decision Making in the Atmospheric Sciences, A. H. Murphy and R. W. Katz, Eds., Westview Press, 379-437. Southern Region Headquarters, Fort Worth, Texas) in -, W.-R. Hsu, and R. L. Winkler, 1982: Subjective probabilistic the design and conduct of the experiment is also grate quantitative precipitation forecasts: some experimental results. fully acknowledged. This work was supported in part Preprints, Ninth Conf. on Weather Forecasting and Analysis, by the National Science Foundation (Division of At- Seattle, Amer. Meteor. Soc., 94-100. mospheric Sciences) under grants ATM-8004680 and and R. L. Winkler, 1977: Reliability of subjective probability forecasts of precipitation and temperature. Appl. Statist., 26, ATM-8209713. 41-47. REFERENCES -, and —, 1982: Subjective probabilistic tornado forecasts: some experimental results. Mon. Wea. Rev., 110, 1288-1297. Bermowitz, R. J., and E. A. Zurndorfer, 1979: Automated guidance and — 1984: Probability forecasting in meteorology. J. for predicting quantitative precipitation. Mon. Wea. Rev., 107, Amer. Statist. Assoc., 79, 489-500. 122-128. National Weather Service, 1980a: The use of model output statistics Brier, G. W., 1950: Verification of forecasts expressed in terms of for predicting probability of precipitation (winter season). Silver probability. Mon. Wea. Rev., 78, 1-3. Spring, MD, National Weather Service, Tech. Proced. Bull. No. Charba, J. P., and W. H. Klein, 1980: Skill in precipitation forecasting 289, 13 pp. (Available from National Weather Service, 8060 in the National Weather Service. Bull. Amer. Meteor. Soc., 61, 13th Street, Silver Spring, MD 20910.) 1546-1555. 1980b: The use of model output statistics for predicting the Committee on Atmospheric Sciences, 1980: Atmospheric precipi- probability of precipitation amount and precipitation amount tation: prediction and research problems. National Academy Page 9
frontal air from Nowra at 1350 and 1500 EDST and in the prefrontal air from Sydney Airport at 1500 EDST are shown in Fig. 13. Also shown are the pre- and postfrontal soundings made by the aircraft. Over the land, the prefrontal air is very hot and dry due to its recent continental trajectory, and the lapse rate is superadiabatic to about 740 mb, or 2.69 km. As this air was advected over the cooler ocean surface, a strongly stable, but shallow, boundary layer developed. The postfrontal soundings show a well-mixed layer, due presumably to mechanical and convective turbulence. At Nowra, there is an appreciable superadiabatic lapse rate in the near surface layer, again evidence of strong instability as the residual heat in the previously hot land surface is released. Accordingly, the mixed layer over the sea is not as deep as at Nowra at 1500 EDST. Although the Nowra soundings indicate an increase in the depth of the postfrontal air, there is little change over the period between the two soundings in the thickness of the frontal inversion layer.
b. The burster of 1 December 1982
Analysis of this event was more difficult than the previous one. Added complications which illustrate the forecasting difficulties are mesolow development on wind shift lines over Victoria, multiple wind shift lines 0600 on the New South Wales coast, and mesohigh devel opment due to the outflow from thunderstorms. This FIG. 8. Isochrones of frontal position between 0600 EDST 25 No- burster had its origins in a trough that lay over Western vember 1982 and 0100 EDST 26 November 1982. Front locations Australia on 27 November 1982. The trough extended over the ocean, determined by the CSIRO aircraft, are shown with a circled cross and the times (EDST) of location are indicated. southeastward toward Spencer Gulf during 28 November and the following day frontogenesis occurred and a prefrontal windshift line developed. During 30 the front, with an initial lapse rate of 7.4°C/100 m November the system moved across western Victoria, (Fig. 12). The increasing stability over the hour prior several mesolows developed on the windshift line and at 1800 EDST there are some data suggesting the forto the front was probably caused by the decrease in mation of a second windshift line (see Fig. 14). This surface temperature which was associated with an increase in middle level cloud. There is little difference complex system moved eastward ahead of a Southern in the temperature changes at Nowra, Sydney Airport Ocean cold front which has passed over the Great Ausand Richmond (Table 2), but the change becomes tralian Bight during the day. By 0300 EDST 1 Decemsmaller as the front moves north of Sydney and at Coffs er 1982 the prefrontal troughs extended from the northwest of New South Wales to the eastern Victorian Harbour the temperature actually rose following the change. This last result is not paradoxical and was coast (Fig. 15). found in some instances by Clarke (1961) who attrib Despite multiple changes evident on the analyses uted it to mechanical stirring of the surface inversion. only one discontinuity was indicated by the Green Cape Garrett et al. (1985) also discuss this phenomenon. anemograph record, and this occurred just before 0900 The strongest wind gust recorded by a permanent EDST with the passage of the surviving windshift line, anemograph at standard exposure was 24.7 m s'at Sydney Airport. Over the six years of the investigation TABLE 1. Temperature difference AT, measured over a distance described in Section 2, seven of the seventeen strong of about a degree of latitude across the front, and temperature of the burster events registered gusts of at least 24.7 ms-'at immediate postfrontal air Tc, on 26 November 1982. Sydney Airport, and the highest gust was 27.3 m s Time AT T. (EDST) (°C) (°C) coast were southerly, further inland they were southeasterly and, on parts of the tablelands had backed fur 0900 26-27 ther to be from the eastnortheast. 1200 14 22 The data from the radiosondes released in the post 1500 22 21 Page 10
previous event, as were the temperature contrasts between pre- and postfrontal air masses (Table 3). The temperature at Gore Hill fell 11°C within an hour of the arrival of the burster. The lapse rate decreased for the three hours up to ten minutes prior to the burster then increased, becoming unstable soon after the passage of the front. Unlike the event on 25 November the increasing stability was not associated with a decrease in surface temperature, but may have been a result of prefrontal subsidence. The low level wind field near Nowra was determined for this event by tracking 30 gm helium filled pilot balloons (pibals) with two theodolites. The theodolites were separated by an east-west baseline distance of about 1.4 km and synchronized readings (azimuth and 17 elevation) from each theodolite were made every thirty TIME (EDST) seconds, corresponding with an incremental balloon FIG. 12. Variations in 10 m level temperature (°C) and lapse rate height rise in still air of about 65 m. In clear conditions, (°C/100 m) between the 10 and 110 m levels at Gore Hill, Sydney, balloons were tracked for up to twenty minutes to an between 1400 and 2000 EDST on 25 November 1982. altitude in excess of 2 km. A time-height cross section of the relative streamlines (x, z) = constant in a plane which by this time was developing frontal character- normal to the surface front is shown in Fig. 18. These . istics. are determined by integrating the formula Interpretation of the anemograph data from the New at South Wales Coast is difficult. Even though only one u(x, z) – C, (1) discontinuity occurred at Green Cape, two were de дz tected at Merimbula, Bermagui, Narooma and Kioloa, with respect to z, u(x, z) being the horizontal wind ( while three were evident at Moruya Heads. The first component in the x-direction, normal to the surface discontinuity at Moruya Heads involved a change in front, with x and u positive in the direction towards wind direction but the second and third were mainly the cold air, and c is the disturbance propagation speed speed changes. Synoptic data indicate that there was a in the x direction, equal to -13.2 m -. In Fig. 18a, lower tropospheric temperature inversion over the is calculated from integration of Eq. (1) assuming South Coast during the morning. The discontinuities that u is measured at a fixed x, equivalent to ignoring . developed as the front moved into this inversion layer. the sloping trajectory of balloons relative to the moving Figure 16 shows the locations of the initial discontinuity disturbance. To take the sloping balloon trajectory into (full lines) at hourly intervals. Also shown (dashed line) account requires that the ascent rate of balloons relative is the position at 0900 EDST of the windshift line which to still air is known as described by Smith and Morton caused the one discontinuity at Green Cape. Discon- (1984; see appendix). The streamlines computed by tinuities at 1000 and 1100 EDST are new develop- integrating Eq. (1) along the sloping balloon trajectory ments. CSIRO aircraft observations show that they x = xblz) are shown in Fig. 18b for the region surwere less than 450 m deep at 1136 EDST. The discon- rounding the frontal nose where errors in the simpler tinuity