You are on page 1of 3

HISTORICAL NOTES

Description of rainfall variability in Brhat -samhita of Varâha-mihira

R. N. Iyengar

Brhat -samhita of Varâha-mihira (5–6th century AD) provides valuable information on the approach in an-
cient India towards monsoon rainfall, including its measurement and forecasting. In this context, we come
across a description of the expected amount of total seasonal rainfall depending on the first rains under the
27 naksatras of Indian astronomy. This provides a rough statistical picture of what might have been the
rainfall and its variability in the region around Ujjain, where Varâha-mihira lived. The coefficient of varia-
tion of the model, described by him, is 37%. This value is close to the present-day climatic variability of sta-
tion-level monsoon rainfall in and around Ujjain, Madhya Pradesh.

Monsoon variability may be defined as Artha-ú âstra Brhat -samhita


the tendency of seasonal rainfall to fluc-
tuate about its long-term climatic normal
value. When year-wise sample observa- A review on measurement of rainfall in There are several editions available for
tions are available, one can find the nor- ancient India has been previously written this ancient Sanskrit classic, with com-
mal value as the time average of the data by Srinivasan1 and hence will not be re- mentaries and translations in various
series. The standard deviation of the peated here. The earliest reference to rain languages. Here, the text edited by
sample provides a simple characteriza- gauges and regional distribution of rain- Ramakrishna Bhat3 is used for further
tion of the variability. It is the usual fall is contained in the Artha-ú âstra2, a reference. BS devotes eight chapters to
practice to represent this deviation, as a treatise on statecraft authored by Kautilya discuss rainfall, including measurement
percentage of the normal or mean value, in the 4th century BC. Kautilya states that and forecasting. In chapter 23, a state-
and refer to it as the coefficient of varia- rainfall over forest districts was 16 ment on the amount of rainfall to be
tion. For example, the all-India (south- Drona (D) and over Avanti it was 23 D. forecast for the season, depending on the
west) monsoon rainfall has a mean value Since the above rainfall values are men- first rains in the month of Jyestha is
of 85.24 cm with a standard deviation of tioned in an administrative manual, these given. The text reads,
8.47 cm. This gives the coefficient of might have been some kind of averages
variation over a period of 130 years observed over a length of time. The an- Hastâpya-saumya-citrâ-pausna -dhanis -
(1871–1990) to be nearly 10%. This cient connotation of Avanti stands for the th âsu sodaú a dronâh |
would be higher for a smaller region region of that name with its capital at Œatabhisag -aindra-svâtisu catvârah
such as a district or a State. It would be Avanti City, generally identified with krttik âsu daú a ||
of interest to know whether this variabil- Ujjain (23°11′N 75°47′E). The extent of Œravane magh â’nurâdhâ-bharan î-mûlesu
ity remains constant over time in a statis- this region is not precisely known, ex- daú a caturyuktâh |
tical sense. If this parameter were to be cept that it overlaps with the present-day Phalgunyâm pañcakrtih punarvasau
stable, it would hint that floods and subdivision no. 19 of the India Meteoro- vimœatirdron âh ||
droughts have been part of this normal logical Department (IMD). From avail- Aindrâgnyâkhye vaiú ve ca vimú atih
climatic variability. There is consider- able data for 100 years (AD 1901–2000), sârpabhe daœa tryadhikâh |
able anecdotal description in ancient lit- the mean value of the June–September Âhirbudhnyâryamna -prâjâpatyesu
erature, to believe that the monsoon monsoon seasonal rainfall of this subdi- pañcakrtih ||
phenomenon was well known in India vision is 86 cm. If the months of October Pañcadaú âje pusye ca kîrtitâ vâjibhe
since early Vedic period (c. 4000–3000 and November are also included, the six- daœa dvau ca |
BC). However, on the quantitative side, month average increases marginally to Raudre ast âdaú a kathitâ dronâ nir u-
descriptions are meagre and sketchy. In- 91 cm. The modern values cannot be padravesvete || (ch.23.6-9)
formation is available on how rainfall compared with the figures of Kautilya,
was measured, but the outcomes of the since he mentions his gauge to be a bowl The rainfall quantified in terms of
measurement are missing, except for with its mouth being about 40 cm (in di- Drona, stated in the verses is depicted in
some brief numbers. Interestingly enough, ameter), with no information on the column 2 of Table 1. The naksatra s are
what little is available has reference to height. Thus, instead of comparing abso- proxies for the position of the Moon.
variability of rainfall in an obscure statis- lute values of rainfall, it would be This is an ancient Indian method of cal-
tical fashion. With this in view, this note worthwhile to compare the past endar reckoning, which is still popular.
presents a brief review of available in- with the present, in terms of a dimen- VM is categorical that rainfall measure-
formation followed by an interpretation sionless parameter. It is in this context ment should start after the full moon in
of a set of statements appearing in the that the set of 27 rainfall quantities men- the month of Jyestha (June–July). Thus,
Brhat -samhita (BS) of Varâha-mihira tioned by VM in his BS acquires signifi- he appears to have been particular about
(VM) from a modern perspective. cance. the onset of monsoon, which he has

