prediction of hydrological events for planning rainfed_hysj_47!03!0435

8/12/2019 Prediction of Hydrological Events for Planning Rainfed_hysj_47!03!0435

1/14

HydrologicalSciences~Journal~des SciencesHydrologiques. 47(3) June 2002 43 5

Prediction of hydrological events for planningrainfed riceB.PANIGRAHI, SUDHINDRA NATH PANDADepartme nt of Agricultural and FoodEngineering, Indian Institute of Technology,Kharagpur 721302, Indiasnp(5)agfe.iitkgp.ernet.inR. MULLInstitute ofWaterResourcesManagement, Hydrology and Agricultural Hydraulic Engineering,Universityof Hanno ver, Appelstrasse9 A,D-30167Hannover, GermanyAbstract Daily rainfall and evaporation data of 30 years (1969-1998) are analysed forprediction of hydrological events , such as occurrence of effective monsoon (EM) andwithdrawal of monsoon (WM), for planning rainfed rice in Eastern India. Thosehydrological events are treated with normal (ND), lognormal (LND), Pearson type 111(P TD ) , and log -Pearson ty pe 111 (LP TD ) distr ibutions. The historical data of hyd ro-logical events are a lso t ransformed by pow er and modified S M EM AX transformat ionsand ND is f it ted. Normal dis tr ibution fit ted to power transformed data is found to bethe best for both the events . Accordingly, these events are predicted at variousprobabilit ies of exceedence (PE). At 50% PE level, the monsoon is effective in theregion for 110 days. The study reveals that short-duration rice will face less droughtstress than long-duration rice during the critical growth stage and will produce higheryield under rainfed conditions. At 50 and 70% PE levels , the s imulated yield of short-duration rainfed rice is observed to be 2265 and 1890 kg ha" 1, respectively, that is 6.7and 7.6%o mo re, respective ly, than the yield of long -duration rainfed rice.Key wor ds critical growth stage; drought stress; monsoon occurrenc e; prediction; probabilityof exceedence; rainfed rice; IndiaPredetermination d vnements hydrologiques pour la planificationde la culture du riz pluvialR s u m Des donnes de prcip i ta t ion e t d 'vaporat ion journal ires sur t rente ans(1969-1998) ont t analyses pour prdterminer des vnements hydrologiques ,comme l 'occurrence ou l 'absence de mousson, afin d 'aider planifier la culture du rizpluvial dans l 'Est de l ' Inde. Ces vnements hydrologiques ont t analyss avec desdistr ibutions normale, lognormale, Pearson type III et log-Pearson type III . Lesdonnes his toriques des vnements hydrologiques ont galement t traites grce des t rans format ions pu issance et SMEM AX modif i , avant a jus tement d 'unedistr ibution normale. La distr ibution normale ajuste aux donnes modifies selon lat ransformat ion puissance donne les mei l leurs rsul ta ts pour les deux types d 'vnements , qui sont ainsi prdtermins avec diffrentes probabilits de dpassement. Pour5 0% de probabilit de dpassement, la mousson est effective dans la rgion sur110 jou rs . L' tud e rv le que le r iz cycle court subit un s tress hydrique m oind re qu ele r iz cycle long durant la priode critique de croissance, et qu' il a un meilleurrende me nt en condition s hum ides. A 50 et 70%> de probabilit de dpa ssem ent, lesrendements s imuls du r iz p luvial cycle cour t sont respect ivement de 2265 et1890 kg ha"1, soit 6.7 et 7.6% de plus qu'avec le r iz cycle long.Mo ts clefs stade critique de croissance; stress hydriqu e; occurrence de mousson;predetermination; probabilit de dpassement; riz pluvial; Inde

INTRODUCTIONRice is the most important crop of eastern India that is generally grown under rainfedconditions. The region has about 67% of the total area of rainfed rice in India; yet itsOpenfordiscussionuntil1 December 2002