located by the aircraft at 1136 EDST may have been the developing front which passed Green Cape just before 0900 EDST. TABLE 2. Temperature changes (°C) in the postfrontal air on 25 The complexity of the synoptic situation is shown November 1982 and the maximum postfrontal wind gust (m s-') by the MSL pressure analysis at 1500 EDST 1 Decem- within one hour of the front. ber (Fig. 17). Thunderstorm activity over northeastern New South Wales and southeastern Queensland had Change in temperature (°C) within the given produced a mesohigh/mesolow coupled with an 11 mb period (min) pressure difference between the centers. As in the pre Front immediately following vious event, movement across the tablelands of New Maximum arrival the front South Wales was slow; however, this front penetrated gust time Location (m s-') (EDST) 10 30 60 further to the west, crossing most of the tablelands area in the south of the state by 0300 EDST 2 December. Nowra 18.5 1342 -18.0 -20.0 -21.2 The 500 mb analysis at 1500 EDST 1 December Sydney Airport 24.7 1550 -18.0 -19.0 -20.0 showed a trough lying from the southeastern to the Richmond 22.7 1722 -17.0 -18.7 - 20.0 Williamtown 21.1 northwestern corners of New South Wales and at 300 1823 -15.0 -16.3 -16.7 Coffs Harbour 15.4 0110* +0.2 +0.4 +1.2 mb a jet stream was between Victoria and Tasmania. Generally gust strengths were less than those of the * The front reached Coffs Harbour on 26 November 1982. Page 11
Christie, D. R., and K. J. Muirhead, 1983: Solitary waves: A low level wind shear hazard to aviation. Int. J. of Aviation Safety, 1, The Bureau of Meteorology, Melbourne, who orga 169-190. (Available from Capstan Press, Queen St. Chambers, nized the supply of Woelfle anemometers and provided Queen St., Exeter, Devon, UK.) hourly GMS data; the New South Wales Regional Of- Clarke, R. H., 1961: Mesostructure of dry cold fronts over featureless fice of the Bureau of Meteorology who sited and re terrain. J. Meteor., 18, 715-735. moved the anemometers, and to Malcolm Down and Coulman, C. E., J. R. Colquhoun, R. K. Smith and K. McInnes, 1985: Orographically forced cold fronts—mean structure and Malcolm Sullivan who assisted with diagram prepa motion. Bound. Layer Meteor., 32, 57-83. ration. The Australian Oil Refining Pty. Ltd., Snowy Colquhoun, J. R., 1981: The origin, evolution and structure of some Mountains Hydro Electric Authority, the University southerly bursters. Tech. Rep. 40. Australian Bureau of Meteof Wollongong, Lucas Heights Research Laboratories orology, P.O. Box 1289K, Melbourne 3001, Australia. Doswell, C. A. III, 1982: The operational meteorology of convective and East Sale Meteorological Office for anemometer weather. Volume 1: Operational Mesoanalysis. NOAA Tech. data. The ACT Regional Office, the Bureau of Mete- Mem. NSW NSSFC-5 (available from NTIS, Springfield, Virorology for data from Woodlawn Mines, the Conser- ginia, 22161). vation and Agriculture Department, and the Australian Gauntlett, D. J., L. M. Leslie and L. W. Logan, 1984: Numerical experiments in mesoscale prediction over southeast Australia. National University and to people who assisted by al Mon. Wea. Rev., 112, 1170-1182. lowing anemometers to be sited on their land. Garrett, J. R., W. L. Physick, R. K. Smith and A. J. Troup, 1985: Discussions with Mr. K. Marriott were helpful. The Australian summertime cool change. II: mesoscale aspects. Daphne Herzog typed a draft of the manuscript. MFM Mon. Wea. Rev., 113, 202-223. model data were kindly provided by Dr. L. M. Leslie. Gentilli, J., 1969: Some regional aspects of southerly buster phenom ena. Weather, 24, 173-180. Data collected by the CSIRO aircraft were processed Gill, A. E., 1977: Coastally trapped waves in the atmosphere. Quart. by kind permission of the Anglo-Australian Observa- J. Roy. Meteor. Soc., 103, 431-440. tory, Epping NSW. The first two authors thank the Hunt, H. A., 1894: An essay on southerly busters. J. Proc. Roy. Soc. Director, Bureau of Meteorology for permission to NSW, 28, 138-184. (Reprinted in R. Abercrumbie, Ed., 1896: publish this paper. Three essays on Australian weather. White Publishing.) Hutchings, J. W., 1944: Orographical disturbances of the pressure field over New Zealand. N.Z. M.O. Ser. A, No. 7,4 pp. (Available REFERENCES from N.Z. Meteorological Service, P.O. Box 722, Wellington, N.Z.) Baines, P. G., 1980: The dynamics of the southerly buster. Aust. Lilly, D. K., 1981: Doppler radar observations of upslope snowstorms. Meteor. Mag., 28, 175-200. Preprints, 20th Conf. on Radar Meteor., Amer. Meteor. Soc. Bosart, L. F., V. Pagnotti and B. Lettau, 1973: Climatological aspects Meteorological Reports 1948: Aviation meteorology, of South Page 12
or at the boundary of the convective regions, it is pos- By so doing we must insure the energetical consistency sible to evaluate the diffusion fluxes by some existing of both the cumulus and diffusion components. Since method. there is now continuity in the definition of Qdif, ez dif, Thus, we postulate the following form of Qı - Qr the cumulus components describe only internal redis the consistency is self-preserved. On the other hand, and -Q2 when deep convection is present: tribution of heat and moisture, and fallout of rain: The 1 as vertical integral of the moist static energy on the conQ- QR + K(sc – s) (3a) vective column must be conserved by the cumulus Cp др components. For the simple case of a convective region bounded by cloud top p, and cloud base Pb, this reads -Q2 = + Kac-9) 3b dp (7) The first terms in the rhs of (3a-b) combine with the 8 vertical advection in (la, b) to give place to the vertical Combining (6a, b) and (7), we obtain a relation between advection by the “real” large-scale velocity. The second K and w* terms describe the detrainment as in Kuo's scheme. There is no turbulent diffusion, since we do not know how to compute it. We note that this formulation is + Fn(pb) – Fulp.) equivalent to a “symmetric Kuo scheme,” especially K (8) if w* behaves like w. However, it is also close to the (he – h) Arakawa and Schubert formulation, since the first 8 terms in (3a, b) may be interpreted as cloud-induced where Fn = F, + LF, where Fn = F, + LF, is the flux of moist static energy. subsidence. Thus, the scheme is closed if we know w* and the cloud To clarify the treatment of the diffusion fluxes in profiles. this approach, we will in the following split the heat An ideal closure equation for w* is not known. source and the moisture sink into their cumulus and However, many of the ideas underlying the Arakawa diffusion components. and Schubert scheme may be used as guidelines. The Qi - Qr = Q. Cu + Q, dif (4a) closure states that there is a quasi-equilibrium between large-scale and cloud-scale processes. Thus, both large-Q2 = -Q2Cu + (-Q2 dif). (4b) scale and cloud-scale information must enter the deAs mentioned before, we assume that outside the re termination of w*. We choose to incorporate the cloudgions of deep convection (or at the boundary) a usual scale information by modeling the vertical structure of formulation of the vertical turbulent exchange is valid. w* after that of a single updraft, thereby introducing Defining Fs, Fq as these fluxes, we have the knowledge of the location of the maximum insta bility in the vertical, and the large scale information Q.Cu = 0 (5a) by relating the overall magnitude of w* to the large scale moisture convergence as in the Kuo scheme. This QzCu = 0 (5b) is still fairly complicated to model, and we will assume 8 that since the cloud accomplishes conversion between a F (50) static and kinetic energy, the single most important Cp op parameter is the excess of moist static energy of the Lg a cloud over the environment he – h. Fq (50) Other parameters have been tried to model the verCp op tical structure along the same lines, e.g., the cloudoutside the convective regions, and inside these regions, environment difference in dry, or moist virtual static we may write formally: energy. The best results have been obtained, however, with he - h. Another advantage of this choice is that as + K(Sc – s) – 8 (6a) hc, as a conserved quantity, is less dependent on the др cloud model. Dimensional analysis suggests that the vertical velocity of the single updraft should be proL af, -QzCu = 8 ) h) w* = a(he - h)/2 = (9) g af li! dif (60) Cp op where a is a dimensional constant containing only large-scale information. To determine a, we use Kuo's (60) (1965) hypothesis, as a simple and efficient procedure: Cp op The total moisture convergence in the convective re
l- clo [ + K(qe – 9) - ] (6b) portional to (he – n'", and we consequently postulate Page 13
