SPACE TIME CHARACTERISTICS OF RAINFALL IN EAST RAJASTHAN

Abstract

Rainfall in semi arid regions is low and uncertain. For long term planning of water resources development and short term planning for agricultural operations it is essential to understand the mechanisms and behaviour of rainfall process and analyse its space and time characteristics. In India, semi arid regions in the north have a distinct wet season restricted to about 80 to 90 days during the summer monsoon season June to September. Rajas than in northwest India has semi arid climate in the east and arid in the west. The region is characterised by low to moderate rainfall with high degree of variability. The region is not only prone to frequent droughts but occasional heavy rain storms. The mean annual rainfall varies from about 100 cm in the south east to 15 cm in the west and 40 cm in the north. The coefficient of variability of annua] rainfall varies from 35% in the southeast to more than 80 % in the west and 40 % in the north. Literature review has indicated that no systematic studies have been carried out for this region to understand the behaviour of rainfall occurrence in space and time. The review has further indicated the possibility of the application of Thomas Fiering model for monthly rainfall data generation and the transition probability matrix and alternating renewal process methods for daily rainfall data generation for semi arid regions. The objective of the study was to analyse systematically the space and time characteristics of monthly, monsoon and annual \ rainfall and rainy days over east Rajasthan. General spatial (1) variability of annual rainfall over whole of Rajasthan was studied using data of 521 long term and short term raingauge stations iri the state. The adequacy 'of•the existing raingauge network in east east Rajasthan was examined and based on the study, catchmentwise $n,i<_, raingauge network has been recommended. . , / ,Jnn-*< :i The study was done broadly on the following lines : '•'• ;" '" ' '••••.' | • • • 'u.rfjij (i) Time series analysis of monthly, monsoon and annual rainfall >•..••.;'• <•••,•;:•'• • • • . , . .,, •iU; '\ -• and rainy days has been carried out to identify (a) persistence in the series • i • r-«iO ./,.• • -• • • ,: •••-,;,.•;.,;' (b) dependance of rainfall of a month or season on the «'*••' ' - I • •' ' ) 1•:• -'.-'' rainfall of the previous month or season and (c) presence of any rising or falling trend and the type (linear or curvilinear) trend (ii } Thomas Fiering model has been used for generation of monthly rainfall data with appropriate transformations of historica,l v monthly rainfall data, • •>>..) > I ;. ••< I ,.*;..., (iii) Two daily rainfall models described above have been used for generation of daily rainfall data, i •-. ».'» . : . ' (iv) Structure of spatial correlation over the region has been analysed and (v) Covariance model has been used for simulation o£ ,rafinfall field within homogeneous rainfall groups . :.r, ,,. 9|1j jB,f: !<),.,; Data of monthly rainfall for the period 1901 -' I^d" for ' 20 }, stations located in the twenty districts in east Rajasthan -has been used for the anlayses. The homogeneity of the rainfall ' and rainy days series has been tested. The series in general were ' found to be stationary. The randomness of the monsoon and; annual i rainfall and rainy days series has been tested using turning.pointi o < ' (ii) test,and through serial correlation. The analysis did not indicate any persistence or non-randomness in the series. J IT! 8maJ**j« 5:fl~T«o7r,i 'B-i^I li«.ls [.tin .:.-,-,..< **r;o! r'%." m <j.i,-. '^Cfossi correlation iainong :the >r.ainfalii jofs ,fqujevinonsoon months June to September was computed ;which <was found .to be low and not significant. Also, no relation was.- found between | the rainfall series of non-monsoon .season and the rainfall of the previous '•' ~i* • ! ';?'•' • HO • '.; ! r»i .7 •< :• .'•>>) „"»• -nit,.: • i monsoon season. This has indicated the independance of rainfall ! ' 'if i*fi I. ♦»!»'!••■ '.iji noi,,.|ib" . -•• -i ;iu -"i i ... of any month or season from the rainfall of any previous month or '.1 i ! I •> ^ i•' ;l:(, ';>', j • •.. season. Analysis for presence of trend has been done by fitting •c?' ••• •.'?-/•:- .o riJni'.a • tt, i • • .1 • linear and polynomial (2nd order) regression to monthly, monsoon and annual series of rainfall and monsoon and annual series of rainy days. Four stations Banswara, Dausa, Pratapgarh and Sirohi have shown trend in the series which was tested to be significant. Two other stations Ajmer, Jalore- and Khanpur (Jhalawar District) have 'also indicated some trend, positive in the case of the earlier two and negative in the case of the later. These were, however,, statistically not significant. Rainy days series of Bharatpur and Dholpur both representing the northeastern part of Rajasthan have indicated a negative trend which was statistically significant. 