2/14

436 B. Panigrahi et al.

contribution to the country's rice production is only about 50% (Ghildyal, 1989). Itcontinues to lag behind other regions of India in terms of rainfed rice productivity, w hichis currently 1586 kg ha"' com pared to the country 's average productivity of1921kg ha"1.Ma ny constraints are associated with the poor economical growth of the region. Frag

men ted land holdings, lack of assured wa ter supply, and poverty o f farmers hindering theiruse of intensive agricultural inputs are the major constraints hindering the adoption ofmodern technology, and hence resulting in an instability in crop production. There is limited scope for further development of surface storage facilities for irrigation water, due toenvironmental problems of land submergence, waterlogging and salinization of canalcom man ds, and higher investment in the area of infrastructural developm ent. Furtherm ore,tubewell irrigation is not feasible in all locations due to hydrogeological and installationcost factors. Therefore, rice growers of the region depend mainly on rainfall.About 70% of the total cultivated area of Eastern India is rainfed. The regionreceives annual rainfall varying from 1194 to 2290 mm, of which about 80% is

received in four rainy months between June and September. During this period, about50% of the annual rainfall comes from a few intense storms. Water received from suchintense storms is subject to high runoff losses coupled with the erratic nature of theonset, distribution and withdrawal of monsoon, which increases the probability ofwater stress at various grow th stages of the rice crop (Bhuiyan & Goo nasekera, 1988).Substantial yield losses due to drought in most upland rainfed rice environments anddue to excess water in low lands have been reported from astern India (Widawsky &O'Toole, 1990). The productivity of the rainfed rice ecosystem is observed to besignificantly lower than that of irrigated rice.In view of the above conditions, there is tremendous scope for rainwater manage

ment and its optimum utilization in Eastern India. In order to improve crop productivityand achieve sustainability of rainfed agriculture, it is essential to plan agriculture on ascientific ba sis, ma king th e best use of the rainfall pattern. This necessitates studying thehydrological events of the occurrence of effective monsoon (EM) and withdrawal ofmonsoon (WM) and preparing the contingent crop plan accordingly.The occu rrence of EM is the vital rainfall characteristic influencing rainfed farming.Agricultural operation in India starts with the onset of the southwest monsoon. It isessential to predict the occu rrence of EM accurately, since a slight delay in the sowing ofrainfed crops may lead to drastic reduction in grain yield and may adversely affect thenext crop too. While determining the criteria of occurrence of EM, the type of soil and

the crop variety to be grown in the field should be considered. Effective monsoon shouldprovide optimum moisture in the effective seeding zone for tillage and for satisfactoryseed germination after accounting for evaporation and runoff losses. Normally 75% ofavailable soil moisture (ASM) in the seeding zone may be considered as the lower limitfor good germination and initial plant growth of monsoon crops.Considering the above factors, Verma & Sarma (1990) proposed the following fourcriteria for determining the occurrence of EM:

(a) The first da y's rain in a 7-day perio d should not be less than e mm, where e is theaverage daily seasonal pan evaporation;(b) The total rain during the 7-day period should not be less than the effective onsetrainfall EOR);(c) At least three out of seven days should be rainy days having a minim um of 2.5 m mrain each day; and


3/14

Predictionofhydro logical even ts for planning rainfedrice 437

(d) If criteria (a)-(c) are satisfied in one week of any month, but are followed by aprolonged dry spell, then it should be considered as a pre-monsoon spell and notthe occurrence of EM.The effective onset rainfall, EOR (in mm), is computed as:EOR = 0.75 FC - WP)D + 5e+R O 1)

where FC and WP (%) are moisture content on a volume basis at field capacity and atpermanent wilting point, respectively, D is the effective seeding zone depth or theploughing depth whichever is greater (cm), e is the average daily seasonal pan evaporation (mm), and RO is runoff (mm). Chore (1998) and Gupta (1999) proposed thefollowing criteria for defining WM:(a) the last two days in seven days of the last wet spell of the season should have atleast 2.5 mm rain,(b) the rain in a 7-day spell should be mo re than 2e + 10 ) mm, and(c) one of the rainy days of the last wet spell should hav e more than emm of rain.Rainfall is the only source of wa ter in rainfed a gricultu re and rainfall deficienciesin the crop growing season cause a drought spell. Reduction of crop yield depends onthe severity of the drought spell and the different crop growth stages during which itoccurs. Various authors have analysed rainfall to define drought spells in differentways and have classified them into several categories depending on their severity(Banerji & Chhabra, 1963; Sh arm ae? a/. , 1979; Ashokraj, 1979).