tions than to adjust zo at an optimal value. It would of the scheme into the 3D numerical model. In some be physically more significant to use the same speci- cases, dry instability and moist instability occur tofication of zo as in the 3D model, and to let the ID gether, and it has been found dangerous to subtract model adjust on a long run. This would give insight the diffusion term from the convection term, making into the quality of the interaction between the diffusion the sum of the two independent of the stability. Indeed, and convection schemes. However, this would require in these occasions, unstable lapse rates have been found the knowledge of the large-scale forcing on the wind, to persist in the simulations. To remedy this problem, and a prognostic equation for the wind. This is outside the vertical diffusion term in (6a) has been replaced by the scope of this study. For the time being, we will only its vertical average through the depth of the convective check the consistency of the surface fluxes with the region; no modification in the other equations of the predicted Q, and Qz. scheme is needed to preserve consistency. (iv) Convection is parameterized according to the method developed in the former Section, with the fol- C. Results lowing choices for the options left open. The cloud profiles Tc and qc are determined as those of a bulk The time-height variations of (Q -Qr)comp and convective updraft starting from the lowest level of the Qzcomp for the 161 single time step predictions are dismodel, as explained in the Appendix. As this level is played in Fig. 2. Clearly, the scheme is successful at usually unsaturated, the moist adiabat construction reproducing the general patterns and magnitude of the starts from the wet-bulb characteristics, i.e. the satu- day-to-day evolution of Q, and Q2. However the seemrated state which has the same moist static energy as ingly good agreement between Figs. 1 and 2 must be the environment. This allows for the computation of critically assessed. First, the main part of the information entering the the properties of the next higher level without computing first the saturation point characteristics by a dry advection of T and q. This one enters also the deter determination of Q. obs and Q2obs is the observed vertical adiabat construction. However, since this method may produce clouds dryer or colder than the environment mination of Q.comp and Q2comp, through the use of the at some levels, and always at the lowest level, the de vertical integral of the large-scale moisture tendency. trainment is not considered at those levels, and (8) is Therefore, the global magnitude of the effect, as meamodified accordingly. sured for instance by the rainfall rate, is biased towards the observations, and only the vertical structure of the The moist adiabat construction itself incorporates results may be considered as a significant result. Second, even in these ideal conditions, there are inentrainment of environmental air, according to the deed discrepancies between the computed and observed method suggested by Molinari (1982), as the clouds values. Negative values of Q. - Qr appear occasionally with no entrainment have been found to overshoot the tropopause. The entrainment rate is given a realistic during the undisturbed periods, and are badly reprovalue of 5 x 10-5 m-'(this is the value used for the values are sometimes predicted near the surface as a duced by the scheme. On the other hand, large positive highest cloud by Arakawa and Schubert, 1974). Finally, ice phase is taken into account by varying the value of consequence of the adjustment in the single time step, ice phase is taken into account by varying the value of which have no counterpart in the observations. The of ice whenever T is below 273.15, disregarding any other prediction of Q2 is in general less good than the one of , complication. Further details are given in the appendix. teresting structures that can be traced in the budget to Q. complication. Further details are given in the appendix. 01. In the boundary layer Qzobs has a number of inOnce the cloud profiles are determined, the convective regions are defined by the following procedure: The the large time-variations of q near the surface during buoyancy of the updraft is measured by the excess of diction which favors a kind of climatological behavior convective events. This is not reproduced by the previrtual temperature Tvc – Ty. If the buoyancy is negative at all levels, there is no convection. If it is positive tion error by of the moisture. We have computed the mean predicat one level at least, the convective region extends from the surface to the highest level where the buoyancy of 1 = E(Q.1,2) the updraft is positive. This has been adopted because Σ Ps of the difficulty to obtain a good prediction of the diffusion fluxes below convective clouds, and it is felt that The result is EC The result is E(Q) = 1.98 K day-!, and E(Q2) = 2.25 the surface fluxes are more reliable than the fluxes pre- K day-!. This may be considered as a good result, if K dicted at the top of the boundary layer by the turbulent compared to the largest terms in the heat and water diffusion scheme. Finally, the vertical integral of the budgets, reaching values of 15 K day-' during convecmoisture convergence on the convective region is tive events. However, if one recalls that the two main computed, and the scheme is activated only if there is terms in these budgets, the vertical advection and the net convergence of moisture. cumulus term, cancel each other to give a residual of A minor modification of the formulation (6) has the order of magnitude of the next important term, been made necessary by the numerical implementation i.e., the radiative cooling ~ 1 K day-', this is clearly Page 14
REFERENCES Johnson, R. H., 1980: Diagnosis of convective and meso-scale motions during Phase III of GATE. J. Almos. Sci., 37, 733-753. Kanamitsu, M., 1975: On numerical weather prediction over a global Anthes, R. A., 1977: A cumulus parameterization scheme utilizing tropical belt. Rep. No. 75-1, 282 pp. [Available from the Dea one-dimensional cloud model. Mon. Wea. Rev., 105, 270 partment of Meteorology, The Florida State University, Talla286. hassee, FL 32306.) Arakawa, A., and W. H. Schubert, 1974: Interaction of a cumulus Krishnamurti, T. N., Y. Ramanathan, H. L. Pan, R. J. Pasch and J. cloud ensemble with the large-scale environment. Part I. J. Al Molinari, 1980: Cumulus parameterization and rainfall rate I. mos. Sci., 31, 674-701. Mon. Wea. Rev., 108, 665-672. Augstein, E., and H. Hinzpeter, 1982: The structure of the atmo S. Low-Nam and R. J. Pasch, 1983: Cumulus parameterization spheric boundary layer under different convective conditions. and rainfall rate II. Mon Wea. Rev., 111, 815-828. GARP Publ. Ser. 25, 345–387. (Available from the Secretariat Kuo, H. L., 1965: On formation and intensification of tropical cyof the World Meterological Organisation, Case Postale no 5, CH clones through latent heat release by cumulus convection. J. 1211 Geneva 20, Switzerland.) Atmos. Sci., 22, 40-63. Betts, A. K., 1974: Further comments on “A comparison of the , 1974: Further studies of the parameterization of the influence equivalent potential temperature and the static energy". J. Atmos. of cumulus convection on the large-scale flow. J. Atmos. Sci., Sci., 31, 1713-1715. 31, 1232-1240. Charnock, H., 1955: Wind stress on a water surface. Quart. J. Roy. Lord, S. J., 1978: Development and observational verification of a Meteor. Soc., 81, 639-642. cumulus cloud parameterization. Ph.D. dissertation, 359 pp. Cho, H.-R., 1977: Contributions of cumulus cloud life-cycle effects (Available from Dept. of Atmospheric Sciences, UCLA, Los to the large-scale heat and moisture budget equations. J. Almos. Angeles, CA 90024.) Sci., 34, 87-97. 1982: Interaction of a cumulus cloud ensemble with the largeand L. Cheng, 1980: Parameterization of horizontal transport scale environment. Part III: Semi-prognostic test of the Arakawaof vorticity by cumulus convection. J. Atmos. Sci., 37, 812-826. Schubert cumulus parameterization. J. Atmos. Sci., 39, 88–103. and A. Arakawa, 1980: Interaction of a cumulus cloud ensemble Cox, S. K., and K. J. Griffith, 1978: Tropospheric radiative divergence with the large-scale environment. Part II. J. Atmos. Sci., 37, during Phase III of the GARP Atlantic tropical experiment. At 2677-2692. mos. Sci. Paper No. 291, 166 pp. (Available from the Department Louis, J.-F., 1979: A parametric model of vertical eddy fluxes in the of spheric Science, Colorado State iversity, Fort Collins, CO.] Page 15
