SR/SSD 97-45

10-15-97

Technical Attachment

MEAN AREAL PRECIPITATION CLIMATOLOGY

FOR TWO LOWER MISSISSIPPI RIVER BASINS

Keith Stellman(1) and John Kuhn

LMRFC Slidell, Louisiana

The National Weather Service (NWS) forecast offices in the Southern Region have been providing the Lower Mississippi River Forecast Center (LMRFC) with quantitative precipitation forecasts

(QPF) of Mean Areal Precipitation (MAP) instead of point precipitation. Since point precipitation has been used by the forecast offices over the years, there was a need to provide forecasters with some sort of climatology on MAPs. So the LMRFC embarked on a program to compute MAP basin climatology.

Using techniques available in the NWS River Forecast System (NWSRFS), point precipitation observations were collected, quality controlled, and processed to compute MAP over a period of 39 years for selected basins in southeast Louisiana and Mississippi. Statistics and graphics were generated for each river basin to show frequency distributions and maxima of 6-hr and 24-hr MAP. In the following sections we discuss how the data were collected and processed, and present general results and conclusions of the MAP study.

Since there is a large volume of precipitation data for the LMRFC area, and computer resources at the LMRFC do not allow for storage of these data easily, two basins in southeast Louisiana and Mississippi were selected for the study (Fig. 1). The larger basin, Pearl River basin at Pearl River (PERL1), is 931 mi2 in area. The smaller basin is the Amite River basin at Denham Springs (DENL1), which is 112 mi2. The dots in Fig. 1 show the precipitation stations used in the MAP computation.

Data were collected from two different sources: the Office of Hydrology (OH) Web site htpp://nhds2.ssmc.noaa.gov/loclsoft.html, and CD-ROMs provided by the National Climatic Data Center (NCDC). Programs developed by OH were used to synthesize and process the precipitation data available in the specific river basins. The first of these programs, stainv, provides an inventory of stations for a given period of record (1955-1993), locations, and amount of missing data, as well as a list and history of the stations with a plot of station locations. All stations near the specific basins with a minimum of ten years of data were selected. In some cases adjoining stations with partial records were combined to provide a complete record.

Once the selection of the stations was completed, the dlytran and hlytran programs from OH and the NCDC CD-ROMs were used to download daily and hourly precipitation data, respectively, for use in the processing program (MAP3).

MAP3 was used to compute MAP from point precipitation data. The MAP3 program also completed consistency checks for erroneous data. Consistency plots compared monthly precipitation from a single station against monthly precipitation from other stations to show any bias in the data. A precipitation adjustment parameter was used to correct for bias. The general input for MAP3 includes the period of record, weighting options, data output options, and station information such as observation time, time changes, and latitude/longitude of the stations. Basin information included the basin size, name, and basin outline in latitude/longitude points.

After the data had been corrected, the program generated 6-hourly MAPs for the specified basins. A local script was used to put the MAP output into the Informix database (Informix is a commercially available relational database). In the last step, we generated statistics using a spreadsheet/script software package called WINGZ which can access the database and put the MAPs into a spreadsheet.

The statistics that were generated included frequencies/number of occurrences by month, value, and total number of events. Chances of exceeding a value, and the top ten values over the period of record for a 24-hr and 6-hr totals were also calculated. In general, statistics indicate the daily mean areal precipitation is lower in the summer than in the winter. The maximum mean areal precipitation occurred during the spring, with April having the largest number of MAP events greater than 4 in. The statistics also show that during the 38 year period of record, the MAP was less than 1.0 in over 90 percent of the time.

Table 1 shows the 6-hr MAPs computed for the period 1955-1993 for the Pearl River basin at Pearl River. During the 39 year period, 6-hr MAP values exceeded 1.0 in 356 times, or a little over nine occurrences a year, on average. A 2.0 in 6-hr MAP occurred only 33 times, or less than once per year on average; a very rare occurrence. Table 2 shows the five highest 6-hr MAPs for the Pearl River basin at Pearl River for the same 38 year period. The highest MAP values occurred in 1961 and in the early 1980s. (The latter period, incidently, coincided with record El Niño conditions.)

Table 3 shows a statistical analysis of the 24-hr MAPs for the period 1955-1993. During this period there were only 38 occurrences of 3.0 in or greater, which averages about once a year. Of these, only seven were 5.0 in or greater, or about once every five years. An event of this magnitude is not very common. Table 4 indicates the top ten 24-hr MAP values, showing that high values again seem to have occurred in the early 1960s and early 1980s, with only a few such amounts in other years.