The hydrological events are stochastic and so must be predicted at a certainprobability of exceedence PE) level by appropriate distribution functions. Mosthydrological data are highly skewed and normal distribution (ND) often providesinadequate fit to a set of observations. Hence, there is a need to obtain a best fittingdistribution function for the random sample. On the other hand, historical data may betransformed and the ND can be fitted. Bethlahmy (1977) suggested SMEMAX (Small,MEdian and MAXimum) transformation, which was used for dry spell analysis byAldabagh et al. (1982). Rasheed et al. (1982) proposed modified SMEMAXtransformation, which was used by Thakur (1997) for consecutive day rainfall analysisfor drainage planning. Chander et al. (1978) and Aldabagh et al. (1982) used Box-Coxpower transformation for normalizing the hydrological data series and reported that ittransformed the data to a near normal series better than SMEMAX.

An attempt was made in the present study in Eastern India to apply ND, lognormal(LND), Pearson type III (PTD) and log-Pearson type III (LPTD) distributions andpower and modified SMEMAX transformations to predict hydrological events such asthe occurrence of EM and WM by the best fitting distribution at different PE levels.The findings of the study should be helpful in deciding on the sowing time of rainfedrice, in crop planning and in proposing measures for a supplementary irrigation fromharvested rainwater. Another objective of the study is to suggest either short- or long-duration rice that will effectively utilize the mons oon rain of the region and give h igheryields under rainfed conditions.

M A T E R I A L S A N D M E T H O D SThe present study was carried out in Eastern India, comprising states of Orissa, WestBengal, Bihar, Assam, eastern Uttar Pradesh, and eastern Madhya Pradesh. The


4/14

438 B.Panigra hi et al.

climate is humid sub-tropical and the mean annual rainfall is 1638 mm, of which about80% is received in the four rainy (monsoon) months of June to September. All thestates of the Eastern India region are hydrometeorologically similar, in terms of rainfalland evaporative demand of the atmosphere (Sharmaetal.,1979). Therefore, instead ofanalysing the data of rainfall and evaporation for different locations in the region, thedata from a representative centre were selected for predicting the hydrological events:daily rainfall and pan evaporation data of 30 years (1969-1998) were collected fromthe meteorological observatory of the Indian Institute of Technology, Kharagpur, WestBengal (India).

The occurrence of EM was determined by adopting the four criteria proposed byVerma & Sarma (1990). Average values ofF C and WP for the dominant sandy loamsoil group ofthe region were found to be 25 and 11%on a volumetric basis, respectively. Rice is the major crop grown in the rainy season in Eastern India, covering about90% ofthetotal cultivated area. The effective seeding zone depth of rice to be wettedfor good germination is assumed to be 15 cm. The maximum possible ploughing depthby bullock power is found to be 20 cm; therefore, for the computation of EORbyequation (1), the value ofDwas taken as 20 cm. Runoffwasconsidered as 10% of theonset-week rainfall for the sandy loam soil ofthestudy area (Verma & Sarma, 1990).Table 1Estimates of occurrence of effective monsoon (EM) and withdrawal of monsoon (WM).Year196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998

EM11 June14 June1 June22 June6 July15 July18 June26 June13 June25 June22 June1 June13 June25 June25 July4 June3 June23 June25 May16 June14 June21 May2 June8 June17 June21 June19 June31 May10 June21 June

WM30 September2 October23 October27 September19 October9 October9 October23 September18 September3 October19 September19 September1 October18 September10 October13 September15 October8 October28 September5 October28 September8 October7 October30 September18 October10 October14 October29 September2 October2 October


5/14

Prediction ofhydrologicalevents for planning rainfed rice 439

The occurrence of WM was identified by the methodology as suggested by Chore(1998) and Gupta (1999).The historical data of occurrence of EM and WM for the 30-year study periodwere com puted and are sho wn in Table 1. The data of all the above m entioned events

were conv erted to the equivalent calendar days (ECD ) taking 1 January as 1, 2 Januaryas 2, . . . , and 31 December as 365 (for a leap year, 31 December is 366). In order totest the normalization effect of data series containing ECD of occurrence of EM andWM, the coefficient of skewness Cs) was computed for the untransformed as well asthe transformed data series. Then these historical data were treated with probabilitydistributions such as ND, LND, PTD and LPTD, and also with ND, after beingtransformed by modified SMEMAX and power. Using the standard software SMADA,the probable values of ECD for the events were determined at various PE levels usingthe same distributions.