with the cold front, but not with the precipitation that provided the information necessary for the diagnostic was occurring along an "unanalyzed" warm front. computations in both cases studied. The outcome, In this paper, I present results of a study that applies though necessarily limited in scope, has been so enQ-vector analysis to weakly baroclinic, summertime couraging and the means required to attain it so nearly situations in order to determine what, if any, relation- trivial, that I could not pass up this opportunity to ship the diagnosed circulations might have with or- share with you the operational possibilities of the ganized convective storm systems, a major summer- scheme with the hope that future tests will bear out time forecasting problem in middle latitudes. our initial enthusiasm. The synoptic conditions associated with many me- The following sections describe the weather situation soscale convective complexes that occur in middle lat- facing a duty forecaster, say, in Denver or Amarillo, itudes have been documented by Maddox (1983). By on the morning of 25 June 1982. Together, we will Maddox's satellite image definition (1980), the cloud look at some (but not all) of the available guidance for shield of an MCC (Table 1 lists the meanings of ac- considering convective storms in the forecast, consider ronyms used in this paper) must exhibit an area of at the results of the omega diagnostics, and then look at least 100 000 km2 in which the IR temperature is at what actually happened that day in eastern Colorado least as cold as - 32°C. Inside this area must appear a and adjacent areas. We will then take a cursory glance colder (< -52°C) region of at least 50 000 km2, and at the temporal continuity of the omega diagnostics both areas must persist for at least 6 h. Typically, MCCS from 3-hourly soundings for a more problematical case, develop in a region of the lower troposphere that is 20 May 1976 AVE-SESAME V. Technical details are characterized by warm air advection, veering of wind relegated to appendices. with height (but with small speed shear), deep moisture content (>10 g kg-' in the lowest 200 mb), and significant convective instability. Initiation of MCCs often 2. The synoptic situation, 25 June 1982 takes place in advance of a weak trough in middle troposphere. An observed characteristic of MCCs is the Some of the first available information sources indevelopment of a meso-alpha-scale (250–2500 km) dicating what the atmosphere is doing prior to the circulation whose duration is considerably longer than morning forecast cycle are the IR satellite images. One the individual convective components (Fritsch and of these (Fig. 1) obtained near the time of the morning Maddox, 1981). rawinsonde release confirms what inhabitants of The tools brought to bear in this study were a firm northern Texas have known all night, that widespread background in objective interpolation techniques and thunderstorms have been occurring, Amarillo being in their implementation on computers, a microcomputer one activity zone. Over the central and northern Rocksystem of modest capacity, a quite rusty and somewhat ies, middle and high cloudiness persists in advance of dated understanding of QG theory, and a group of in- a short wave trough at 500 mb (Fig. 2) that is about to terested colleagues urging, guiding, and educating me penetrate through a larger scale ridge. The desert states along the way. Rawinsonde pressure-height data alone of the southwest and most of California are experienc ing only patchy middle and high clouds. At first glance, it looks like another standard summertime day over Table 1. Explanation of acronyms used in the text. the mountainous west: more or less scattered thunAFOS Automation of field operations and services dershowers over the mountains, more in the north, less AVE Atmospheric variability experiment in the south. The Amarillo forecaster's experience with CVA Cyclonic vorticity advection nighttime thunderstorms probably tells him that those IBM PC International Business Machines personal computer in his area will dissipate by noon (Wallace, 1975). IR Infrared If, however, our two forecasters recognize surface LFM Limited-area Fine-mesh Model Lifted index features associated with high-plains severe thunderMCC Mesoscale convective complex storms (Doswell, 1980), they may experience their first MOS Model output statistics doubts about how typical the day will be when they NASA National Aeronautics and Space Administration look at the 1200 GMT surface map. Principal features NMC National Meteorological Center NSSFC National Severe Storm Forecast Center (abstracted in Fig. 3 from the NMC version) constitute NWS National Weather Service the warning flags: continental polar air of high moisture POP Probability of precipitation content streaming upslope over the high plains from PVA Positive vorticity advection; same as CVA in Texas to Wyoming behind a front that has penetrated Northern Hemisphere across the mountains into Utah. The 35 m s-'jetlet at QG Quasi-geostrophic QPF Quantitative precipitation forecast 200 mb over southern California (Fig. 3, inset) lends SESAME Severe Environmental Storms and Mesoscale a further complication to the forecast picture. Very Experiment likely our forecasters already have strong suspicions SUNYA State University of New York at Albany that something more than a standard thunderstorm TDL Techniques Development Laboratory forecast is called for today. Page 16
25 JUN82 122 500-mb_Eta & dZ
mb layer suggests that an analysis of geostrophic forcing 12 this 107 based only on thermal advection of cyclonic vorticity may not tell the whole story. The deformation com10 ponents, which the Sutcliffe (1947) analysis ignores, could be significant. In Hoskins et al. (1978) Q-vector formulation of the diagnostic omega equation, the divergence of Q is proportional to omega (vertical motion in a pressure coordinate system). In the version employed here 10 (which is described in appendix A), the troposphere is divided into two overlapping layers, 850-500 mb and 700–300 mb, and divergence of Q is computed at two (10-5 S-1) levels, 700 and 500 mb. The Q-vectors for the 700 300 mb layer are shown in Fig. 13. Hoskins and Pedder FIG. 11. 500 mb absolute vorticity (10-45-'; vy lines) and 700300 mb thickness contours (interval 40 m; light lines). Compare with (1980) instruct us to expect rising motion where Fig. 4a. vectors converge, sinking where they diverge. Furthermore, they demonstrate that in the lower levels of the layer ageostrophic wind is proportional to and in the (CVA) in Utah just as Fig. 4a does. In this case, a fore direction of the Q-vector. The pattern of Q-vectors in caster could deduce from either thickness pattern that Fig. 13 and the resulting pattern of omega (actually, CVA is occurring in Utah, northwestern Colorado and divergence of Q in Fig. 14) indicate that a relatively southwestern Wyoming. However, for determining strong ageostrophic circulation is occurring over mean temperature advection in a layer, the 1000-500 southern California and western Arizona. Possibly this mb thickness with the 500 mb geostrophic winds would is associated in part with the jetlet at 200 mb noted not be an appropriate combination. An important dif earlier (Fig. 3 inset); possibly it is due in part to oroference exists in the sign of the curvature (Laplacian) graphic forcing (results not shown). Being near the edge of the thermal pattern, especially over western Colo of the data domain, we cannot be confident that the rado. Figure 4a shows a thermal ridge (negative Lapla- height fields have been represented with sufficient accian) in the 1000-500 mb layer along the Utah-Col curacy. Even if the circulation has been diagnosed with orado border, while Fig. 11 shows the thermal ridge in some measure of fidelity, the upward branch is working the 700–300 mb layer to be farther to the east over with low humidity air that is quite stable (Fig. 6), so central Colorado with a different orientation through thunderstorms are not expected to develop over western Wyoming toward the northwest. In this upper layer, Arizona. there is an indication of a thermal trough (positive La The next most significant upward motion center is placian) over western Colorado. This difference is suf- more confidently analyzed well within the data domain ficient to change the computed sign of thickness ad and is associated with the short wave trough in Utah. vection at 500 mb from warm to cold over the region At this point in our experience with these diagnostics, from southwestern Wyoming to north-central New Mexico (results not shown). As will be explained later, the correctly computed cold advection in the 700–300 mb layer has an important bearing upon the stability 11 10 1011 12 12 11 tendency over the region from central Wyoming to eastern New Mexico. At 700 mb (Fig. 12), a significant cyclonic vorticity center is already on the Utah-Colorado border, indi 0 10 cating a westward tilt of the axis of maximum CVA N with height, a typical configuration for a developing 10 short wave. An area of CVA and low level warm ad 16 10 vection (not shown) extends from north-central New 9 Mexico through the western half of Colorado and into southern Wyoming, suggesting that the area of maxi N mum ascent in the lower troposphere is already east 9 of the area predicted by the LFM to occur 12 h after 08 this diagnosis (Fig. 5). Note that the depth of the polar 25 JUN82 12Z 700-Mb Eta & dz6 10 air behind the surface front is apparent in the relatively (10-5 5-1) 7. 8 9 98 strong thickness gradient extending from north-central Colorado through the Texas panhandle. The presence Fig. 12. 700 mb absolute vorticity (heavy lines) and 850-500 mb of a strong (for the time of year) front in the 850-500 thickness contours (light lines). Units are same as in Fig. 11. Page 17
FIG. 27. Time-series plots of 500 mb heights (m) at the 23 NWS sounding stations participating in the AVE-SESAME V experiment. Heights observed at 1100 GMT 20 May 1979 are at the right of the horizontal time lines, which represent the ensemble average height equal to 5787 m. The scale of height deviations from the ensemble average height is shown at the lower left.
