HYDROLOGICAL PROCESSES, VOL. 5,243-250 (1991)

111: METHODS OF CORRECTING FOR SYSTEMATIC ERRORS IN ATMOSPHERIC PRECIPITATION MEASUREMENTS IN

CZECHOSLOVAKIA

M. LAPIN AND F. SAMAJ Slovak Hydrometeorological Institute, Bratislava, CSFR

ABSTRACT

The methods of correcting for systematic errors in precipitation measurements using the Czechoslovak gauge METRA 886 are presented. This gauge has an orifice area of 500 cm2 and is elevated 1 m above the ground. The wetting correction amounts to 0.1-0.2 mm per measurement. The evaporation correction ranges from 3 per cent in frost-free periods to 10 per cent in frost periods. The wind-induced correction amounts to 5 per cent for rain and 45 per cent for snow. The total sum of corrections on average 10-15 per cent per year in lower localities.

KEY WORDS Precipitation measurement Error correction

INTRODUCTION

In the period 1971 - 1985, a stage in research into the methods of atmospheric precipitation measurements and into systematic errors in precipitation measurements was completed by the Hydrometeorological Service of Czechoslovakia (CSFR) within the framework of the state research task. This research was aimed mainly at the Czechoslovak standard precipitation gauge METRA 886, which differes significantly in its parameters from the precipitation gauges most frequently used elsewhere in the world. In connection with the research described here, experimental areas for an extensive programme of intercomparison measurements were built at the meteorological observatories in Bratislava and Ostrava. Intercomparison precipitation measurements were taken to a lesser extent at six other meteorological stations. A limited measurement programme has continued since 1985.

The aim of this research was mainly the analysis of three systematic errors of precipitation measurement (wetting, evaporation, and wind effect), comparison of the results with errors of other national precipitation gauges, and, on the basis of these, the development of correction methods for monthly precipitation totals for all precipitation stations in the CSFR. In this paper, we describe briefly the methods of systematic error corrections for precipitation that are

accepted in the CSFR (Lapin and Priadka, 1987; Lapin et al., 1985; Samaj, 1973), describe the methods of computer technique calculations and the database system METEOSYS (IDMS), as well as the evaluation of corrected precipitation totals in the territory of Slovakia for the period 1981-1988 (Lapin el al., 1990).

CHARACTERISTICS OF THE METRA 886 PRECIPITATION GAUGE AND THE PRECIPITATION MEASUREMENT SYSTEM IN THE CSFR

The METRA precipitation gauge has been used in the CSFR without modification since 1946; before then, the most frequently used precipitation gauge had comparable construction parameters. The METRA precipitation gauge has an orifice on 0.05 m2 (500cm2) with a height of 1 m above the ground. It is manufactured from galvanized plate with a hardened top edge and is painted outside in either white or a dull silver colour. The standard version is used without a wind shield. Specific versions of this precipitation gauge

0885 -6087/9 1/030243 -08W5.00 0 1991 by John Wiley & Sons, Ltd.

244 INTERNATIONAL WORKSHOP ON PRECIPITATION MEASUREMENT

are used in the cold and warm periods of the year, which significantly influences mainly the first two systematic errors of measurements (wetting and evaporation). The winter version of this precipitation gauge consists of a large vessel only, while the summer version is equipped with a funnel that drains the failing liquid precipitation into a small vessel (Figure 1).

In the CSFR, the standard method of measuring precipitation totals is once daily at 07.00 hr, mean local time. The winter version of the precipitation gauge is used in the lowlands from 1 November to 31 March, and all the year round at altitudes above 2000 m. Daily precipitation totals, daily snow cover, data about the kind and form of precipitation are simultaneously, in monthly cycles, checked and stored on computer media in a database (Lapin el al., 1990).

DESCRIPTION OF BASIC SYSTEMATIC ERROR CORRECTIONS

The corrections of the METRA precipitation gauge systematic errors due to wetting of the gauge and the calibrated vessel (K,) were derived from a great number of simulated measurements of various precipitation totals under various conditions (Lapin et al., 1985). It was found that the distribution of these errors is in the range 0-06 mm to 0.30 mm per one measurement, depending on the version and age of the precipitation gauge. Because of simplification, on the basis of the statistical processing of these measurements, only two corrections were determined: 1. 0.2 mm per one measurement using the winter version of the METRA precipitation gauge for measured

precipitation totals of 1 mm or higher. 2. 0.1 mm per one measurement in all other cases.

