PAWEES 2012International Confere ce on

"C al enges of Water & Environmental Managementin Monsoon Asia"

27-29 November 2012Royal Irrigation Department (Pakkred),

Thailand

ttJDth IofVIrvelJ411 ofClwl, fng,"""1'lg lOfJ

CONTACT INFORMATION AND DETAILS:

For US-related questions, please contact:Contact Person: Dr. Sucharit KoontanakulvongWater Resources System Research UnitRoom 203, Bldg. 2 Faculty of Engineering,Chulalongkorn University, BangkokTel: +66-2-218-6426, +66-81-646-9750Fax: +66-2-218-6425Email: sucharit.k@chula.ac.thFor more detailed information you are more than welcome to visit our website:http://project-wre.eng.chula.ac.th/watercu engl

11

Co c Organizing Institutes:

International Society of Paddyand Water EnvironmentEngineering (PAWEES)

Faculty of Engineering,Chulalongkorn University (CU),

Thailand

Faculty of Engineering atKamphaengsaen,

Kasetsart University (KU),Thailand

Royal Irrigation Department (RID),Thailand

Irrigation Engineering AlumniAssociation under H.M.

The King's Patronage (lEA),Thailand

IV

Advisory Committee

1. Mr. Lertviroj Kowattana2. Or Pirom Kamol-ratanakul, MD3. Mr. Vudtechai Kapilakanchana4. Dr. Tai-cheol Kim5 Or. Yohei Sato6. Or. Hiroyuki Konuma7. Or. Sho Shiozawa8. Or. Sun-joo Kim9. Mr. Wei-Fuu Yang

Scientific Committee

10. Or. Bancha Kwanyuen11. Dr. Ekasit Kositsakulchai12. Dr. Masaru Mizoguchi13. Or. Yoshiyuki Shinogi14. Or. Nobumasa Hatcho15. Dr. Yutaka Matsuno16. Dr. Seong Joon Kim17. Dr. Sang Min KIM18. Or. Fi-John Chang19. Or. Ke-Sheng Cheng20. Dr. Sucharit Koontanakulvong21. Dr. Chaiyuth Sukhsri22. Dr. Aksara Putthividhya23. Dr. Somchai Donjadee24. Or. Chirakarn Sirivitmitrie25. Or. Pongsathorn Sopaphun26. Or. Wisuwat Taesombat27. Mr. Santi Thongphamnak28. Dr. Varawoot Vudhivanich29. Dr. Watchara Suidee

Organizing Committee

30. Or. Bonsoom31. Dr. Sucharit32. Mr. Va-Son33. Or. Thanet34. Or. Bancha35. Or. Ekasit36. Or. Yutaka37. Miss Mada

LerdhirunwongKoontanakulvongBoonkirdAksornKwanyuenKositsakulchaiMatsunolaumsupanimit

Royal Irrigation DepartmentChulalongkorn UniversityKasetsart UniversityPAWEESPAWEESFAOJSIDRE, JapanKSAE, KoreaTAES, Taiwan

Kasetsart UuniversityKasetsart UuniversityUniversity of Tokyo, JapanKyusyu University, JapanKinki University, JapanPAWEES, Kinki University.JapanKonkuk University, KoreaGyeongsangNationalUniversity,KoreaNational Taiwan University.TaiwanNational Taiwan University.TaiwanChulalongkorn UniversityChulalongkorn UniversityChulalongkorn UniversityKasetsart UniversityKasetsart UniversityKasetsart UniversityKasetsart UniversityKasetsart UniversityKasetsart UniversityRoyal Irrigation Department

Chulalongkorn UniversityChulalongkorn UniversityRoyal Irrigation DepartmentRoyal Irrigation DepartmentKasetsart UuniversityKasetsart UuniversityPAWEES Kinki University.JapanChulalongkorn University

v

CO TE TS

WELCOME ADDRESS 111

CONFERENCE ORGANIZING INSTITUTES IV

CONTENTS VI

AIMS AND SCOPE 1

PROGRAMME 2

FEATURED SPEECH

Participatory management for irrigation projects

Va-son Boonkird 6

Radioactive contamination of paddy soil and its transfer to rice in fukushima

Sho Shiozawa 7

Farmers' responses to climate change adaptation in irrigation project (Thailand case study)

Assoc. Prof. Chaiyuth Suksri. 8

ABSTRACT

SESSION A Climate Change and Uncertainty

PAPER ID: 000

A review of land-use change scenarios by new climate change scenarios in Korea 10

PAPER ID: 008

Simulation of GHG emission from paddy by DNDC model for climate change impact in Korea .... 11

PAPER ID: 025

Copingwith uncertainties in climate change by stochastic storm rainfall simulation 12

PAPER ID: 032

The study on parameter sensitivity analysis of the Denitrification-Decomposition

(DNDC) model 13

PAPER ID: 046

Climate change impact assessment in Sukhothai Province: intercomparisqn between three Global

Climate Models : 14

PAPER ID: 049

Probability based assessment of Climate Change Impact on Irrigation Systems in Upper Chao

Phraya Basin 15

PAPER ID: 051

Assessment of water supply capacity of agricultural irrigation facilities using MODSIM-DSS

Coupled with SWAT 16

VI

PAPER ID: 052

Assessment of Climate Change Impact on Multi-purpose Dam based on RCP emission scenarios

Using SWAT model 17

PAPER ID: 057

Mitigation method of irrigation systems against climate change in the Chao Phraya Basin,

Thailand 18

PAPER ID: 075

Analysis Framework for Water Resource Policy Decision-Making under

Effects of Climate Change 19

SESSION 8 Participatory Management for Irrigation ProjectsPAPER ID: 005

Relationship between irrigation water management by farmers' group and the levy system 21