storms cannot be explained by QG theory, but we must lieve there is a tendency among many meteorologists continue to look for general relationships in large-scale to give more weight to “CVA” than to the thermal processes as they may influence the overall structure advective contribution in winter, and conversely in of thunderstorm groups. summer. Perhaps the sum of their experiences tells them this interpretation works for the most part. How7. Discussion ever, it is easy to find examples where this predeliction will lead to an incorrect analysis, and a numerical What has been demonstrated here is the apparent computation is required to correctly estimate the largeutility, and somewhat surprising power, of some simple scale forcing. At the present time, operational numerdiagnostic products that are based on nothing more ical forecasts produced by the LFM do not perform than geopotential height fields and quantities derived reliably in summertime convective situations assofrom them. That they seem to add valuable information ciated with weakly baroclinic systems. Of course, Sutto a fictitious forecasting process2 is decidedly encour cliffe's scheme can be used as the basis for such operaging. However, a considerable amount of testing and ational computations, as Trenberth (1978) has shown. evaluation remains before their general value can be However, since Hoskins et al. (1978) Q-vector forassessed. mulation accounts for the complete geostrophic forcing One of the problems with qualitative application of including the deformation components, it would apSutcliffe's scheme to vorticity and thermal patterns on synoptic charts is determining the relative contributions the usual formulation because it does not require the pear to be a better choice for general application than of vorticity advection and thermal advection to the computation of the difference between two nearly cangeostrophic forcing. Although I have no statistics, I be celling effects. Neither Q-vectors, divergence of Q, nor the inferred vertical motions are easily visualized from conventional 2 A comparison could have been made with the actual forecasts representations of synoptic data, i.e., constant pressure issued from the Denver and Amarillo NWS offices, but I chose not to, reasoning that the purpose here is to introduce diagnostic products charts and thickness patterns. Fortunately, the comthat are not now available to NWS forecasters, not to test the skill puter algorithm required to solve for the Q-vector fields of a hindcast against an operational forecast. is relatively simple, especially when written in terms Page 18
APPENDIX A strophic wind. The vertical difference in this quantity Omega Diagnostics Formulation is a measure of the local tendency of stability (Saucier, 1955; p. 397), where stability is defined as in Appendix S. C. Lin (personal communication, 1984) derived A, but where 6 is replaced by virtual 0, and by definition the QG omega equation in the pressure coordinate sys- refers to dry adiabatic processes only. I offer here a tem simple demonstration of proportionality rather than a aw rigorous proof. The stability tendency can be written до Do do V.Vo (B1) rameter, and the components of Q are expressed by at Dt др av ΟΦ av ao For present purposes, we neglect material changes (Do/ Q (A2) дх др ду др, Dt = 0) and apply the approximation (V = V2). Since we are interested in areas in which dolap is small and Lin's formulation also included forcing terms for dia- decreasing, we omit the vertical advection term. In such batic heating and the longitudinal variation of Coriolis cases, omega is likely to be small and increasing, but (B), both of which I have chosen to omit. For the present w <V and dolap < Vo. application, I am more interested in the geostrophic With these approximations forcing prior to convection rather than the contribution to omega from convection. The effect of the diurnal до (B2) heating cycle also has been neglected, although it is at likely to be an important contributor to omega at 700 mb over the mountainous west. The term involving In a manner equivalent to the definition of o used in is usually an order of magnitude smaller than the other (A1), Hoskins and Pedder (1980) write o = - hd7/dp where h = (R/p)(p/po)" forcing terms and is here omitted for simplicity of where h = (R/p (p/po)" = a/8. Taking the horizontal 0. computation. variation of o in a layer defined by P, and P2 and re oc On the basis of (A2), I computed the Q-components calling that Pp - $pi = dø ac 7, we have using the horizontal variation of geostrophic wind at Voo V[(89)/ap). (B3) 700 mb for the 850–500 mb layer, and at 500 mb for the 700–300 mb layer. This was done as a convenience. What I have calculated to produce the tendency con a , since the actual mean pressure levels for these layers tours in Figs. 3 and 20 is the differential advection of are 652 and 458 mb. Thickness of the appropriate layer thickness by the geostrophic wind, which can be exwas determined from the differences in the reported panded into two terms, heights. Horizontal gradients of thickness also representa[V .\(89)]/ap horizontal gradients of mean virtual potential temperature. Using thickness temperatures eliminates the need [(avg/ap). (89)] + [V .v(89)/ap)]. (B4) to input both temperature and humidity observations But the first scalar product is zero (the component of as is required for a single-level Q-vector computation the thermal wind along the thickness gradient) and the in order to compute virtual temperatures explicitly. second is, according to (B3) and (B2), proportional to At the interior of the grid, centered differences were the local stability tendency. used for the gradient computations; at the edges, the gradient at the two adjacent interior points was extrapolated outward. The Q-components were then cal APPENDIX C culated at each point of the mesh, but only values at the interior points are displayed. The Interpolation Function Rather than solving (Al) explicitly for omega, I chose to depict only the Q-forcing, since it is really that pattern and its sign that are important for present purposes Interpolation of height data from reporting stations to gridpoints was accomplished through a two-pass rather than omega's magnitude. Thus, spatial variations in static stability have not been accommodated Barnes (1964; 1973) analysis for which the weight function has the form in the results. WT = exp(-R2/4C), (C1) APPENDIX B where R is distance from data point to gridpoint and Stability Tendency C is a selectable parameter that governs the response of the interpolated pattern to the initial data. That is, Since the necessary data were available for each layer, C determines the amount of smoothing and the wave C Page 19
of the associated Euler equation with Neumann lateral where S, is the heat source or sink and Fo the turbulent boundary conditions, or by inversion of the matrix ob- subgrid-scale dissipation of 0. tained from the discretization of F through finite dif- In order to arrange (8) for deducing the horizontal ference approximation. mean values of @ci, the following hypotheses are made: Solutions to (6) are not unique and are known only (i) The subgrid scale dissipation Fo is assumed to be to within a constant. Thus, this method allows to retrieve the deviations II* and one of the pressure and negligible with respect to the other terms; (ii) The heat source or sink S, are expressed as fol"virtual cloud temperature" fields from their horizontal lows: mean values P and T: • When air is saturated, s, is supposed to result TIT(x, y, zo) = II/(x, y, zo) – P(zo) , y1 y, , ) from saturated adiabatic vertical (up or down (7) 0*1(x, y, zo) = 0c1(x, y, zo) – T(zo) ward) motions: Brandes (1984) has proposed a method where the So = WYSA three-dimensional buoyancy field is first deduced from where w is the vertical velocity of air and Ysa the integration of a two-dimensional elliptic equation is the saturated adiabatic lapse rate; for buoyancy (obtained by dot-multiplying the curl of . When air is unsaturated, Se is supposed to result the vorticity equation by k,the vertical unit vector) with from the evaporation of raindrops: the hypotheses that the pressure dependence is negli L gible with respect to the other terms and that the buoy So E ancy is equal to zero on the lateral boundaries. Then, Collo the three-dimensional pressure field is derived from where L is the latent heat of condensation, Ilo the integration of the three-dimensional elliptic equa is the nondimensional pressure and E is the tion (derived by taking the divergence of the motion evaporation rate of rain. equations). Similarities between both approaches result from analogies between the elliptic buoyancy equation Owing to the complexity of the microphysical and the Euler equation of (6) for the virtual cloud temperature, and between the elliptic pressure equation process involved in the net production of ice particles, their contribution in the heat source or sink So is not and the Euler equation associated to the minimization of the functional F, written as: taken into account here. Nevertheless, their influence on the thermodynamic field may be estimated from 1 Dv recent studies with numerical models of convective , (Covo LDt clouds including the ice phase (e.g.: Cotton et al., 1982; Lin et al., 1983; Lord et al., 1984). It is Oct generally observed that the release of latent heat 9. F dxdydz. 