Power transformationPower transformation by the Box-C ox m odel follows that:

Yi= f--l)/X X*0 (2)7, =log/c X= 0 (3)

where / is the variety of the given series, 7, is the transformed variety, and X is aconstant of transformation, which generally ranges from -1 to 1.

An iterative solution was carried out to estimate the value of X so that the serieswas alm ost transformed to a norm al series with value ofCsapproaching zero.

Modif ied SMEMAX transformationA cco rding to this me thod , the transform ed variate , 7,-, is given as:

=Y,[1 - Xm-Xi)l Xm- Xs)] X,Xm (5)

whereX\, X, andXs are highest, median, and lowest data in the untransformed series,X; is the variate of the series (untransform ed), Ym is a constant assumed as 50, 7; is thetransformed value of the variate of the series, Xf, and i is the number of data in therange 1, 2, 3, ..., N.This transformed 7; series was assum ed to be norma lly distributed. H enc e, i t waspossible to estimate the variable Y for any specified PE level by the proceduresuggested by Kottegoda (1980). The original X variable at any PE level can bedetermined corre spond ing to the F value by:X = Xm + Xm - Xs) {7/5 0 - 1} if Y> 50 (6)X=X,+ Xi~~X, {7/50- 1} if 7 < 50 (7)The estimates of ECD of occurrence of effective monsoon (EM) by ND, PTD,powe r and modified SM EM AX transformations are shown in Fig. 1 and those by LN D

and LPTD are shown in Fig. 2. Similarly, the estimates of ECD of occurrence of


6/14

440 B. Panigrahi et al.

OO

210200

190

180170

160

150140

O b s e r v e d N o r m a l ^ P e a r s o n T y p e IIM o d i fi e d S M E M A X o P o w e r

10 2 0 ~~i 1 1 1 r~30 4 0 50 60 70 8 0 90 95Probability of exceedence, PE (%)Fi g. 1 Occ urrenc e of EM at various PE by norm al, Pearson type III , modifiedSMEMAX and power t ransformat ion .

400

Qo

i 1 1 1 r20 30 40 50 60 70Probability of exceedence, PE (%) 95Fig. 2 Occurrence of EM at various PE by lognormal and log-Pearson Type IIIdis tr ibution.

withdrawal of monsoon (WM) by ND, PTD, power and modified SMEMAXtransformations are shown in Fig. 3 and those by LND and LPTD are shown in Fig. 4.The curves in Figs 1-4 are drawn using ORIGIN software, with a probability scale in


7/14

Prediction of hydrological events for planning rainfed rice 441

DO

300

2 9 0 -

2 8 0 -

2 7 0 -

2 6 0 -

250

O b s e r v e d N o r m a l * P e a r s o n T y p eM o d if ie d S M E M A X ^ o ^ P o w e r

10 9 0 9 50 30 40 50 60 70 80Probability of exceedence, PE (%)

Fig. 3 Occurrence of WM at various PE by normal, Pearson type III, modified SM EM AXand power transformation.

QO

400

300

200

-Observed log norm al- log Pearson type III

I 1 1 1 1 r -20 30 40 50 60 70

Proba b i l i t y o f exce ede nce , PE (%)

Fig. 4 Occurrence of W M at various PE by lognormal and log-Pearson Ty pe HI distribution.

the abscissa ranging from 2 to 9 5% . The o rdinate is on a linear scale for Figs 1 and 3(drawn on normal probability paper) and on a logarithmic scale (logio) for Figs 2 and 4(drawn on lognormal prob ability pap er).


8/14

442 B.P anigrah i et al.

Rice is to be grown under rainfed conditions so that it effectively utilizes the entiremonsoon rain of the region. In this study, two types of rice were considered: short-duration rice that can be harvested before the monsoon withdraws, and long-durationrice that is harvested after the withdrawal of monsoon. In order to compare the yieldpotential of both types of rice, a water balance model of rainfed rice was run withoutconsideration of supplemental irrigation (SI). The root zone soil moisture status(RZSMS), which affects the growth and, ultimately, the yield of rice was computed byrunning the water balance m odel c onsidering the effective root zone (45 cm) as asingle layer. The water balance model of rainfed rice, neglecting the contribution dueto the capillary rise of groundwater (the groundwater table in the region lies more than1.5 m below the effective root zone depth of rice) is given as:

SMC,=SMQ. i+Pi- AETi - SP, (8)where SM C is the soil moisture content (mm), P is rainfall (mm), SP is seepage andpercolation loss (mm),AET is actual vapotranspiration (mm ), and /' is the time index,taken as 1 day in the study.