PAPER ID: 033

Duration of irrigation vulnerabilities according to agricultural water supply and demand 22

PAPER ID: 044

Assessment of farmer participation in irrigation management for

paddy irrigation development in Myanmar 23

PAPER ID: 045

Impact of participatory approaches on irrigation development and management for Communal

Irrigation System (CIS): Case studies in three upland provinces of north, Philippines 24

PAPER ID: 071

Investigation of paddy field irrigation activities by farmers aiming for

demand-oriented irrigation service 25

SESSION C Emerging Technologies in Water ManagementPAPER ID: 003

Dam management at Drought in dry season paddy irrigation, Case study of Mae Suai Dam ..... 27

PAPER ID: 012

Assessing flood damages of rice in the Chao Phraya Delta, using

MODIS satellite imageries 28

PAPER ID: 015

Optimizing non-flooded irrigation regime under system of rice intensification crop management

using Genetic Algorithms 29

PAPER ID: 018

Investigating the interactive recharge mechanisms between surface water and groundwater over

the Jhuoshuei River Basin in Central Taiwan 30PAPER ID: 020

jlrAls: Agent-Based Modeling for Simulating Irrigation Water Use in Paddy Land 31VII

PAPER ID: 023

A bayesian uncertainty analysis of the modelled surface- and ground-water flows in an agricultural

watershed. .. .. . .. . 32

PAPER ID: 024

A study on drainage efficiency of shortcut canal project in the Lower The chin River. 33

PAPER ID: 031

Drought response by farmers after decreasing water supply 34

PAPER ID: 036

Development of automated irrigation system for food production land 35

PAPER ID: 047Satellite Data Application for flood simulation 36

PAPER ID: 048

Statistical forecasting of rainfall with ENSO index in the Chao Phraya River Basin in Thailand ..37

PAPER ID: 055

Est mation of Streamflow by SWAT Model in Sedone River Basin, LAO PDR '38

PAPER ID: 059

Yoshino and Nan River basins development and management comparative study 39

PAPER ID: 061

System of Environmental-Economic Accounting for water in case of Thailand .40

PAPER ID: 062

Ap lication of Input-Output Table for future water resources management under policy and climate

ch nqe in Thailand: Rayong Province Case study 41

PA ER ID: 063Wc er Footprint of Bioethanol Production in Thailand .42

PA ER ID: 073

Ot.. si Real-time monitoring system for Informing the Optimum Planting and Harvesting Dates of

Ca sava in Rain-fed Upper Paddy Field in Northeast Thailand .43

pp ER ID: 077

Re ional difference in the citizen's consciousness of water resources : .44

SE SION 0 Environmental Sustainability in Paddy Irrigation and DrainagePAPER ID: 001

Car asian experience be transferred to Africa? -Lessons learned from drafting a rice production

me ual in Africa .45

PA ER ID: 002

Wa 'r quality constituents export from paddy field in Southern Korea .46

PAl ER ID: 004

Ml . regression analysis of water quality characteristics in lowland paddy fields .47

VIII

PAPER ID: 006

Evalcation of field measurements and estimated rice crop water requirements .48

PAPER ID: 007

Effect of rice straw mat mulch and soil amendments on runoff under

laboratory rainfall simulations .49

PAPER ID: 009

Runoff and NPS pollution discharge characteristics from sloping upland fields in Korea 16 50

PAPER ID: 011

Aqrlcuitura: infrastructure database establishment for vulnerability assessment according to climate

chan e in Korea 51

PAPE RID: 013

Genr ic diversities and population structures of small freshwater fishes in

Mekong River basin 52

PAP[R ID: 014

Char cteristics of drainage water quality and loading from paddy field under cyclic irrigation and its

mancqement options 53

PAP RID: 017

Estin ating regional total phosphate concentration in a river basin through the NARX network ...54

PAP RID: 021

Deve opment of the world atlas of irrigated agriculture for sustainability science 55

PAP RID: 022

Influ ntial factors in determining the timing of transplanting lowland rice:

case study in Lao PDR 56

PAP RID: 027

Hab t potential maps of three frog species for paddy field areas of

the I 'ddle Sakura River basin, Japan 57

PAP RID: 040

Cha --nges in the decontamination of radioactive cesium of Fukushima: a rural planning

pers ective 58

PAF RID: 058

Soil acro Nutrient (N, P, K) during Growth Stages under Conventional and SRI

(Sy_m of Rice Intensification) Practices in Tropical Soil. 59

PA RID: 069

A si lyon the reason why the reported yields of the System of

Rice ntensification (SRI) are so widespread 60

PA RID: 072

Mal xing and Analyzing of Soil Water Pollutant Loads in Greenhouse 61

IX

IN OF AUTHO S 62

R Y L IRRIGATIO DEPARTMENT (PAKKRED) MAP 63

x

PA\\ EES 2012 International Conferencehe lenges . Water & Envrronmental Management In Monsoon ASIa

'7-2'-1 member 20 2. Thailand

Optimizing on-flooded Irrigation Regime under System of RiceIntensification Crop Management using Genetic Algorithms

Chusnul Arif. Budi Indra Setiawan, Masaru Mizoguchi and Ryoichi Doi

Abstract: In this study, an optimal non-floodedIrrigation regime that maximizes both the yield andwater productivity of System of Rice Intensification(SRI) crop management was simulated by geneticalgorithms (GAs) model. Here, the field was classifiedinto wet (\\I). medium (M) and dry (D) conditions ineach growth stage, namely initial, crop development,mid-season and late season stages according to the soilmoisture level. The simulation was performed based onthe identification process according to the empirical dataduring three cropping seasons. As the results, theoptimal combination was 0.622 (W), 0.563 (W), 0.522(M), and 0.350 crnvcrrr' (D) for initial, cropdevelopment, mid-season and late season growth stages,respectively. The wet conditions in the initial and cropdevelopment growth stages should be achieved toprovide enough water for the plant to develop root, stemand tiller in the vegetative stage, and then the field canbe drained into medium condition with the irrigationthreshold of field capacity to avoid spikelet fertility inmid-season stage and finally, let the field dry to savemore water in the late season stage. By this scenario, itwas simulated that the yield can be increased up to6.33% and water productivity up to 25.09% with savingwater up to 12.71% compared to the empirical data.