00 through the conversion of cloud water and raindrops to cloud ice, snow and graupels reinforces the updrafts, Here F, differs from the expression of (6) for pressure, while the melting processes which concentrate the for the three components of the motions equations, cooling in a layer below the 0°C isotherm contribute C therefore the gradients of II, parallel to x, y and z, are to initiate or to strengthen the downdrafts. Therefore considered. The following section shows an alternative the neglect of the ice phase will probably reduce way for retrieving the complete thermodynamic fields. slightly here the mean vertical temperature profile Here, the pressure dependence of By in (4) is calculated since warming in the upper levels and cooling in the from the retrieved pressure perturbations and no hy- lower levels will be little underestimated. pothesis is made for the temperature values on the (iii) Due to lack of information about the temporal boundaries. A simplified thermodynamic equation is evolution of the thermodynamic field, the temporal used to calculate the horizontal mean values T for derivative of 0 will not be taken into account here. This temperature at each altitude, then the horizontal mean means that, among, among the various pressure and values P for pressure are deduced from the vertical temperature fields solutions of (1) and (8) for the ob projection of the equation of motion. served wind field, only the stationary solutions will be considered. Then, the temperature field is developed b. The processing of thermodynamic equation at the first order as The thermodynamic equation for potential temperature e comes from the first law of thermodynamics 0(x, y, z) = 0.(z) + 0,(x, y, z). with the addition of diffusion (Wilhelmson and Ogura, 1972), it may be written as Then, (8) may be written as Page 20
between the radar derived relative virtual cloud tem- bations II* and on their horizontal mean values T and perature at the lowest level (where the contribution of P for each altitude. The non-dimensional pressure 9c in Oci is probably small) and the surface network perturbations II, are transformed in dimensional presmeasurements. It has been shown in RTPP that a sat- sure perturbations p, through (19). The two-dimenisfying agreement between the two kinds of results is sional fields of pressure perturbations pi(x, z) and virobtained when a value of 4°C is subtracted to the radar tual cloud temperature 0c1(x, z) are shown in Fig. 6. data. Therefore, T (1 km) is taken here as equal to The thermodynamic structure of the frontal con(-4°C) and the vertical profile of T deduced from in- vective region of this squall-line will not be detailed, tegrating [dT|dz], with this initial value is also dis- since it has already been discussed in RTPP, but the played in Fig. 4. features which did not appear from the only relative The vertical gradient (dP /dz] of the horizontal mean perturbations of pressure and of virtual cloud tempervalues of the pressure perturbations II 1, deduced from ature, will be emphasized here. The two-dimensional these results through (18), are shown in Fig. 5. Likewise, field of oci (Fig. 6a) confirms the analysis made in as deduced in RTPP from comparison between the RTPP, but displays more clearly the dynamical proradar-derived relative perturbations of pressure and the cesses occurring within this region of the squall line. simultaneous mesonetwork measurements the initial At low levels, one can clearly see, when comparing value for the integration of (dP/dz] will be taken to be with Fig. 2, that the frontward flow is a cold one with 0.6 mb. For convenience these mean values P of pres- maximum negative perturbations up to -5.5°C near sure perturbation are expressed in terms of dimensional the ground and far from the interface between the two pressure P (in hPa or mb) through: flows. Lessening of this temperature deficit near the entering flow results probably from mixing between 1 P(mb) (Covopo)P (19) the two air masses. The virtual cloud temperature per100 turbations retrieved in the frontal updraft display a pewhere ovo and po are the virtual potential temperature culiar thermodynamic structure. Although the inflowand the air density in the unperturbated state at the ing air in the lowest levels is associated to slightly pos itive perturbations, negative values are observed up to altitude considered. The vertical profile of P is also 3 km within the updraft. As detailed in RTPP, this displayed in Fig. 5. Then, the pressure perturbations II, and the virtual situation reveals the influence of the propagation of cloud temperature perturbations Oct can be deduced the cold low-level frontward flow: initiation of the upthrough adding to the radar-derived relative pertur- bility but it results here from an upward pressure gra ward motions is not a consequence of thermal instadient force due to the nonhydrostatic pressure perturbation at the interface between the two oppositely directed flows. Morever, as deduced from the 0348 raALTITUDE (km) diosounding, the temperature excess of an undiluted air parcel ascending moist-adiabatically from the ground level should remain smaller than 1°C to 2 km. - 10 Taking account of the cloud water loading, as long as the precipitation process is not triggered, the effective buoyancy (temperature buoyancy minus cloud water loading) should remain close to zero up to 2.8 km, so that the contribution of (-9.00] in the virtual cloud temperature 0c, may compensate the actual temperature perturbation 0,. At higher altitudes and in the inner part of the squall line, once the cloud water loading has been released through the fall-out of the heavy precipitations, larger perturbations are observed. The maximum values of +4.5°C at (35 < x < 50 km) are in good agreement with the maximum temperature excess of 5°C at the altitude of 4.5 km deduced from the 0348 radiosounding. The negative perturbation observed in association with the air entering the squall line between altitudes 5 and 7.5 km may result from the slight lifting (due to the frontal updraft below) of the colder layer observed between 4 and 6 km on the -5 12 0348 sounding (see RTPP). Likewise the negative per.0: dP/dz (10'm') turbations above altitude 10 km are probably due to : P (mb) forced lifting of stable layers in front of the squall line, FIG. 5. As in Fig. 4 but for (dP/dz) and P. morever evaporation of precipitations condensed below Page 21
deduced from a multiple-Doppler radar experiment. =24°C The reliability of the fields of relative pressure and virtual cloud temperature has been previously insured in RTPP through comparisons between the radar-derived fields at low altitude and surface network measurements. The similarity of the constants determined for the two and three-dimensional cases shows the “stability" of the method and the obtained fields of pressure and virtual cloud temperature seem realistic when compared with previous results. The pressure field with a low center within the inner convective region resem bles to that displayed in LeMone et al. (1984) for fast Ī AFTER moving squall lines, and the different vertical forces are qualitatively and quantitatively similar to those deduced from three-dimensional models of deep convective clouds (Schlesinger, 1984). Despite these encouraging features, several points still remain unsatisfying in this method. First, the saturated adiabatic lapse rate (in term (c) of (10)] is supposed to be constant at each altitude, depending only on the pressure and temperature in the environment. Although this hypothesis is approximately correct since VIRTUAL TEMPERATURE (°C) the vertical gradients of pressure and temperature are larger than the horizontal ones, it becomes less justified when large perturbations occur within the cloud. One Fig. 11. Vertical profiles of virtual temperatures in the upstream consequence is that the adiabatic lapse rates will be environment (before), deduced from the 0348 radiosounding, and in the outflow of the convective region (after). The vertical profiles underestimated in warm regions and overestimated in of virtual temperature for two saturated adiabatic curves (0 - = 20°C cold regions. For instance, at the 600 mb pressure level, and e'w = 24°C) are also indicated. where the potential temperature in the environment is 317.5°K and the saturated adiabatic lapse rate Ysa is 5.7°C km-!, a temperature perturbation of 5°C would ciated to a positive vertical gradient force with a small induce a variation of 