For the determination of evaporation errors in precipitation meaurements, there were also extensive experiments with simulated precipitation during days with non-precipitation weather (Lapin et al., 1985). It was found that, using the winter version of the precipitation gauge, the evaporating surface of water in the precipitation gauge is approximately 10 times greater than that using the summer version. Errors as a result of evaporation in winter are comparable with high summer; the greatest errors are in the warm weather at the end of the period of use of the winter version of the precipitation gauge, that is in March and April. The correlation dependencies between evaporation from the precipitation gauge and other meteorological elements were also analysed. In spite of the fact that the closest relationships were found between evaporation and global radiation for the winter version of the precipitation gauge and between evaporation and saturation deficit for the summer version, for practical reasons regression equations of dependence on mean air temperature for a time interval between the start of precipitation event measurement time at 07.00 hr were accepted for the determination of error corrections for evaporation from precipitation gauge KE. Graphs of derived regression equations are shown in Figure 2. It can be seen that curves for correction calculations are very close for various totals of measured precipitation. Only for low precipitation with the winter version of the precipitation gauge are they significantly differentiated. With the winter version, the errors due to

A 8 C Figure I . Rain gauge METRA 886 (CSFR) without base. (A-Winter version, B-Summer version, C -Calibrated vessel)

Ill: CORRECTING PRECIPITATION ERRORS

0,OS-

245

/3

0

Figure 2. Dependence of mean hourly evaporation, A R E , from the METRA 886 rain gauge on air temperature. (Winter version of the gauge-solid lines; summer version of the rain gauge-dashed lines. 1 -for daily precipitation total < 1 mm, 2-for 1-49 mm. 3-for > 5 mm)

evaporation are 4 to 5 times greater than with the summer version of the precipitation gauge at the same air temperature.

Deriving correction for the precipitation measurement errors due to the aerodynamic effect (wind effect) K, appeared to be the most complicated. The reasons for this are analysed in Lapin et al., 1985. The basic problem was that we were not successful in finding a suitable area for the installation of double fences as wind shield to eliminate the effect of wind on falling precipitation. Instead of this, the METRA precipitation gauge was used, with a modified Tretyakov wind shield (Figure 3). It was found that this shield reliably eliminates the effect of wind only to a wind velocity 3 m s- at a height of 2 m above ground, for solid precipitation. At higher wind velocities, the efficiency of this wind shield gradually decreases. For mixed and liquid precipitation, this shield is reliable also at greater wind velocities. Besides this, intercomparison measure- ments were made using the Soviet precipitation gauge 0- 1, ground level precipitation gauges, and precipitation gauges with other types of wind shield. On the basis of the results of these measurements, regression curves were derived of dependencies of correction coefficient K, on wind velocity at a height of 2 m above ground (Figure 4). For solid precipitation, a curve for wind velocities greater than 3 m s - ' was extrapolated in the form of a quadratic function up to the value K, = 2.00, i.e. to a wind velocity of approximately 5 m s - ' . Above this wind velocity, we recommend the use of K, = 2.00 (Lapin et al., 1990) for solid precipitation. In spite of this, this correction coefficient cannot be used for windy exposed mountain areas.

CORRECTION DETERMINATION O F PRECIPITATION TOTALS MEASURED IN A NETWORK IN SLOVAKIA

Methods of precipitation total corrections came out mainly from the extensive intercomparison measure- ments in the experimental polygon in Bratislava-Koliba. This locality is characterized by significantly windy weather with relatively high snow precipitation in winter. Measurements at the other stations were used for verification or for defining the obtained results more precisely.

In the application of corrections in a network, it is necessary to take into account mainly the great complexity of natural conditions in Slovakia and their effect on systematic measurement errors. Likewise, it is an important fact that only for some of the stations are there available all the input data needed for the calculation of the corrections.

In Slovakia, in a territory of about 50000 km2, there are approximately 700 precipitation stations, of which 105 also measure other meteorological elements including air temperature and wind velocity. Approximately

246 INTERNATIONAL WORKSHOP ON PRECIPITATION MEASUREMENT

Figure 3. The Czechoslovak precipitation gauge METRA 886 with the Tretyakov wind shield (orifice 500 cm2, height above the ground 1 m). It is used on a limited number ofstations, simultaneously with the precipitation gauge METRA 886 without a wind shield, which is the Czechoslovak standard precipitation gauge (photo: Z. Kmentova)

37 per cent of the territory lies above 500 m in altitude, where 32 per cent of the precipitation stations are situated (Lapin, 1990).