Keywords: system of rice intensification (SRI), non-flooded irrigation, crop productivity, water productivity,genetic algorithms.

Chusnul Arif «(8])Department of Civil and Environmental Engineering,Bogor Agricultural University (IPB), Bogor, IndonesiaDepartment of Global Agricultural Sciences, theUniversity of Tokyo, JapanPhone: (62)-251-8627225, (81)-3-5841-1606Fax: (62)-251-8627225, (81)-3-5841-1606Email: chusnul_ar@yahoo.com.chusnutarif@ipb.ac.id

Budi Indra SetiawanDepartment of Civil and Environmental Engineering,Bogor Agricultural University (IPB). Bogor, IndonesiaEmail: budindra@ipb.ac.id

Masaru Mizoguchi and Ryoichi OoiDepartment of Global Agricultural Sciences, theUniversity of Tokyo, JapanEmail: amizo@mail.ecc.u-tokyo.ac.jp;aroird@mail.ecc.u-tokyo.ac.jp

Introduction

Recently, the scarcities of water resources andcompetition for their use have made water saving themain challenge in maintaining the sustainability of ricefarming. Therefore, water saving technology becomesone of the priorities in rice research (Barker et al. 2000).From the previous findings, rice is highly possibleproduced under water saving technology with System ofRice Intensification (SRI) crop management in whichcontinuous flooded irrigation is not essential anymore togain high yield and biomass production (Zhao • al.201 I; Sato et al. 201 I; Lin et al. 201 I).SRI is well-known as a set crop management practicesfor raising the productivity of irrigated rice by changingthe management of plants, soil, water and nutrients.Although some critics were dismissed to the SRI(Sinclair and Cassman 2004; Sheehy et al. 2004;Oobermann 2004), however, its benefits have beenvalidated in 42 countries of Asia, Africa and LatinAmerica (Uphoff et al. 2011). In the SRI paddy field,non-flooded irrigation is applied in which the field isallowed dry during particular time instead of keepingthem continuously flooded, a practice called alternatewetting and drying irrigation (AWOl) (Van der Hoek etal. 2001).Many experiments have been conducted by comparingcontinuous flooded and non-flooded irrigations underSRI crop management (Choi et al. 2012; Zhao et al.2011; Sato et al. 2011; Hameed et al. 2011; Barison andUphoff 2011; Chapagain and Yamaji 2010). Waterproductivity can be raised by saving water significantly,as reported in studies that provide data for differentcountries, e.g., 28% in Japan (Chapagain and Yamaji2010),40% in Eastern Indonesia (Sato et al. 2011), and38.5% in Iraq (Hameed et al. 2011). Also by SRI cropmanagement, the land productivity raised more thandouble in Madagascar (Barison and Uphoff 2011), 78%in Eastern Indonesia (Sato et al. 2011), 65% inAfghanistan (Thomas and Ramzi 2011), 42% in Iraq(Hameed et al. 2011), and 11.3% in China (Lin et al.2011). However, the optimal wet and dry levels(represented by soil moisture) in each growth stage isstill unclear because there is lack information study onoptimizing water irrigation regime under SRI cropmanagement. Thus, the current study was undertaken tofind optimal soil moisture level in each growth stage tomaximize both land and water productivity duringcultivation period.

P \ \\ Ef S 2012 International Conference( tal enges of \\ ater & Environmental Management In Monsoon ASia'7-11) ver-ber 2012 Thailand

In the irrigation planning model, it has always been adifficult problem to find optimal or near optimalsolutions with traditionally dynamic optimizationsmethod because of multi-factors, uncertainty and non-lmearity in the model (Zhang et al. 2008). Also,traditional optimizations have limitations in findingglobal optimization results for a complex irrigationplanning problem because they search from point topoint (Kuo et al. 2000). Thus, genetic algorithms (GAs)model proposes global optimization search with manyremarkable characteristic by searching the entirepopulation instead of moving from one point to the nextas the traditional methods (Kuo et al. 2000).GAs model has ability to rapidly search a global optimalvalue of a complex objective function using a multi-point search procedure involving crossover andmutation processes (Goldberg 1989). GAs model differsfrom traditional optimization and other searchprocedures in the following ways: (l) GAs works with acoding of the parameter set, not the parametersthemselves, (2) GAs searches from population of points,not single point, (3) GAs uses objective functioninformation, not derivatives or other auxiliaryknowledge, and (4) GAs uses probabilistic transitionrules, not deterministic rules (Goldberg 1989). GAsmodel has been applied to several irrigation planningapplications (Zhang et al. 2008; Ward law and Bhaktikul2004; Raju and Kumar 2004; Kuo et al. 2000).However, optimizing non-flooded irrigation regime byfinding optimal soil moisture in each growth stage hasnot yet been achieved particularly under SRI cropmanagement.Therefore, the objective of current study was to findoptimal soil moisture level in each growth stage ofpaddy rice under SRI crop management using GAsmodel to maximize both land and water productivity.

Materials and Methods

Field ExperimentsThe optimization process was carried based the fieldexperiment in the experimental paddy field in theNagrak Organics SRI Center (NOSC), Sukabumi, WestJava, Indonesia located at 06°50'43" S and 106°48'20"E, at an altitude of 536 m above mean sea level (Fig. 1)during three cropping seasons (Table 1).