0.7°C km-'in Ysa (the pressure contribution of positive buoyancy at x = 5 km, there- dependence is negligible as compared to the temperfore, in this case, the driving force is the pressure one. ature dependence). The method described here allows The vertical thermodynamic structure of atmosphere the complete determination of the pressure and temwithin the inner convective region can hardly be an- perature fields, so it should be possible to use these alyzed in terms of perturbations with respect to a basic results iteratively through calculating on each point hydrostatic equilibrium. Although such an analysis is better estimates of the input adiabatic lapse rates as justified in the very frontal part where the updraft de- functions of the output retrieved values (taking into velops almost in the upstream air, the concept of per account, if possible, the contribution of the mixing raturbation is no longer adapted in the inner region. A tios of water vapor and of cloud water in the virtual vertical thermodynamic structure different from the cloud temperature). environmental (upstream) air takes place with colder The second point deals with the mixing ratios of air in the frontward flow below 3.5 km and warmer water vapor and cloud water. Although the presence air above, resulting from the lifting of the unstable up- of precipitation is simply determined by reflectivity stream layers. Then, the large values of vertical pressure values larger than 0 dBZ, one must specify a priori if gradient (Fig. 12c) and virtual cloud temperature the air is saturated or not. The relatively simple strucbuoyancy (Fig. 12d) denote rather a different hydro- ture of the frontal part of this squall line, the availability static equilibrium, although convective motion (as seen of independent surface network measurements and the on the vertical acceleration in Fig. 12a) may exist in results obtained in HAC concerning the water vapor the inner region. and cloud water fields facilitate the preliminary de scription: air is saturated everywhere except in the high 4. Discussion level entering flow and in the low level frontward flow. Moreover, since the vertical motions in both regions It has been shown that the processing of the equa- are small, the influence of these hypotheses on the caltions of motion and of the thermodynamic equation culated horizontal mean values T and P is not very allows retrieval of the thermodynamic fields associated important here. However, the situation could be very with the three-dimensional wind and reflectivity fields different for more complex events such as isolated Page 22
The Global Climate for December 1984-February 1985: A Case of Strong Intraseasonal Oscillations
Climate Analysis Center, National Meteorological Center, NWS/NOAA Washington, DC 20233
1. Introduction mean SLP for the period 1935-84. During January the SLP rose to above normal at both stations. This inStrong intraseasonal oscillations in atmospheric cir phase SLP fluctuation may be considered an example culation and related fields characterized December of what Trenberth (1984) defines as “noise due to small1984-February 1985 (DJF) over many regions of the scale or transient phenomena." The Southern Osciltropics and extratropics. Convective activity, tropo lation Index (SOI) or "signal” remained nearly constant spheric winds and sea level pressure (SLP) within the and close to zero during both December and January. tropics underwent large and coherent oscillations that appear to be related to fluctuations in geopotential but remained above normal at Tahiti , resulting in a height and temperature over and in the vicinity of rise in the SOI. North America. These oscillations exhibit a behavior consistent with that associated with the 30-60 day os b. Sea surface temperature cillation described by Madden and Julian (1971, 1972), Weickmann (1983) and Weickmann et al. (1985). The seasonal mean and anomaly fields of SST are Tropical oceanic and atmospheric anomaly patterns shown in Fig. 1. The anomaly pattern shows a decrease for the entire season are similar to those found in com- in positive anomalies in the Gulf of Guinea, an increase posites for the northern winter prior to the onset of El in negative anomalies in the equatorial central and Niño-Southern Oscillation episodes (Rasmusson and eastern Pacific and the development of positive anomCarpenter, 1982; Arkin, 1982). This Southern Oscil- alies east and northeast of New Zealand relative to the lation signal is quite weak however, especially when values observed during the preceding season. The compared to the “noise” level produced by the intra- anomaly pattern in the tropical Pacific is similar to the seasonal oscillations. composite antecedent conditions for the year preceding In section 2 we consider seasonal and intraseasonal El Niño (Rasmusson and Carpenter, 1982). features of the global tropics. This is followed by a discussion of features for the Northern Hemisphere (sec- Outgoing longwave radiation (OLR) in the tropics tion 3) and Southern Hemisphere (section 4). The mean and anomaly fields of OLR (Fig. 2) in2. The global tropics dicate more active-than-normal convergence zones in the South Pacific, South Atlantic and in the equatorial a. Atmospheric indices in the tropical Pacific Atlantic and greater than normal convective activity over Indonesia. Positive OLR anomalies, indicating As in each of the last two seasons (Ropelewski, 1985; weaker than normal convection, lie along the equator Dickson, 1985), the outgoing longwave radiation, Pa- in the central and eastern Pacific coinciding with the cific 200 mb zonal wind and central and eastern Pacific region of negative SST anomalies. 850 mb zonal wind indices (Table 1) exhibit positive During DJF considerable variability in convective values for December 1984-February 1985 (DJF). The activity occurred on the intraseasonal time-scale as is variable behavior of the western equatorial Pacific 850 evident in the time-longitude plot of half-monthly mb zonal wind, noted in the preceding season (Dick- OLR anomalies averaged from 10°N to 20°S (Fig. 3). son, 1985), continued during DJF. Considerable vari- Greatest variability is found in the region from eastern ability is also noted in the SLP at Tahiti and Darwin. Africa eastward to the central Pacific (30°E-150°W). Pressure fell dramatically at both stations during De- These anomalies propagate eastward with time. The cember with Tahiti registering the lowest December region of South America (30-70°W) also shows conTABLE 1. Atmospheric and sea surface temperature (SST) indices for the tropical Pacific. Atmospheric indices are mean values of departures from long term averages divided by appropriate standard deviation; SSTs are mean departures of °C. Tahiti minus Darwin SOI is a Southern Oscillation Index based on standardized mean sea level pressure anomaly difference. Positive values of 850 mb zonal wind indices represent anomalous easterlies. Page 23
height anomalies are found in the Gulf of Alaska and western United States, where early season snow cover over the southeastern and eastern United States. in the northern Rockies and interior of the Northwest The precipitation percentiles (Fig. 9) show that for provided favorable conditions for radiational cooling the United States the greatest positive rainfall depar- and trapping of cold air in the valleys, and Alaska where tures (percentiles > 70) extend from the Southwest much warmer than normal conditions prevailed due northeastward to the upper Great Lakes and over to the stronger than normal ridge near the British CoAlaska. Largest rainfall deficiencies (percentiles < 30) lumbia coast. are found in the Northwest, including the northern Rockies, and along the entire East Coast. Eastern and b. Monthly circulation and climate anomalies northern Europe generally received below normal precipitation while western Russia experienced above 1) DECEMBER 1984 normal precipitation. Very wet conditions occurred along the South China coast. Large positive 700 mb geopotential height anomalies The seasonal temperature anomaly pattern (Fig. 10) (Fig. 11) were observed near Scandinavia during Deshows that much of Europe and the central and western cember associated with a persistent nearly month-long portions of Russia experienced 2-6°C negative anom- block in the vicinity of 40°E. Large negative surface alies. Most of North America had seasonal anomalies temperature anomalies over central and western Russia with absolute value less than 2°C except for the north- (Fig. 12) accompanied the block. Anomalous ridging
Fig. 9. Percentiles of precipitation for DJF 1984/85 based on a gamma distribution fit to 1951-80 base period data. Hatched area <30%; Stippled area > 70%.
FIG. 10. Temperature anomalies for DJF 1984–85. Anomalies >2°C (< -2°C) are shaded (hatched). Page 24
FIG. 21. (a) Southern Hemisphere mean 200 mb height and (b) anomalies for DJF 1984/85. Mean (anomaly) interval is 14 dam (2 dam).