In the METEOSYS database system, the Slovak Hydrometeorological Institute (SMHI) has stored daily and seasonal data of meteorological measurements from about 100 stations per year since 1961. Since 1981, there have also been stored daily data from 700 precipitation stations. For calculations of corrected monthly precipitation totals, data from the database was used exclusively and the whole data processing was done automatically on an EC 1055 computer (Lapin er al., 1990). Because of some problems with standardization and with the complexity of the measurements of some of the required input data, it was necessary to simplify some calculations.

For the calculation of wetting correction K , it is necessary to know the date of the change from the winter version of the precipitation gauge to the summer one, and vice versa. In spite of the fact that, with continued warm weather, the observers prolonged the period of use of the summer (more precise) version of the precipitation gauge, we accepted for calculation purposes the following dates as the most probable for the changeover: in localities up to 300 m above sea level-1 April and 31 October; to 500 m above sea level- 16 April and 15 October; to lo00 m above sea level- 1 May and 30 September; to 2000 m- 16 May and 15 September; and above 2000 m, the winter version is used all year round.

For calculation of the evaporation correction KB it is necessary to know the duration of the precipitation from a fixed start and the air temperature during this time. Because these time data have not been stored in precise form in the database, on the basis of calculation from a smaller number of observations we made a simplified assumption that, in the warm period of the year, precipitation starts to fall on average at 17.00 hr and in the cold period of the year a t 15.00 hr, local mean time. Mean air temperature can be calculated in a

111: CORRECTING PRECIPITATION ERRORS 247

Figure 4. Dependence of correction coefficient K, on wind velocity u measured at a height of 2 m for solid (S), mixed (M), and liquid (L) precipitation

simplified form from weighted mean measurements taken at 14.00,21.00, and 07.00 hr, while the temperature is taken twice at 21.00 hr. Because of this fact, corrected precipitation is used for the shortest period of 1 month; this simplification cannot effect significantly the accuracy of the results.

For calculation of wind effect correction K, it is necessary to know the mean wind velocity during the precipitation periods at a height of 2 m above ground. In the database are stored the mean wind velocities at 105 stations from measurements taken at 07.00,14.00, and 21.00 hr, but anemograph sensors are mostly at a height of 10 m and very often they are relatively far from the precipitation gauges. For the calculation of wind velocities at a height of 2 m, we used logarithmic formulae of wind profile (Lapin et al., 1990), while for individual stations there were determined altogether six various parameters of ground level roughness. The existence of the greater snow cover depth of 5 cm was also taken into account.

Correction calculations were made on the computer in four steps. In the first step, daily correction errors for wetting for all stations (approximately 700) were calculated. For each station an extract of monthly and annual sums of measured precipitation R, and corrected precipitation sums R, was made for individual years.

Correction coefficient K,,, = R J R ,

At the same time, for approximately 100 stations with air temperature and wind velocity measurements, calculations of corrections for evaporation and wind effect were made in a similar way.

and

248 INTERNATIONAL WORKSHOP ON PREClPlTATlON MEASUREMENT

Table 1. An example of calculated correction coefficients K,, KE, K,, K, and corrected precipitation totals R, for station Bratislava airport for 1987. Calculations were made on the basis of daily values of measured meteorological elements by computer EC1055 and by database system METEOSYS according to SHMI methods; R, represents measured monthly precipitation totals in mm

R I I1 111

R , 92.8 67.7 31.1 K,, 1.040 1.037 1.084 K , 1-031 1663 1.115 K, 1.253 1-061 1-178 K 1.344 1.170 1.423 R, 124.7 79.2 44.3

1v 13.8 1.116 1.151 1 a45 1.343

18.5

V

84.5 1.028 1 048 1.033 1.113

94.0

Months v1 VII

59.9 47.8 1.027 1.025 1-053 1-060 1.030 1.030 1.114 1.120

66.7 53.5

VlII

45.1 1.040 1.082 1-027 1.155

52.1

IX X XI XI1 Year

15.1 18.7 44.9 76.8 598.2 1066 1 0 4 3 1.067 1.031 1.041 1.124 1.070 1.162 1.065 1.071 1.027 1.029 1.033 1.033 1.076 1.230 1.148 1.280 1.134 1.200

18.6 21.5 57.5 87.1 717.7

The final correction coefficient is:

All monthly and annual values of these coefficients and precipitation totals are issued in the form illustrated in Table I.