Table I. Cultivation period of each cropping season

Period Harvesting dale SeasonPlanting dateFirstSecondThird

14 October 20 I020 August 201122 March 2012

8 February 20 I I15 December 20115July2012

WetDry- WetWet- Dry

There were four plots and each plot was planted with thevariety of rice tOryza saliva L), Sintanur using thefollowing SRI crop management: single planting ofyoung seedlings spaced at 30 cm x 30 cm, applying anorganic fertilizer at 1 kg/m ' in the land preparation, butno chemical fertilizer. The weeding was performedevery 10 days in the period between 10 and 40 days

after transplantation supplying local indigenousmicroorganism to enhance biological activity in the soils(Uphoff et al. 2011)

Source: earth.google.com (2012)Fig. I Experimental field location 111 West Java.Indonesia.Each plot was irrigated under different water irrigationregimes according to the growth stages, namely, initial,crop development, mid-season and late season stages(Vu et al. 2005; Tyagi et al. 2000; Alien et al. 1998;Mohan and Arumugam 1994). Non-flooded conditionwas applied in all regimes and in each growth stage thesoil was classified into three conditions i.e. wet (W),medium (M) or dry (D) to realize the soil moisturedescribed by changes in soil suction head (i.e. pF value)(Arif et al. 2011 a) presented in Fig. 2. The wet conditionwas achieved when pF value was between 0 and 1.6which was the air entry value for this soil. The mediumcondition was achieved when pF value was between 1.6and 2.54 which was the field capacity value. When thesoil was drier than the medium condition, the conditionwas regarded as the dry condition.

5.0

4.0

3.0u,Cl,

20

LO

LOVolumetric water content (cm'jcm')L_ _ _

Fig. 2 Classification of soil moisture condition duringthe cultivation period.

P'\ WEES 2012 International Conferencel 'iallcngcs of Water & Env ronmental Management in Monsoon ASIa27-29 lovernber 2012 fhailand

oil moisture and meteorological data consisting of airtemperat re, wind speed, relative humidity, solarradiation and precipitation were collected by a quasi realtin-e monitoring system (Arif et al. 20 I1 b). Here,meteorological data were used to calculate referenceevapotranspiration (ETo) based on the FAO Penman-Monteith model (Alien et al 1998). Then, usingmonitoring data. total irrigation, crop evapotranspirationwere estimated b) Excel Solver (Arif et al. 2012a).

Modeling Approach

Identification procedureTo determine optimal soil moisture level in each growthstage to maximize yield and water productivity, therelationship between their values and the yield as well aswater productivity should be identified firstly based onempirical data. Since there was no mathematical modelbetween them, identification process was carried outgradually and its procedure as follow:I. Yield as function of plant height and tiller

numbers/hillThe yield has positive correlation to plant growthrepresented by plant height and tiller numbers/hill.Therefore, we used multiple linear regressions to showthe correlation between the yield and plant height aswell as tiller numbers.hill, The basic formula was givenas follow:Y=aPH+bTH+e (I)Where Y is yield (ton/ha), PH is plant height (cm), THis tiller numbers/hill, a, b, c are coefficients of plantheight, tiller numbers/hill and the intercept, respectively.Plant height and tiller numbers/hill were measuredmanually every 5 days and we used their average valuesin the end of mid-season stage when maturity time wasstarted to show their correlation to the yield (Alien et al.1998). Here, all coefficients in equation I weredetermined empirically according field experiments bymultiple linear regressions.2. Plant height and tiller numbers/hill as function of

soil moisturePlant height and tiller numbers/hill are affected by soilcondition represented by the soil moisture level in eachgrowth stage. In fact, it is difficult to identify therelationship between soil moisture and plant growth bymathematical model because it is characterized by non-linearity. Thus, we implemented neural networks modelto show its correlation since neural networks modeldeals with complex system such as in agriculturalsystem (Hashimoto 1997). The model consisted threelayers, i.e. input, hidden and output layers. Soil moisturein the initial (SM 1), crop development (SM2), mid-season (SM3) and late season (SM4) were used as in theinputs, while plant height (PH) and tiller numbers/hill(TH) as the outputs (Fig.3).3. Water productivity with respect total water inputSince we focused to find the minimum water input asmuch as possible, thus we defined water productivity

with respect total water input (Bouman et al. 2005) withthe following equation:

Wp= yL(! +P)

Where I is total irrigation (mm) that affected by the soilmoisture level in each growth stage, P is precipitation(mm) and WP is water productivity (g grain/kg water).

Input layer Hidden layer Output layer

(2)

SMl

TH

5M2 PH

5M3

5M4Fig. 3 Structure of neural networks model to identifyplant height and tiller numberslhill as affected b) soilmoisture.

Optimization procedureThe objective function can be described as:F(SM 1,SM2,SM2,SM4)= dY +eWP (3)

Maximize F(SMI ,SM2,SM3,SM4)Subject to:

0.586~SMI~0.622(ern3fern3) (4)

0.563 ~ SM2 ~ 0.593 (cm 3/cm ') (5)

0.455 ~ SM3 ~ 0.522 (cm3/cm ') (6)

0.350 ~ SM4 ~ 0.505 (cm i/cm ") (7)

Where d and e are weights for the yield (Y) and waterproductivity (WP) and their values were 0.6 and 0.4,respectively. Since both Y and WP have different unit,thus their values were normalized using their maximumand minimum values based on empirical data. Here,GAs model searched optimal combination of SM I,SM2, SM3, and SM4 with their interval (minimum andmaximum soil moisture in each growth stage) accordingexperiment data to maximize the objective function bymulti-point searching procedure.In order to employ GAs model, some parameters such asindividual, population, fitness function and operators ofGAs should be defined firstly as follows:I. Definition of individual

An individual represented a candidate for theoptimal solution that consisting of particular valuesSM I, SM2, SM3 and SM4. Meanwhile a setindividual was called population. An individual wascoded as six-bit binary strings as illustrated asfollow:Individual = SMI, SM2, SM3, SM4

= 000101, 001000, 001100, 100001(binary string)

PA\\ EES 2012 International ConferenceCha enge of Water & Em ironmental Management In Monsoon Asia:7-29 ove+ber 2012 Thailand

0.598, 0.567, 0.468, 0.431 (decimalvalues in cm'zcrrr')

2 Fitness functionFitness function is an indicator to show the qualityof an individual. All individuals in a populationwere evaluated based on their performances inwhich the higher fitness function, the better abilityto survive In this problem, fitness function wasgiven same as an objective function (equation 3).