Computation of the parameters AET andSP in the above water balance model wasdone according to the methodologies suggested by Panigrahi & Panda (2001). Thevalues of crop coefficient of rice during the crop establishment (CE), crop development (CD), mid season critical growth stage (CGS) and late season stage (LS) wereassumed as 1.05, 1.10, 1.10 and 0.95, respectively (Doorenbos & Pruitt, 1977). For

0 4 DAS 29 DAS 49 DAS 78 DAS 104 DASCropestablishmentstage

Crop^development^stage *Mid season staqe(Critical growth)

Late season

Germinationstage0 4 DAS 34 DAS 64 DAS 103 134 DAS

CropestablishmentstageCrop__ developm ent^

* stage *Mid season staae(Critical growth)

GerminationstageFig . 5 Schematic presentation of different growth cycles of rice: (a) short duration riceand (b) long duration rice.


9/14


short-duration rice of 100 days, the duration of each different growth stage, for CE,CD,CGS and LS, is 25, 20, 30 and 25 days, respectively, whereas, for long-durationrice of 130 days, the stages are 30, 30, 40 and 30 days, respectively (Fig. 5). Of thefour growth stages of rice, the most important is the CGS, which includes booting,panicle initiation, flowering and milking, and which occurs between 49 and 78 daysafter sowing (DAS) for short-duration rice of 100 days and 64-103 DAS for long-duration rice of130days (Mandai, 1990).The simulation was run for 30 years (1969-1998) from the onset day of EM untilthe harvest of each year using the available meteorological, soil and crop data of thestudy site. The total duration of simulation for short- and long-duration rice was 104and 134 days, respectively, considering the four days required for germination aftersowing on the day of occurrence of EM. During the first four days of thegerminationphase, the soil is under bare conditions. Bare soil evaporation was computed asproposed by Jensen etal. (1993) and was used in place ofAET in the water balancemodel.The soil moisture stress during the CGS of rice affects the yield more than thatoccurring in the other stages. It was reported by Doorenbos & Kassam (1979) andAllenetal.(1998) that the yield of rice declines when the RZSMS is depleted by 20%of saturation m oisture content during the CG S, referred to herein as the threshold valueof RZSMS from the yield reduction point of view. In the present study, a day isdefined as a drought day if the RZSMS of rice is below the threshold value and thediscussion is limited to the CG S. Ifthedrought stress occurs for more than a day, thenit is designated as drought spell. The percentage of probabilities of the different days tohave drought during the CGS were estimated by the Markov probability model(Kottegoda, 1980) for both short- and long-duration rice (Fig. 6).The yield of rice under a rainfed farming system was also simulated for both short-and long-duration rice varieties using the multiple regression based model (Panigrahi,

Short Duration Rice- Long Duration Rice

91 94 97 100 103

Fig. 6 Probability of drought on different days during CGS of short- and long-durationrice.


10/14

444 B.Panigrahiet al.

2001), which relates the yield of rice, Yr, to the values ofA ET during different growthstages as:Y r = -13 .0 6 + 0.05AETCE +0.07AETCD+0.75AETCGS+0A5AETLS R2= 0.87) (9)where Yr is in 10 kg ha"1, AET is actual vapotranspiration of rice (cm). The abovemodel was validated using two years of field data (1999-2000). The simulated andobserved yields of rice were found to be statistically significant with high values ofcoefficient of determination for both the years R~ > 0.90). The water balance modelwas run to determine AET on a daily basis for both varieties of rice. From the dailyvalues, AET during the different growth stages was computed. Finally, using themultilinear regression model (equation (9)), the yield of rice for both the varieties wassimulated for 30 years. The value of F, varies from year to year depending on therainfall and other meteorological data. So, to predict Yr at different PE levels, the dataseries containing the yields of both short- and long-duration rice were treated withpower transformation and using the standard software SMADA (Fig. 7).In order to verify the adequacy of the residual RZSMS for the germination ofseeds at the sowing time of another short-duration dry land crop in the winter season,the water balance model (equation (8)) was extended from the harvest day of the riceuntil the sowing of the next dry-land crop, which generally requires 15 days after therice harvest for land preparation. In the water balance model, AET was replaced bybare soil evaporation during the 15-day period (Sanchez-Cohen et al., 1997). If theresidual RZSMS in the seeding zone (top 15 cm) is less than 75% ASM, then the dryland crops cannot germinate and hence the crop cannot be grown (Verma & Sarma,1990). A value of 75%ASM in the seeding zone was identified as the optimum soilmoisture content for seed germination. The soil moisture content at the sowing time ofthe second dry land crop in the winter season within 45 cm depth was computed by thewater balance model and, i ts uniform distribution over the entire 45 cm depth wasassumed. Then, the soil moisture values for the top 15 cm layer, after cultivation ofboth short- and long-duration rice in the preceding rainy season were determined forthe 30 years of simulation. Finally, the soil moisture storage values of the top 15 cm