In the second step, statistical evaluations of coefficients K,,,, Km K , and K were made for individual months and years and for degrees of altitude and locality according to wind velocity. An example of output for coefficient K for the whole of Slovakia is given in Table 11. From this table it is possible to find out the distribution and frequency of coefficient occurrence depending on altitude, to find some errors in correction calculations. In most cases it is evidently possible to approximate the dependence of correction coefficients on altitude by a straight line.

In the third step, coefficient calculations are made of regression equations of the dependence of KE, K , and K on altitude for individual months. The general form of the equation is

Ki = a,, + a,h + afi2

where Ki is the correct coefficient, h is the altitude, and am a,, and a2 are coefficients of the regression equation. Table I11 gives example coefficients of regression equations and calculated values of correction coefficient K for altitudes graded by 100 m for weak and strong wind localities on Slovakia in 1987.

In the fourth step, calculations of Kh, K, and K are made, for individual precipitation stations for which there are no wind velocity of air temperature measurements, based on regression equations and the altitudes of the stations for individual months. For the practical utilization of corrected monthly precipitation totals, Lomputer printouts for all stations in Tables I and 11 have been stored. At the same time there are tables of the calculations KE, K,, and K for altitudes graded by 100 m. Correction coefficients K for individual stations and for individual months are stored on computer media in the database.

CONCLUSIONS

Methods of correcting for systematic errors in monthly precipitation totals and the results of the application of these methods to a network in Slovakia demonstrate that there is a series of problems in this sphere that is difficult to get over for objective reasons. The most serious problem is the fact that the methods were derived only on the basis of a limited number of intercomparison measurements of solid precipitation. We were not successful in the realization of representative measurements using a perfect wind shield, although the modified Tretyakov seems to be very effective, especially at wind velocities up to 3 m s - * . Another serious problem is the insufficient basis of input data for the correction calculations for precipitation stations in the network. By the application of methods, for this reason, there was used a series of effective simplifications which would not substantially affect the correction of monthly precipitation totals.

Ill: CORRECTING PRECIPITATION ERRORS 249

Table 11. Absolute frequencies of correction coefficients K, (weak windy meteorological stations) and K, (mild and strong windy stations) for annual sums of measured precipitation in Slovakia in 1987 in distribution according to grades of altitude

Altitude in m to 201 301 401 501 601 701 801 901 1001 1201 1401 above 200 300 400 500 600 700 800 900 lo00 1200 1400 1600 1600

Kl < 1.00-1.02) < 1.02- 1.04) < 1-04-1.06) < 1.06- 1.08) < 1.08-1.10) < 1.10-1.12) < 1.12-1.14) < 1.14-1.16) < 1.16-1.18) < 1.18-1.20)

2 1.20

K2 < 1.00- 1.02) < 1-02-1.04) < 1.04- 1.06) < 1.06- 1.08) < 1.08-1.10) < 1.10-1.12) < 1.12-1.14) < 1.14-1.16) < 1.16-1.18) < 1.18-1.20)

2 1.20

Table 111. Values of coefficients of regression equations (urn u,, 0,) for calculation of correction coefficients of precipitation K , and K2 on the territory of Slovakia and calculated values K, and K, for altitudes from 100 to loo0 m for selected months and year 1987. (Coefficients K, and K , are used for simplified calculation of corrections of monthly precipitation totals)

Months a, a, x lo3

K , February April June August October December Year

1.225 0.4131 0.967 0.7323 1.105 - 0.0690 I * 120 - 0.0596 1.071 0.3961 1.129 0.2616 1.134 -0.0023

u2 x 106

-0.3635 -0.5473 0.0496 0.0192

- 0.2338 -0-21 53 - 0.0077

Altitude in m 100 200 300 400 500 600 700

1.26 1.29 1.32 1.33 1.34 1.34 1.34 1-04 1.09 1.14 1.17 1.20 1.21 1.21 1.10 1.09 1.09 1.09 1.08 1.08 1.08 1-12 1.11 1.10 1.10 1.10 1.09 1.09 1.11 1.14 1-17 1.19 1.21 1-23 1.23 1-15 1.17 1.19 1.20 1-21 1-21 1-21 1.13 1.13 1.13 1-13 1.13 1.13 1.13

800

1.32 1.20 1.08 1 *09 1.24 1.20 1.13

900 lo00

1.30 1.27 1.18 1.15 1.08 1.09 1.08 1.08 1-24 1.23 1.19 1.18 1.13 1.12

K2 February 1.289 0.3855 -0.0650 1.33 1.36 1.40 1.43 1.47 1.50 1.53 1.56 1.58 1.61 April 1.141 0.0577 0.0758 1.15 1.16 1.17 1.18 1.19 1.20 1.22 1.24 1.26 1.28 June 1.109 -0.0560 0.0456 1.10 1.10 1.10 1.09 1.09 1.09 1.09 1.09 1.10 1.10 August 1.139 -0.0844 0.0554 1-13 1.13 1.12 1.11 1.11 1.11 1.11 1.11 1.11 1.11