3. Operators of GAs modelThe main operators were crossover and mutation.Crossover combined features from two individualsbased on crossover rate (Pc). It is operated byswapping corresponding component in the binarystrings represented an individual. Mutation invertedone or more bit binary string (also called gene) ineach indiv idual from 0 to 1 or 1 to 0 based onmutation rate (Pm). Here, Pc and Pm were 60% and5%, respectively.

K=OInitial population

No

New population

ote:GN = generation numbers; 200

Fig. 4 Searching procedure of GAs model in the currentstudy.We adopted the searching procedure as developedpreviously (Suhardiyanto et al. 2009) and could befurther explained as follow:I. An initial population consisting of ten individuals

were generated randomly as the first generation2. Fitness function was calculated to show the

performance of each individual3. The performance of each individual was evaluated

by sorting maximum to minimum fitness value.Here, 60% of all individuals with the highestperformance were selected to next step, anotherfitness values were eliminated

4. Crossover and mutation operations were applied tothe selected individual based on Pc and Pm rates

5. Then, new population was created. Here, 60% ofselected individuals previous generation werecompared to the new population. We implementedthe elitist strategy to find global optimum by sortingthe fitness value and then selected ten individualswith highest performance as the next generation.

6. Steps 2 to 5 were repeated until the requiredgeneration numbers achieved. The optimal valueswere given as an individual with highest fitnesswhen the result was convergent

We developed both neural networks and GAs models inMicrosoft Excel 2007 with Visual Basic Applicationwith our own codes.

Results and Discussions

Correlation between plant growth, yield and soilmoistureBased on empirical data, it was cleared observed thatboth plant height and tiller numbers/hill have positivecorrelation to the yield with R2 of 0.86 and 0.89,respectively (Fig.5). These results were consistent withprevious findings that higher plant height with moretiller numbers/hill promoted more yield and vice versa(Kumar et al. 2012; Zeng et al. 2003). Therefore, thecorrelation between the yield and both plant height andtiller numbers/hill can be presented well by multiplelinear regressions with R2 of 0.94 (Fig.6). Commonlythe yield of rice is highly dependent upon the number offertile tiller numbers/hill, thus more tiller numbers/hillhave high probability to produce more fertile tiller(Zeng et al. 2003).

a12

v 02329)1-1853110 ""'WI'" s'>

~ 6"0-.;;;:

4

2

050 70 90 110 130 ISO

Plant height [crnl

b12

y 0.536,· 6.742110 "'/'" 8

""c.g6

-o-.;>: 4

1

010 IS 20 25 30 3S 40

Tiller numb ers /hdl

Fig. 5 Correlation between plant height, tillernumbers/hill and yield: a) plant height vs yield, b) tillernumbers/hill vs yield

PA\\ EES 2012 International ConferenceCha.lenges of Water & Environmental Management In Monsoon ASIa27-2'1. ovember 2012 Thailand

12--; y a.107PH t 0 323TH- 13.03

10 R- 0.9355 0cg

8'0-a:;s,'0 6

'"toe: 4:;;w

2

00 4 6 8 10 12

Observed yield(ton/ha)

Fig. 6 Multiple linear regression plant height and tillernumbers/hill on the yield.The yield also has positive correlation to the cropevapotranspiration (ET) with R~ of 0.94 as presented inFig. 7. It was indicated that higher cropevapotranspiration promoted more evaporation processfrom the soil and transpiration process from the plant,thus more yield was achieved as well as biomassproduction (Shih 1987). The linear correlation betweencrop evapotranspiration and the yield is not only foundfor paddy rice (Shih et al. 1983), but also other cropssuch as corn (Ko and Piccinni 2009), cotton and wheat(Jalota et al. 2006). Therefore, crop evapotranspirationis usually used to estimate the yield (Shih 1987).

12.0

10.0

~ 8.0

?,g,6.0

'0Qj> 40

20

00310

Y = 0.l2ET - 33.9R' = 0.94

• • Ydata

-Ymodel

350 360 370320 330 340

Crop evaporranspuatioo (mm)

Fig. 7 Correlation between yield and cropevapotranspiration.Crop evapotranspiration is vital for irrigation schedulingand water resource allocation, management andplanning (Jensen et al. 1990). Also, it is a maincomponent of water consumption in paddy fields, thusits rate is depend on water availability in the fieldrepresented by soil moisture. In other words, plantgrowth and the yield were clearly affected by soilmoisture level in each growth stage (Arif et al. 2012b;Anbumozhi et al. 1998). However, there is nomathematical model was found showed the relationbetween soil moisture level in each growth stage andplant growth as well as the yield.Here, we used neural networks model to estimate plantgrowth represented by plant height and tiller

numbers/hill as affected by soil moisture level in eachgrowth stage. Fig. 8 shows the comparisons betweenobserved and estimated values of both plant height andtiller numbers/hill by neural networks model. It can beseen that the estimated values were closely related to theobserved values with R2 of 0.82 and 0.92 for estimationof plant height and tiller numbers/hill, respectively. Thismeans that more than 80% of observed values can beexplained linearly by neural networks model. Therefore,the results suggest that a reliable simulation model byneural networks could be obtained to estimate plantheight and tiller numbers/hill.