'Correlation between rainfall and rainy days at some stations indicated that the number of rainy days influence to some extent the total .rainfall .amounts received in a month or season. Correlation between rainfall amount and intensity (amount of rainfall per rainy day) is comparatively less and there is no relation between rainy days and intensity. >Thoma"s tiering model has been used for generation of monthly rainffeiil dafta''by incorporating modifications and transformation of ( iii ) the historical series. The use of square root transformation and power transformation gave better results than log transformation. Transition Probability Matrix ( TPM ) method and the Alternating Renewal Process (ARP) method have been used for simulation and generation of daily rainfall data. In the TPM method, the rainfall is modeled to occur in seven classes. Daily rainfall data for 90 years is generated from 30 years of historical series (1961-1990) normalised using the Box-Cox transformation. In the ARP method, the run length of a dry or wet spell is modeled by a two parameter truncated binomial distribution, and the rainfall amounts above a given rainfall threshold are generated by the two parameter Gamma distribution. Daily rainfall data of 30 years has been generated using 30 years of historical data as above. The performance of the two methods was evaluated by comparing the statistics of the generated series with the statistics of the historical series . The TPM method was found to have performed well in comparison to the ARP method based on the reproduction of the statistical parameters like mean, standard deviation, average number of wet days and highest daily rainfall which were used as criterion for comparison of generated and historical series. Cluster analysis has been used for .sub-dividing the study area and grouping the different raingauge stations. The clusters identified by this way have been used for selecting the representative stations from each cluster for studying the space time characteristics of rainfall in each cluster. ( iv ) Correlation structure of 20 stations ' representing the 20 districts in the region has been studied by computing pairwi'se correlation among the monthly, monsoon and annual seriesi of rainfall and monsoon and annual series of rainy days.'It was seen that the spatial correlation structure, i.e the relationship between inter-station correlation and inter-station distance could be mathematically represented by an exponential relation. In the case of rainy days, the curve representing the above relation was decaying less monotonously indicating widespread occurrence of rainfall over space during the monsoon season. i, Rainfall field has been simulated by the covariance model using the Williams and El Kadi algorithm. The rainfall field simu lation for homogeneous rainfall regions within the study area has given satisfactory results. The study has brought out some important findings regarding the rainfall behaviour in the semi arid monsoon climate. These are (!) The rainfall and rainy days series for different time scales such as month, season (monsoon) and an year are, generally,; random and there is no persistence. (ii) There is no trend in the series of either rainfall or rainy days excepting in a few cases which have indicated positive or negative trends . (iii) The Thomas Fiering model using square root transformation was found to be suitable for generation of monthly rainfall data for regions with semi arid monsoon climate. (iv) The transition probability matrix method was found to be satisfactory for generation of daily rainfall data in regions (v) of semi arid climates with a prominent humid season. (v) Spatial correlation structure is well defined and could be represented mathematically. (vi) The covariance model was found to be capable of simulating the rainfall field satisfactorily. It is believed that, the analysis and procedures used in the present study would be useful for similar studies in other semi arid regions for a better understanding of the behaviour of rainfall occurrence and its space time characteristics. The study has pointed towards some topics needing further investigation. These include the assessment of effects of affores tation and deforestation on rainfall and monitoring of dust content in the atmosphere close to the ground and studying its influence on rainfall in semi arid regions.