December 1.159 0.1254 0.0781 1.17 1.19 1.20 1.22 1.24 1.26 1.29 1.31 1.34 1.36 Year 1.181 -0.0699 0.0983 1.18 1.17 1.17 1.17 1-17 1.18 1.18 1.19 1.20 1.21

October 1.143 0.2124 -0.0632 1.16 1.18 1.20 1.22 1.23 1.25 1.26 1.27 1.28 1.29

250 INTERNATIONAL WORKSHOP ON PRECIPITATION MEASUREMENT

Obtained results of corrected systematic precipitation errors are compared with parallel precipitation measurements by precipitation gauges with wind shields and with measurements by totalizers. It is satisfying that precipitation totals measured by totalizers and the corrected precipitation totals are very similar in mountain localities (FaSko and Lapin, 1988; Lapin, 1990). The first applications of corrected precipitation totals are very encouraging for the aims of analysis of water balance in catchments. Precipitation totals corrected for systematic errors are used in the practical operations of the SHMI for they are provided to external users, mainly from the departments of water management and they are used in the solution of research tasks. Using corrected precipitation totals presents the necessity of re-evaluation of calculation methods of determination of areal evaporation and runoff in some catchments.

REFERENCES

FaSko, P. and Lapin, M., 1988. ‘Meranie atmosferickych zri2ok pomocou totalidtorov na uzem; Slovenska za obdobie 1951-1985 (Atmospheric precipitation measurements by totalizers on the territory of Slovakia for the period 1951 - 1985)’. Meteorologicki zprhuy,

Forland, E. and Aune, 9. 1986. ‘Comparison of Nordic methods for point precipitation correction’, in Sevruk, 9. (Ed.), Corrections of Precipitation Measurements. Ziiricher geographische Schrifren, ETH, 23,239-244.

Lapin, M. 1990. Problematics of Measurements and Elaboration of Atmospheric Precipitafion in Mountainous Areas of Slovakia, in Molnar, L. (Ed.) Hydrology of Mountainous Areas, IAHS Publication No. 190.47-55.

Lapin, M. and Priadka, 0. 1987. ‘Korekcie systematickych chyb merani atmosferickych zraiok (Corrections of systematic errors of atmospheric precipitation measurements)’, MeteorologickP zprauy, 40, 1, 9-19.

Lapin, M., FaSko, P., and KoSt’alova, J. 1990. ‘Zhodnotenie zr2kovych pomerov na kemi Slovenska PO korekcii systematickych chyb merani zratok (Evaluation of precipitation conditions in the territory of Slovakia after correction for systematic errors of precipitation measurements)’, Meteorologicke zpravy, 43, 4, 101- 105.

Lapin, M., Lednicki. V., and Priadka, 0. 1985. Upresnenie systematickjch chjb feskoslovenskjch Standardnjch zr&fkomerou METRA 886 (Zcluerefnu sprava ujskumnei ulohy) (Making accurate the systematic errors of the Czechoslovak standard precipitation gauge METRA 886), Bratislava, SHMU.

Sevruk, 9. and Hamon, W. R. 1984. ‘International comparison of national precipitation gauges with a reference pit gauge’, Instrument and Observing Methods Report No. 17, Geneva, WMO, 86 pp.

Struzer, L. R. and Golubev, V. G. 1979. ‘Methods for the Correction of the Measured Sums of Precipitation for Water Balance Computations’. Report for the UNESCO/WMO International Workshop on Water Balance of Europe, Leningrad, 23 pp.

Samaj. F. and Lapin, M. 1986. ‘Results from the study of methods for measuring precipitation and systematic errors of rain-gauges in Czechoslovakia’, in Sevruk. 9. (Ed.), Corrections of Precipitation Measurements, Zurich, ETH. Ziircher Geographische Schriften, 23,

Samaj, F. 1973. Systematicke chyby merania zraiok. Vplyv klimatickjch prvkou nu hydrologicke procesy, (Systematic errors of

41, 5 , 136-140.

176-185.

precipitation measurements. Effect of climatic elements on hydrological processes), Prace a Studie, Bratislava, HMU, 83- 105.