"1 n n i n p i i i f n2 5 10 20 30 40 50 60 70 80 90 95Probability of exceedence, PE (%)Fig. 7 Variation of rice yield at different probabilities of exceedence.


11/14


I 10co" SMC after harvest of short duration rice-| - -* - -S M C after harvest of long duration rice 5 'E Optimum germination moisture content"5

0 J r , - , - 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998

YearsFig . 8 Validation of available soil mo isture content in seeding zone in different year s .

soil depth were converted to volumetric values (%) by dividing the soil moisturestorage by the depth of the soil layer. A programme was written in C language tocompute the soil moisture content at the sowing time of the dry-land crop to identifythe possibilities of crop germination. Figure 8 represents the variation of soil moisturecontent in the seeding zone of the dry-land crop.RESULTS AND DISCUSSIONOnset of effective monsoon (EM)The study reveals that occurrence of EM in Eastern India is highly fluctuating, rangingfrom 21 May to 25 July (Table 1). It was observed that the EM data series is skewed tothe right with aCs value of 1.061. This reveals that ND would not fit well to the series.The Cs value of the power transformation series was found to be very low (-0.00012)and so it is presumed that ND can fit well to this transformed series. Next to power,modified SMEMAX transformation was found to be the best with a Cs value of 0.01.The estimates of the ECD of occurrence of EM by different distributions andtransformations at variou s PE levels are show n in Figs 1 and 2. At lower PE levels upto 15% , the values of EM pr edicted by variou s distributions are seen to differ greatlyfrom the observed values. However, at higher PE levels, the curves are closer to theobserved series. Occurrence of EM at anyPE level predicted by power transformationwas found to be closest to the observed values, followed by modified SMEMAXtransformation (Fig. 1). At 5 0%P E level, the ECD of the occurrence of EM by powerand modified SMEMAX transformation were found to be 167 and 166, correspondingto 16 and 15 June, respectively (Fig. 1).


12/14

446 B.P anigrah i et al.

The model for predicting the occurrence of EM was tentatively evaluated bycomparing the predicted date of EM with available observations from the field on theaverage sowing dates in 1998. The observed dates of sowing were found to range from12 to 24 June, with the average date of sowing being 17 June, which succeeds thepredicted date of EM (16 June, predicted by power transformation at 50%PE) by oneday only. Thus the comparison is in favour of accepting the model as far as occurrenceof EM is concerned.

Withdrawal of monsoon (WM)Dates of WM in the study area were found to range from 13 September to 23 October(Table 1). The data series of the ECD of WM is skewed to the left with a Cs value of-0.087. With power transformation, the value of Cs was found to be -0.0003,indicating that the series is normally distributed. Modified SMEMAX transformationis the next best transformation for studying the normalizing effect of WM data serieswith a Cs value of-0 .035.Figures 3 and 4 present the estimates of the EC D of WM bydifferent distributions and transformations at various levels ofPE. The closeness of thevalues of WM with observed values at different PE levels signifies that powertransformation is the best followed by modified SMEMAX for the prediction ofoccurrence of WM. At 50% PE, both power and modified SMEMAX transformationgive the date of WM as 3 October (ECD = 276, Fig. 3).