1000-700 mb thickness anomaly for the region 30- Comparing Figs. 3 and 20 we note that the thickness 45°N, 75-90°W shown in Fig. 20. The remaining time anomaly tendency for the eastern United States is posseries in Fig. 20 are discussed in the next section. itive when negative OLR anomalies (enhanced con vection) shift from the Indian Ocean to Indonesia. This c. Intraseasonal oscillations corresponds to ridging over the eastern United States a feature of the December and February monthly mean Considerable intraseasonal variation characterized charts (Figs. 11 and 17). Negative thickness anomaly DJF 1984-85. As evident from the thickness anomaly tendencies and eastern North America troughing are time series in Fig. 20, intraseasonal oscillations were observed in early January and in March when negative present during the northern fall and continued at least OLR anomalies approach the central Pacific. The reuntil early northern spring. Also shown in Fig. 20 are lationship between the position of enhanced tropical time series of a zonal index, the zonal mean SLP convection and the circulation pattern over North anomaly' for the latitude band 0°-20°S and the zonal America observed during DJF 1984-85 is consistent mean 200 mb zonal wind anomaly for the equatorial with that found by Weickmann et al. (1985) associated belt 5°N-5°S. After October 1984 the fluctuations in with intraseasonal oscillations and with that for El tropical SLP (0°-20°S) are generally in phase with the Northern Hemisphere zonal index and out of phase lace, 1981). The relationships discussed here suggest Niño-Southern Oscillation episodes (Horel and Walwith the eastern United States thickness anomalies of that by monitoring tropical convection and the evothe zonal mean equatorial 200 mb zonal wind. The lution of the intraseasonal oscillations information can approximate period of the fluctuations is 60-75 days, be obtained that may prove useful in long-range prewhich is in close agreement with the period noted for diction. fluctuations in tropical OLR and in zonal wind at 850 and 200 mb discussed in section 2. 4. The Southern Hemisphere extratropics seasonal circulation features 'The SLP anomalies for 0°-20°S were obtained by subtracting the climatological mean SLP, estimated from Table 8 and Fig. 1 of The DJF 200 mb mean geopotential height and Trenberth (1981) and interpolated to obtain values representative of each half of each month from the half-monthly mean SLP obtained Page 25
A vector representation of the BOMEX thermodynamic budget data is presented which shows graphically the relationship of the fluxes and the mean layer structure.
1. Introduction differences on a thermodynamic diagram. With this technique the extent and limitations of the information The BOMEX undisturbed Tradewind budget data content in the thermodynamic budget data become has already been analyzed several times (Holland and readily visible. We shall see that the BOMEX budget Rasmusson, 1973; Nitta and Esbensen, 1974; Betts, data suggests some role for cloud-top entrainment, but 1975; Nitta, 1975) and the purpose of this short paper in view of the corrections made to the humidity data, is simply to re-express the fluxes in a simple vector this evidence can only be regarded as encouraging for format using the saturation point (SP) notation intro- further analysis of data along these lines. We also are duced by Betts (1982a). This shows very clearly the reminded that radiative cooling off cloud-tops may play coupling of the fluxes, the changing Bowen ratio with an important role in the budgets. Following Betts (1975), we may express the fluxes in the tradewind layer of the product as a mass flux a spectral model which assumed lateral entrainment (Arakawa and Schubert, 1974). This technique has gF(p) = w*(p)[C(p) – E(p)). (1) been widely applied in various forms to diagnostic studies of cumulus budget data (Ogura and Cho, 1973; We use a vector notation where C, E are representative SPs for a cloud, and environment parcel (Betts, 1984). Johnson, 1976; Houze and Leary, 1976). Betts (1975) In general, C and E vary with height. For a small fracshowed that a single mass flux represented the BOMEX a tional cloud cover, E(p) is closely the mean stratificabudget data well, and that agreement was slightly better tion, while the variation of C(p) will depend on the with a nonentraining cloud model than one with lateral entrainment. Beniston and Sommeria (1981) have cloud-environment mixing processes. If there is no mixing, then stays at B, the SP of air ascending shown that the single mass flux parameter model is a very good fit to the fluxes predicted by a three-dimen through cloud-base, which is closely the SP of the nearly well mixed subcloud layer. In the budget method we sional (3-D) simulation of a shallow cloud field. This generally know Eand F. What can be determined about suggests that perhaps the 3-D model statistics could be used to estimate in-cloud mixing processes. There has convective mass flux w* and the mean cloud SP, C? Since w* is a scalar, the flux vector F has the same recently been new observational support for an old idea (Squires, 1958), namely, that cumulus clouds may en vector orientation as the difference AS = C – E. If we train primarily at cloud-top (Paluch, 1978; Boatman consider AS to have components A01, Aqi, then the ratio of the conserved heat flux to the total water flux and Auer, 1983; LaMontagne and Telford, 1983). Betts is BR = CP40,/LAQ, (2a) parcel saturation point (Betts, 1982a) which enables a simple vector representation of fluxes in terms of SP where BR is the Bowen ratio for the conserved fluxes (Betts, 1984). In parameterizing the flux data we shall approximate static energy differences with the differVisiting scientist, NASA-Goddard Space Flight Center, Code 613, ences CpA0. Thus, we can determine the orientation of Greenbelt, MD 20771. C-E on a thermodynamic diagram directly from the
© 1985 American Meteorological Society Page 26
in GD and SL, as demonstrated in Fig. 1, the difference in our conclusions is not due to the different model formulations. In SL, the thermodynamic properties of 300 the air upstream are described on the basis of a single sounding taken from Amini Devi Island at 0000 GMT 1 July, 1979. In GD, we examined 31 dropwindsonde profiles taken between June 17 and June 30 1979 in a 400 region well upstream of the Ghats. The soundings were sorted according to whether or not offshore convection occurred on that day. The composite soundings for 500 convective and nonconvective days appear as Figs. 13 수 and 14 in GD. Inspection of the Amini Devi sounding (Fig. 9 in SL), shows that it closely resembles the com 600 posite sounding for days without offshore convection; note in particular, the 10°C dewpoint depression at 700 700 mb. The fact that the SL airflow model does not predict a destabilization of this sounding is consistent 800 with the findings in GD. We feel confident that the technique employed in SL would predict destabiliza 900 tion of the composite convective-day sounding shown 1000 in Fig. 13 in GD. The character of the Amini Devi sounding suggests that it should be considered a nonconvective day sounding. However, this appears to conflict with the FIG. 2b. Comparison of 1200 GMT 24 June 1979 Amini Devi daily rainfall totals observed on July 1 at Mangalore sounding (open circles, solid line) with 0000 GMT 1 July sounding and Calicut, as shown in Fig. 4 of SL. In fact, the rainfall from SL (triangles, dashed line). There was significant offshore con vection on 24 June 1979. records are misleading. The first problem is that, although only 24 days of rainfall data are plotted, the axis for the chart is labeled with 25 calendar dates. This introduces a one day ambiguity in the plotted records so that the significant rainfall at Mangalore and Calicut could have begun on either the first or second of July. Inspection of satellite data for the period surrounding July 1, 1979 (Young et al., 1980) shows that whereas 300 there was little deep convection during the period June 28–30, convection was widespread on the days July 25. The transition between these regimes occurs on July 1. An examination of the Amini Devi soundings for 400 this period shows that the 0000 GMT July 1 sounding used in SL has the same character as the other sound ings taken between 28 and 30 June, in particular, they 500 all exhibit an approximate 10°C dewpoint depression at 700 mb (the one exception to this was the 1200 GMT sounding on 30 June which was more moist). 600 In Fig. 2a, the 0600 GMT June 29 Amini Devi sound ing is compared to the 0000 GMT July 1 sounding 700 used in SL. They are very similar and, inasmuch as June 29 was clearly not a convective day, this suggests 800 that the sounding used in SL is not characteristic of 900 the atmospheric structure which is associated with convection in the summer monsoon. 1000 What do the Amini Devi soundings look like on a convective day? The soundings from 2 July to 4 July should serve as good examples, but unfortunately that data is missing. The only soundings available after the FIG. 2a. Comparison of 0600 GMT 29 June 1979 Amini Devi sounding (open circles, solid line) with 0000 GMT 1 July sounding one used in SL were taken at 0600 and 1200 GMT 1 from SL (triangles, dashed line). There was no offshore convection July; they show a considerable moistening of the 850– on 29 June 1979. 700 mb layer, but may still not be favorable for con |
When reviewing articles in a peer-reviewed publication, the reader is assured that the articles are
Related Posts
Copyright © 2024 ShotOnMac Inc.