a130

E~ 120~00ijj

110s:C'"Ci 100'0B'"~ 90w

8080

y = 0.8008x + 21.109R' 0.8181

o

o

90 100 110 120

Observed plant height (cm)

130

35

35b

i: Y 0.9231> + 1.8232'?;;- R' = 0.9243Q:; 30.D

E:JC

~25

'0Bro 20E~w

1515 20 25

o

30

Observed tiller numbers/hill

Fig. 8 Observed and estimated plant growth as affectedsoil moisture level by neural networks model: a)Observed vs estimated plant height, b) Observed vsestimated tiller numbers/hill.

Optimal soil moisture levels by GeneticAlgorithmsFig. 9 shows the evolution curves of fitness values(equation 3) between their maximum, average andminimum values in each generation. All of valuesincreased sharply from the first to the tenth generation,and then increased gradually until the fortiethgeneration. After the fortieth generation, the all fitnessvalues were convergent until the end of generation andtheir values were 0.68. This means that global maximumvalue was obtained because all of maximum, averageand minimum values were same.Fig. 10 shows the evolution curves of soil moisture levelin each growth stage in obtaining fitness values

PA" EES 2012 International ConferenceChallenges of Water & Environmental Management m Monsoon Asia27-29 November 2012, Thailand

presented in Fig. 9. SM I and S'vI2 were convergentfaster than others growth stages in which their valueswe-e reached before the tenth generation. Meanwhile,SM3 was convergent latest in the fortieth generation,thus fitness value was also convergent started at thismoment (Fig. 9). This means that the optimal soilmoisture level in each growth stage that maximizes theyield and water productivity was obtained from themodel simulation based on the GAs procedure shown inFig. 4 after the fortieth generation.

70 r---------------------------------,./

',5 ,", I

/",.,,

.60 ,,I

f'i'-r

5'. i,",,

0.500 '0 20

- - - Mlllfitne:S5

'0 40 50 so 70 80 ?O 100

General 0" numbers

Fig. 9 Evolution curves in searching for a maximal valueof itness function.

0700 r---------------------------------,.....•. . . .

-------------------,--------!'il~c.seo

0.300 '--------------------------'

-..,,• M • ~. ~ •••••••••••• _ ••••••• ~ •••• __ •• ••••••••••••• _

N W ~ ~ m m m % ~ene at on "U'TI~'S

Fig. I0 Evolution curves in searching the optimal values ofsoil moisture in each growth stage.Table 2 shows the optimal soil moisture level in eachgrowth stage obtained from the GAs simulation, whileFig. II shows the optimal output simulated by GAsmodel. Four irrigation regimes with some combinationsof soil moisture level from the field measurements arealso represented in the table with the same precipitationduring cropping season as the comparison. The optimalcombination of soil moisture level in each growth stageobtained in this study was 0.622 (wet), 0.563 (wet),0.522 (medium), and 0.350 crrr'zcrn:' (dry) for initial,crop development, mid-season and late season growthstages, respectively. By this scenario, it was simulatedthat the yield can be increased up to 6.33% and waterproductivity up to 25.09% with saving water up to12.71% compared to the first regime (as base line).From this simulation, it can be concluded that during thefirst to the second stages keeping the field in the wetcondition is important to fulfill plant water requirement.

In the SRI, avoid continuous flooding is one of the theirelements because rice plants cannot grow best underhypoxic soil conditions, thus plants should be given justenough water at saturated condition to meet theirrequirement for root, stem and tiller development(Uphoff et al. 2011). Then, the field can be drained intomedium condition in the third stage when plantsfocusing on reproductive stage (flowering and panicledevelopment). The medium condition with the thresholdof irrigation was reduced to the field capacity (pF 2.54)is important in developing aerobic condition to avoidspikelet sterility particularly around flowering time(Bouman et al. 2005). Finally, the field should bedrained into dry condition in the last stage to save thewater input as reported previous studies (Uphoff et al.2011; Doorenbos and Kassam 1979; Zawawi et al.2010).

a12.0

10.0

'"-E. 8.0c:~ 6.0'0Oi;;:

4.0

2.0

0.0

10.6310.00 --r- 9.38 9.38

r- 8.75 -r-

Regime 1 Regime 2 Regime 3 Regime 4 OptimalRegime

b141.2

~~ 10JP 0.8?;;;OD 06:Ec,

0.4~0.2

0.0

125112 1.11 1.14 r-

- -- 102 ---- []VJitttr

producttvitv 1&grain! kR water]

Regime 1 Regimt> 2 Regime 3 R(>gIl'fH:04 OptmalRegIme

Fig. II Optimal the yield and water productivitysimulated by GAs model and their comparison to theempirical data.