Crop planning in rainfed ricelandsAt 50% PE, it was observed by the best fitting distribution that monsoon starts on16 June and withdraw s on 3 O ctober; thus remaining for 110 days in the region.Hence, to utilize the monsoon rain effectively, short-duration rice of about 100 daysma y be grow n in the rainy season in the study region. On the other hand, long-durationrice may face moisture stress during the last part of the CGS and, consequently, thechance of getting a good yield is diminished. Study of the water balance modelrevealed that there is at least 50% probability that short-duration rice will sufferdrought or water stress during the last seven days of the CGS (72-78 DAS). Thus,there is a likelihood of a reduction in the yield of rice grown under rainfed conditionswithout any supplementary irrigation provision. For long-duration rice, the probabilityof occurrence of drought during CGS is more than 50% between 72 and 103 DAS. Therisk of drought in long-duration rice is more than 70% during the last nine days (95-103 DA S) of the CG S (Fig. 6). Thu s, the study reveals that the likelihood of a reducedyield for long-duration rice is greater than that for short-duration rice under the rainfedsystem.

The study shows that the simulated yield of short-duration rice is greater than that oflong-duration rice for all the years. The smaller yields of long-duration rice may be due tothe longer-duration drought spells occurring during the CGS, which adversely affectgrowth and, consequently, the yield. Th e soil rem ains under m oisture stress for more daysduring the last part of the C GS , which is the "m ilking" stage when the rice grains swell.The simulated yields of short- and long-duration rice ranged from 1060 to 3650 andfrom 1015 to 3610 kg ha ' , respectively. The Csvalue of the data series containing y ieldsof both short- and long-duration rice are skewed to the right with values of 0.052 and


13/14

Predictionofhydrological events for planning rainfedrice 447

0.068, respectively. However, with power transformation, the series were found to benormalised with zero values ofCs. Figure 7 indicates that the yield of short-duration riceis greater than that of long-duration rice at anyPE level. At 50 and 70%PE levels, theyields of short-duration rice we re 2265 and 1890 kg ha"1, respectively, wherea s those forthe long-duration rice at the corresponding PE levels were 2123 and 1756 kg ha"1,respectively (i.e. the yields of short-duration rice at 50 and 70%PE levels, were 6.7 and7.6% higher, respectively, than those of long-duration rice at corresponding PE levels),leading one to conclude that short-duration rice produces greater yields under rainfedconditions than long-du ration rice.

The study further reveals that, in 15 out of 30 years, it was not possible to grow adry-land crop in the winter after the harvest of short-duration rice. In the case of long-duration rice, the next dry land crop could not be grown after the harvest of rice in27 out of 30 years (Fig. 8). Thus, to allow two crops to be grown in most years underthe rainfed farming system in the study region, short-duration rice should be grown inthe rainy season. A survey of cropping patterns followed by farmers in the studyregion in 1998 shows that about 45% of those who grew short-duration rice (of about100 days) in the rainy season could grow another dry-land crop such as mustard (oilseed) in the winter after the rice harvest. However, 55% of farmers in the region whogrew long-duration rice (of about 130 days) could not grow a second crop in thefollowing winter. Moreover, the survey found that the average yield of long-durationrice in 1998 was 1913 kg ha"1 under rainfed conditions, whereas that of short-durationrice was 2009 kg ha"1(i .e. 5.02% m ore).

CONCLUSIONSAt 50%) PE, it was observed by the best fitting distribution that monsoon starts on 16June and withdraws on 3 October; thus remaining for 110 days in the region. A long-duration rice grow n in the rainfed system has a greater cha nce of suffering frommoisture stress during the CGS stage as, by this time, the monsoon is either weak or isabout to terminate in the region. This will result in severe yield reduction of rainfedrice in the absence of supplementary irrigation. On the other hand, short-duration riceof 100-110 days duration can effectively utilize the entire monsoon rain. This willminimize the chance of moisture stress during the CGS and thus improve the yield.Moreover, in 50% of the years, the residual soil moisture after the early harvest ofshort-duration rice allows another short-duration dry-land crop such as mustard (oilseed) to be grown in the winter without requiring much irrigation. The study showsthat the residual soil moisture after the harvest of long-duration rice reduces thepossibility of raising a second crop in the winter to only 10% of the years.