CURRENT SCIENCE, VOL. 87, NO. 4, 25 AUGUST 2004 531


HISTORICAL NOTES
placed after the Jyestha full moon, quantity makes one wonder whether it is values. For this purpose, it is expedient
spread over the next 27 days. Even possible to convert the values stated in to represent time-series data in the stan-
though VM starts chapter 23 of BS with a BS into modern-day equivalents. There dard form y = (R – m)/σ to visualize fluc-
qualitative prognosis based on first rains are many difficulties in converting tuations about the mean and beyond the
in the asterism of Pûrvâsâdhâ , he states Drona to metric units. Srinivasan1 states one-sigma level. For Ujjain town, the
the expected (or forecast) seasonal rain- that a Drona is equal to about 5.1 cm of modern rainfall data are available for
fall in terms of the values given in Table rainfall. But, Balkundi4, a meteorologist, short lengths only. However, for Indore
1. Obviously, he was aware that the first has found this conversion factor to be near Ujjain, reliable time-series data are
rain could happen under any of the 27 6.4 cm. In Table 1, rainfall figures of BS available. These two stations have com-
stars. Hence, the above may be taken as are converted with 1D = 6.4 cm and parable coefficients of variability, as
an ancient way of describing what is shown for easy reference. Average rain- seen from Table 2. With this in view, sta-
presently known as variability of rainfall. fall, according to this ancient text, will tion data of Indore are standardized and
The numbers themselves would have be 99.78 cm with a standard deviation of presented for the period 1901–2002 in
come from observations, which were 36.67 cm. In Table 2, a few current aver- Figure 1. For the ancient model of VM, a
definitely in vogue as understood from age and standard deviation values for time-series has been artificially simu-
the details given by Kautilya and other Ujjain and nearby stations are presented. lated according to the probability distri-
ancient writers on the subject. Keeping in view the uncertainty involved bution of Table 1. This is also shown in
Since there is no connection between in converting the ancient Drona measure Figure 1. This simulated time series has
the first rains and the seasonal total (ex- into modern figures, the present-day been based on a sequence of independ-
cept to the extent the former is included rainfall values seem to be broadly in the ent, uniformly distributed random num-
in the latter), the rainfall given in Drona same range as the figures in BS. During bers from 1 to 27. This is done to see
should be interpreted as the climatic ancient times, some parts of present Ra- how the two data compare, about their
normal value, in a statistical sense. The jasthan and Gujarat were included in the mean levels, when equal sample lengths
day on which the rain starts is a random geopolitical region of Avanti/Ujjain. are considered. Evidently, the simulated
variable and over a long period, this Quite clearly, the coefficient of variation sample has no specific starting year and
would be under any one of the 27 increases as one proceeds from Ujjain is devoid of any natural inter-annual
naksatra with equal probability. Thus, if towards Jaipur in Rajasthan. Thus, the variability pattern. Nevertheless, visual
the rainfall amounts mentioned are model of VM is consistent with the pre- comparison of the two time series high-
meaningful, they may represent values sent-day understanding of rainfall distri- lights the similarity of the variations
observed with a probability of 1/27 for bution over the target region. therein. This can be better seen by con-
each naksatra . For example, in modern structing the relative frequency diagram
notation, probability of rainfall equal to Time-series and probability or the probability density function for the
10D will be the same as probability of distribution two series on the same scale, as in Figure
first rains in Krttik â, which will be 1/27. 2. It is observed that in the central parts
Similarly, The mean (m) and standard deviation (σ) of the distribution, the probabilities are
are not efficient in reflecting extreme comparable. Towards the right tail, indi-
Probability (R = 15) = Probability
(first rains in Pusy a) + Probability
Table 1. Rainfall variability model of VM
(first rains in Pûrvâbhâdra)
= 1/27 + 1/27 = 2/27. Serial number of 27 naksatra Rainfall in Rainfall in cm Probability of
Probability (R = 4) = 3/27; starting with Kr
ttik â Drona (1D = 6.4 cm) occurrence
Probability (R = 25) = 4/27.
1 10 64 1/27
Rainfall in cm and such associated prob- 2, 9, 10, 24 25 160 4/27
abilities are listed in the last two columns 3, 11, 12, 18, 21, 25 16 102.4 6/27
of Table 1. This provides a rough statis- 4 18 115.2 1/27
tical picture of what might have been the 5, 14, 19 20 128 3/27
climatic variability of rainfall in the re- 6, 23 15 96 2/27
7 13 83.2 1/27
gion surrounding Ujjain, where VM lived
8, 15, 17, 20, 27 14 89.6 5/27
during 5–6 century AD. Rainfall with the 13, 16, 22 4 25.6 3/27
above discrete probability distribution 26 12 76.8 1/27
has a mean value (m) of 15.59 D and
standard deviation (σ) of 5.73 D. This
gives the dimensionless coefficient of Table 2. Monsoon seasonal rainfall over central Indian stations (1901–1980)
variability (σ/m) as 37%. This value is Standard Coefficient of
close to the present-day variability fig- Station N E Mean (cm) deviation (cm) variability (%)
ures in the western part of Madhya
Pradesh, including Ujjain (Table 2). Ujjain 23°11′ 75°47′ 82.57 28.13 34
So far, no units have been invoked in Indore 22°43′ 75°48′ 86.45 26.22 31
discussing variability, which is the ratio Guna 24°39′ 77°19′ 97.83 32.27 33
of the standard deviation to the mean Dhar 22°36′ 75°18′ 85.74 25.92 30
Jaipur 26°49′ 75°48′ 55.57 22.61 42
value. However, the close match of this