Table Z Optimal soil moisture level in eacbg;ro\'\th stagf' andits ccmpansce to the empmcal data

Field :Ex.pmrnenrs GAs optmuzerRegime Regsne Regime Regime (Jptm,.,

Components 1 1 3 4 Regone Ccndmon

Soil moisture {cm3!cm3)

lrutial (SM 1) 0.622 C 602 0.611 0586 O.6:!2 W<t

Crop development (SM.:!) 059" 058; 0.593 0'63 0.563 Wet

Mid- season(SM3) 0.522 o JgS 04-1 0455 o 5~~ Mednan

Late st:a..son{S_M4) o 50~ CAOI 0.~j6 0350 0350 DryYield (ton ha) 1000 938 S -5 938 1063

Tctalcrigaricn (mm) 343 295 305 --, :99

Tot alpreciprtation (mm) 551 551 551 ~51 551Waterprcducrrviry(g gramkg+waterj 112 1.11 102 114 1 1~W.ters3;\'ing{*1.) BueliDe 13.S6+'b 11Cl% !O_65~G 1':.-1°~

P.c\w EES 2012 International ConferenceChallenges of Water & Environmental Management in Monsoon Asia27-29 November 2012 Thailand

Conclusions

In this study, optimal combination of soil moisture levelin each growth stage that maximizes both the yield andwater productivity of SRI crop management wassimulated by GAs model. The simulation was performedbased on the identification process according to theempirical data during three cropping seasons. As theresults the optimal combination was 0.622 (wet), 0.563

• 3 1(wet), 0.522 (medium), and 0.350 cm /cm (dry) forinitial. crop development. mid-season and late seasongrowth stages, respectively. The wet conditions in theinitial and crop development growth stages should beachieved to provide enough water for vegetativedevelopment, and then the field can be drained with theirrigation threshold of field capacity to avoid spikeletfertility in mid-season stage and finally, let the field d~yto save more water in the late season stage. By thisscenario it was simulated that the yield can be increasedup to 6.33% and water productivity up to 25.09% ~ithsaving water up to 12.71% compared to the first regime(as base line).

Acknowledgments

We are grateful to the Directorate of Higher Educatio?,Ministry of National Education, Republic of Indonesiafor generous financial support .through gr~nt. ofInternational Research Collaboration and ScientificPublication and the Japan Society for the Promotion ofScience. Also. the study was partially supported byGRE E (Green Network of Excellence) of MEXT inJapan.

References

Alien RG, Pareira LS, Raes D, Smith M (1998) CropEvapotranspiration Guidelines for computingcrop water requirements. FAO - Food andAgriculture Organization of the UnitedNations, Rome

Anbumozhi Y, Yamaji E, Tabuchi T (1998) Rice cropgrowth and yield as influenced by changes inponding water depth, water regime andfertigation level. Agr Water Manage 37(3):241-253

ArifC, Mizoguchi M, Setiawan BI, Doi R (201Ia)Water Management Evaluation of AlternateWetting and Drying Irrigation in Paddy Fieldsby Considering Monitored Soil Moisture. Paperpresented at the Soil Moisture Workshop 2011,Tokyo, 22 December

ArifC, Setiawan BI. Mizoguchi M, Doi R (20 12a)Estimation of Water Balance Components inPaddy Fields under Non-Flooded IrrigationRegimes by using Excel Solver. Journal ofAgronomy II (2):53-59.doi:1 0.3923/ja.20 12.53.59

ArifC. Setiawan BI, Mizoguchi M, Doi R, Saptomo SK,Ardiansyah., Ito T (20 11b) Field etworkSystem to Monitor Paddy Fields in the Systemof Rice Intensification in Indonesia. Paperpresented at the CIGR 20 I1 conference onSustainable Bioproduction, Tokyo, 19-23September

ArifC, Setiawan BI, Sofiyuddin HA, MartiefLM.Mizoguchi M, Doi R (20 12b) Estimating CropCoefficient in Intermittent Irrigation PaddyFields Using Excel Solver. Rice Science 19(2): 143-152. doi: 10.10 16/s 1672-6308( I2)60033-x

Barison J, UphoffN (2011) Rice yield and its relation toroot growth and nutrient-use efficiency underSRI and conventional cultivation: an evaluationin Madagascar. Paddy Water Environ 9 (I ):65-78. doi: I0.1007/s I0333-0 I 0-0229-z

Barker R, Dawe D, Tuong TP, Bhuiyan SI, Guerra LC(2000) The outlock for water resources in theyear 2020: challenges for research on watermanagement in rice production. InternationalRice Commission ewsletter 49:7-21

Bouman BAM, S.Peng., Castaneda AR, Yisperas RM(2005) Yield and water use of irrigated tropicalaerobic rice systems. Agr Water Manage74:87-105. doi: 10. 10 I 6/j.agwat.2004. I 1.007

Chapagain T, Yamaji E (2010) The effects of irrigationmethod, age of seedling and spacing on cropperformance, productivity and water-wise riceproduction in Japan. Paddy Water Environ 8(I ):81-90. doi: I0.1007/s 10333-009-0187-5

Choi JD, Park WJ, Park KW, Lim KJ (2012) Feasibilityof SRI methods for red uction of irrigation andNPS pollution in Korea. Paddy Water Environpublished on line by Springerlink Feb. 9.doi:10.1007/s10333-012-0311-9

Dobermann A (2004) A critical assessment of thesystem of rice intensification (SRI). Agr Syst79 (3):261-281. doi:10.1016/S0308-521 x(03)00087-8

Doorenbos J, Kassam AH (1979) Yield response towater. FAO Irrigation and Drainage Paper 33.FAO, Rome

Goldberg DE (1989) Genetic algorithms in searchoptimization and machine learning. Addison-Wesley, Reading, Massachusetts

Hameed KA, Mosa AKJ, Jaber FA (2011) Irrigationwater reduction using System of RiceIntensification compared with conventionalcultivation methods in Iraq. Paddy WaterEnviron 9 (1):121-127. doi:10.1007/sI0333-010-0243-1

Hashimoto Y (1997) Applications of artificial neuralnetworks and genetic algorithms to agriculturalsystems. Comput Electron Agr 18 (2-3):71-72

Jalota SK, Sood A, Chahal GBS, Choudhury BU (2?06)Crop water productivity of cotton (Gossypiumhirsutum L.)-wheat (Triticum aestivum L.)