Figure 2a compares monthly average point observations and related basin MAPs. The two precipitation stations used for comparison are the New Orleans International Airport and WSO Baton Rouge. These are compared with the Pearl River and Denham Springs basins, respectively. The graphs show similar trends for the stations and basins, even though the size of the basins differ greatly. This is because the precipitation is averaged over one month. There is a peak in precipitation in July and a relatively dry period in October, with only minor fluctuations during the winter and spring.

Figure 2b shows the monthly variation of mean daily MAP computed for 6-hourly periods, including only days when MAP exceeded .01 in, and the mean daily 24-hr MAP. Figure 3 shows the maximum 6-hr and 24-hr MAPs for each month for the Pearl River basin. The two figures clearly show a minimum in the summer months, especially July, and a maximum in the winter and early spring. The fact that point precipitation is higher in July while maximum MAPs are lower indicates that July precipitation includes isolated heavy amounts, but they are not widespread. The upward trend in October reflects fewer, but heavier, precipitation events during that time of the year. From fig. 3 it can also be seen that the max 6 hr MAPs from noon to 6 p.m. shows very little variability in amount during the year, unlike the other three periods. The midnight to 6 a.m. period has the most variability during the year, perhaps reflecting synoptically driven nighttime events.

Figure 4 shows the monthly distribution of 24-hr MAP events greater than 1.0 in for the Pearl River basin. The graph indicates fewer heavier events during the summer months and a peak in late spring. The numbers also show that there has not been an event greater than 3.0 in during the months of June and July for the period of record. Rainfall observations (Fig. 2a) show that July is the wettest month for south Louisiana, but that is not reflected in the frequency of heaviest areal day-to-day averages. The shape of the graph during the summer months can be attributed to convective rainfall from afternoon heating.

Figures 5a and 5b allow comparison of basins of different size for the 6-hr period from noon to 6 p.m. The Denham Springs basin is only 112 mi2, compared to the 931 mi2 Pearl River basin. The size difference is clearly reflected in the MAP values, particularly during the summer months. The larger basin has more low MAP values and fewer high MAP values than the smaller basin. Two factors may explain why the smaller basin has a larger number of MAPs greater than 0.5 inches. Figure 1 shows that observations are closer together in the vicinity of the smaller basin, so there is a greater chance of heavier (convective) rain amounts affecting multiple sites. There is also less area over which to spread the point values in computing the MAPs.

The MAP data and statistics can be very useful to forecasters who issue 6-hr QPFs on a daily basis. By using climatology as a guideline, the forecaster will be able to see the probability of exceeding

a particular MAP value and the return period of a MAP for a basin. When the forecaster has a low confidence in the QPF forecast, statistics from the MAP climatology could be used to verify statistically the QPF amounts for that month or season.

However, the data and statistics should be used with caution. One limiting factor is that the most complete data which can be used to compute the MAPs span only a 39 year period. This factor came into play recently when MAPs greater than 3 in over a large area were observed in June for the first time. Another limiting factor is that the stations with a 39 year period of record are more sparsely distributed than the gauging network we have today. This factor alone can cause MAPs to be biased either high or low because of the weighting factors involved with a dense network and a sparse network. A sparse network would have a higher weight for a station than would be the case for a dense network.

Several stations contained bad and missing data over the 39 year period which the MAP3 program estimated from surrounding stations. This technique may work adequately in winter months when precipitation is distributed more evenly, but it fails during summer months when precipitation is more likely convective in nature and therefore more variable in amount and distribution. Another problem with the MAP3 program is that if a station has missing data, precipitation will be distributed to that station through a weighting problem from surrounding stations. This process can even occur when a site did not receive precipitation. Only with a longer and more complete data record will we be able to generate a more complete MAP climatology for an area.

Even though there are limiting factors involved with generating mean areal precipitation, the climatology should be useful for forecasting QPF. One goal at the LMRFC is to generate MAPs and statistics for the entire LMRFC area, similar to the work done on these two basins. The other goal is to take this one step further and generate a model for the forecaster to use on a daily basis that uses climatology, synoptic and mesoscale patterns, and statistics.

5. Acknowledgments

The authors wish to thank Dave Reed (HIC), Jeff Graschel (Sr. HAS forecaster), and Susan Van Cooten (HAS forecaster) at the LMRFC for their proofreading and editing.

1. NWS co-op student, now attending Florida State University