532 CURRENT SCIENCE, VOL. 87, NO. 4, 25 AUGUST 2004


HISTORICAL NOTES

x = (RF–m)/σ

Figure 1. Standardized actual (IMD) and simulated (VM) rainfall time-series at Indore.

Figure 2. Comparison of probability density functions at Indore. a, IMD (actual); b, VM (simulated).

cating floods, the model of VM appears erable yearly variation. The list of 27 2. Shama Sastry, R. (ed. and translator) Artha-
curtailed or saturated. The reason for this expected rainfall values, based on the ú âstra of Kautilya , Mysore, 1988, 9th edn.
is traceable to the 27 number of states occurrence of first rains, as stated in the 3. Bhat, M. R., Brhatsamhita of Var âhami-
used by him. Had he used another calen- BS is amenable for statistical investiga- hira (text with translation), M. Banarsi-
dass, New Delhi, 1981.
dar, he would have come out with 30 or tion. Since, the conversion of Drona
4. Balkundi, H. V., Commentary in ‘Krsi -
more states resulting in a more spread- measure to present-day linear measure is Parâú ara’, translated by Sadhale, N., Agri-
out distribution. However, on the left not conclusively established, it is not History Bulletin No. 2. Publ. Asian Agri-
side of the distribution, the two prob- possible to directly compare the present- History Foundation, Secunderabad, 1999.
abilities again appear to be comparable. day average rainfall with the ancient
Conclusions that are more specific are value. However, the non-dimensional ACKNOWLEDGEMENTS. Thanks are due
not possible. variability, defined as the ratio of stan- to Dr B. P. Radhakrishna for useful discus-
dard deviation to climatic mean, of rain- sions and Dr Rajeevan Nair, IMD Pune for
Conclusion fall in the central part of India has providing rainfall data of Ujjain and Indore.
perhaps remained stable over a long Sri S. T. G. Raghukanth helped with the com-
Ancient texts such as Artha-ú âstra and period. putations.
BS written before 6th century AD pre-
serve vague, but definitively quantitative R. N. Iyengar is in the Centre for Atmos-
information on the amount of monsoon 1. Srinivasan, T. M., Measurement of rainfall pheric and Oceanic Sciences, Indian
rainfall. It appears that VM had recog- in ancient India. Indian J. Hist. Sci., 1976, Institute of Science, Bangalore 560 012,
nized that monsoon rainfall had consid- 11, 148–157. India. e-mail: rni@civil.iisc.ernet.in

CURRENT SCIENCE, VOL. 87, NO. 4, 25 AUGUST 2004 533