PAWEES 2012 International ConferenceChallenges of Water & Environmental Management In Monsoon Asia27-29 member 2012. Ihailand

system as influenced by deficit irrigation, soiltexture and precipitation. Agr Water Manage84 (1-2):137-146. doi:DOI10.1016lj.agwat.2006.02.003

Jensen ME, Burman RD, Alien RG (1990)Evapotranspiration and Irrigation WaterRequirements. The American Society of CivilEngineers. ew York,

Ko J, Piccinni G (2009) Corn yield responses under cropevapotranspiration-based irrigationmanagement. Agr Water Manage 96 (5):799-808. doi:DOI 10.1 016/j.agwat.2008.1 O.OlD

Kumar KYK, Yellareddygari SK, Reddy MS, KloepperJW, Lawrence KS, Zhou XG, Sudini H, GrothDE. Raju SK, MillerMILLER7 ME (2012)Efficacy of Bacillus subtilis MBI 600 AgainstSheath Blight Caused by Rhizoctonia solaniand on Growth and Yield of Rice. Rice Science19 (1):55-63. doi:10.1016/S1672-6308( 12)60021-3,

Kuo SF. Merkley GP. Liu CW (2000) Decision supportfor irrigation project planning using a geneticalgorithm. Agr Water Manage 45 (3):243-266

Lin XQ, Zhu OF, Lin XJ (20 11) Effects of watermanagement and organic fertilization with SRIcrop practices on hybrid rice performance andrhizosphere dynamics. Paddy Water Environ 9(I ):33-39. doi: I0.1 007's 10333-0 I0-0238-y

Mohan S, Arumugam "'l (1994) Irrigation cropcoefficient for lowland rice. Irrigation andDrainage Systems 8: 159-176.doi: 10.10071BF00881 016

Raju KS, Kumar ON (2004) Irrigation planning usingGenetic Algorithms. Water Resour Manag 18(2): 163-176

Sato S, Yamaji E, Kuroda T (2011) Strategies andengineering adaptions to disseminate SRImethods in large-scale irrigation systems inEastern Indonesia. Paddy Water Environ 9(l ):79-88. doi: I 0.1 007/s 10333-0 I0-0242-2

Sheehy JE, Peng S, Dobermann A, Mitchell PL, FerrerA, Yang JC, Zou YB, Zhong XH, Huang JL(2004) Fantastic yields in the system of riceintensification: fact or fallacy? Field Crop Res88 (1): 1-8. doi: I0.1 0 16/j .fcr.2003.12.006

Shih SF (1987) Using crop yield and evapotranspirationrelations for regional water requirementestimation Water Resour Bull 23 (3):435-442.doi:10.lllllj.1752-1688.1987.tb00821.x

Shih SF, Rahi GS, Snyder GH, Harrison DS, SmajstrlaAG (1983) Rice Yield, Biomass, and Leaf-Area Related to Evapo- Transpiration. T Asae26 (5):1458-1464

Sinclair TR, Cassman KG (2004) Agronomic UFOs.Field Crop Res 88 (I ):9-10.doi: I0.1 0 16/j.fcr.2004.0 1.00 I

Suhardiyanto H, Arif C, Setiawan BI (2009)Optimization of EC Values of NutrientSolution for Tomato Fruits Quality in

Hydroponics System Using Artificial NeuralNetwork and Genetic Algorithms. ITB J Sci 41A (I ):38-49. doi: 10.5614/itbj.sci.2009.41.1.3

Thomas V, Ramzi AM (20 11) SRI contributions to riceproduction dealing with water managementconstraints in northeastern Afghanistan. PaddyWater Environ 9 (I): I0 I-I 09.doi: I 0.1 007/s I0333-0 I0-0228-0

Tyagi NK, Sharma DK, Luthra SK (2000)Determination of evapotranspiration and cropcoefficients of rice and sunflower withIysimeter. Agr Water Manage 45 (I ):41-54.doi: I 0.1 016/S0378-3774(99)00071-2

UphoffN, Kassam A, Harwood R (2011) SRI as amethodology for raising crop and waterproductivity: productive adaptations in riceagronomy and irrigation water management.Paddy Water Environ 9:3-11.doi: 10.1 007/s I0333-0 I0-0224-4

Van der Hoek W, Sakthivadivel R, Renshaw M, SilverJB, Birley MH, Konradsen F (200 I) Alternatewet/dry irrigation in rice cultivation: a praticalway to save water and control malaria andJapanese encephalitis? Research Report 47.International Water Management Institute,Colombo, Sri Lanka

Vu SH, Watanabe H, Takagi K (2005) Application ofFAO-56 for evaluating evapotranspiration insimulation of pollutant runoff from paddy ricefield in Japan. Agr Water Manage 76 (3): 195-210. doi: I 0.1 0 16/j.agwat.2005.0 1.012

Wardlaw R, Bhaktikul K (2004) Application of geneticalgorithms for irrigation water scheduling. lrrigDrain 53 (4):397-414. doi:Doi 10.1002IIrd.121

Zawawi MAM, Mustapha S, Puasa Z (20 I0)Determination of water requirement in a paddyfield at seberang perak rice cultivation area.Journal - The institution of Engineers 71(4 ):32-41

Zeng LH, Lesch SM, Grieve CM (2003) Rice growthand yield respond to changes in water depthand salinity stress. Agr Water Manage 59(I ):67-75

Zhang B, Yuan SQ, Zhang JS, Li H Study of cornoptimization irrigation model by geneticalgorithms. In: JFfP International Federationfor Information Processing, 2008. Springer, pp121-132. doi: 10.1 007/978-0-387-7725 1-6 14

Zhao LM, Wu LH, Wu MY, Li YS (2011) Nutrientuptake and water use efficiency as affected bymodified rice cultivation methods with reducedirrigation. Paddy Water Environ 9 (I ):25-32.doi:10.1007/sI0333-011-0257-3