R E F E R E N C E SAldabagh, A. S., Rasheed, H. & Ramamoorthy, M. V. (1982) Dry days analysis for planning the supplemental irrigationschemes . Trans. Am. Soe. Agric. Engrs 25(1), 150-153,159.Allen, R. G., Pereira, L. S., Raes, D. & Smith, M. (1998) Crop Evapotransp iration. Guidelines for Comp uting W aterRequirements. PAO Irrigation and Drainage Paper no. 56, Rome, Italy.Ashokraj, P. C. (1979) Onset of effective monsoon and critical dry spells. IARI Research Bull. 11, 6-1 8. W ater

Technology Centre, Indian Agricultural Research Institute, New Delhi, India.Banerji, S. & Chhabra, B. M. (1963) Drought conditions in the Telengana division (Andhra Pradesh) during southwest

monsoon season. Ind. J. Met. Geogr. 14(4), 403- 415.Bethlahmy, N. (1977) Flood analysis by SMEMAX transformation. J. Hydraul. Div. ASCE. 103(HYI), 69 -78 .


14/14

4 4 8 B. Panigrahi et al.

Bhuiyan, S. . & Goo nasekera, K. (1988) Rainwater managem ent for increasing production office and upland crops. Paperpresented at the International Rice Research Conf International Rice Research Institute, Los Banos, ThePhilippines.

Chander, S., Spolia, S. K. & Kumar, A. (1978) Flood frequency analysis by power transformation. J. Hvdraul. Div. ASCE104(HYII), 1495-1504.

Cho re, K. K. (199 8) Optimum design of on-farm re servoir for supplemen tal irrigation in rainfed agriculture ofChhotanag pur plateau. MTech. Thesis, Indian Institute of Technology, Kharagpur, India (unpublished).Doorenbos, J. & Kassam, A.M. (1979) Yield response to water. FAO Irrigation and Drainage Paper no. 33, Rome. Italy.Doorenbos, J.& Pruitt, W. O. (1977) Guidelines for predicting crop water requirements. FAO Irrigation and DrainagePaper no. 24, Rome, Italy.

Ghildyal, B. P. (1989) Rice production in eastern Indiaissues in rice research. J. Agric. Issues 1, 79 - 91 .Gupta, P. K (1999) Drought analysis of Kymore plateau and Satpura hills of M.P. MTech. Thesis, Jawaharlal Nehru KrishiVishwa Vidyalaya, Jabalpur, India (unp ublished).Jensen, J. R., Mannan, S. M. A. & Uddin, S. M. N. (1993) Irrigation requirement of transplanted monsoon rice inBangladesh. Agric. Wat. Manage. 23(3), 199-212.Kottegoda, N. T. (1980)Stochastic Water Resources Technology. John Wiley & Sons Inc., New York, USA.Mandai, D. K. (1990) Studies on water and nitrogen balance under rice-wheat sequences in lateritic sandy loam soil. PhD

Thesis, Dept. of Agric. and Food Engineering, Indian Institute of Technology, Kharagpur, India.Panigrahi, B. (2001) Water balance simulation for optimum design of on-farm reservoir in rainfed farming system. PhD

Thesis, Dept. of Agric. and Food Engineering, Indian Institute of Technology, Kharagpur, India (unpublished).Panigrahi, B. & Panda, S. N. (2001) Simulation of ponding and soil moisture status through water balance model forrainfed upland rice.Agric. EngngJ. 10(1), 213-230.Rasheed, H. R. , Ramamoorthy, M. V. & Aldabagh, A. S. (1982) Modified SMEMAX transformation for frequencyanalysis. Wat. Rsout: Bull. 18(3),509-512 .Sanchez-Cohen, I., Lopes V. L., Slack, D. C. & Fagel, M. M. (1997) Water balance model for small-scale water harvestingsystems. J. Irrig. Drain. EngngASCE 123(2), 123-128.Sharm a, H. C , Chau han, H. S. & Ram, S. (1979 ) Probability analysis of rainfall for crop planning. J. Aric. Engng. IndianSoc. Agric. Engrs 16(3), 87-94.Thakur, N. K. (1997) An integrated simulation model for assessment of surface drainage requirement of crops. MTech.

Thesis, Dept. of Agric. and Food Engineering, Indian Institute of Technology, Kharagpur, India.Verma, II. N. & Sarma, P. B. S. (1990). Criteria for identifying effective monsoon for sowing in rainfed agriculture.J. Irrig. Power April, 177-184.Widawsky, D. A. & O'Toole, J . C. (1990) Prioritizing the Rice Biotechnology Research Agenda for Eastern India. TheRockefeller Foundation, New York, USA.

Received 7 August 2000; accepted 10 January 2002

prediction of hydrological events for planning rainfed_hysj_47!03!0435

Documents