Air Temperature Variation and Productivity in Bangladesh CERES Rice Model

Ecological Modelling 106 (1998) 201–212

Air temperature variations and rice productivity in Bangladesh:a comparative study of the performance of the YIELD and the

CERES-Rice models

Rezaul Mahmood *

Department of Geography, College of Geosciences, The Uni6ersity of Oklahoma, Norman, OK 73019, USA

Accepted 24 October 1997

Abstract

Potential increase in air temperature due to climatic change and inter-annual climatic variability and its impacts oncrop productivity is of major concern to crop scientists. A number of physically-based models have been developedand applied to estimate crop–environment relationships. In the present study the performance of two such models(the YIELD and the CERES-Rice) are discussed. These two models are used to estimate boro rice productivity undernormal and abnormal climate scenarios in Bangladesh. This study finds that boro rice productivity at Mymensinghpredicted by the YIELD is higher than the prediction by the CERES-Rice. Productivity estimates for Barisal by thesetwo models are almost identical. Assumptions of non-identical management practices, different soil characterizationprocedures, different methods for calculation of dry matter production by these two models and the range of diurnaltemperature variations played an important role in productivity estimates. The YIELD model predicted the lengthsof the growing season under the normal and abnormal thermal climate conditions and they are to be shorter than thelengths predicted by the CERES-Rice model. The YIELD model’s assumption of higher threshold temperature anda relatively simple relationship between phenology and air temperature has produced such estimations (shortergrowing season). The complex data required by CERES-Rice may be an impediment for its extensive use. If inputdata for the CERES-Rice is not available, the YIELD model can be considered as a possible tool for variousapplications in crop–environment relationships. © 1998 Elsevier Science B.V. All rights reserved.

Keywords: Rice; Crop models; Bangladesh

1. Introduction

Rice is an important staple food for half of theworld’s population. Rice farming employs a sig-

* Corresponding author. Tel: +1 405 3255325; fax: +1 4053256090; e-mail: [email protected]

0304-3800/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved.

PII S 0 304 -3800 (97 )00192 -0

R. Mahmood / Ecological Modelling 106 (1998) 201–212202

nificant number of people in many rice economies.Climatic variability and predicted climatic changeis of major concern to rice crop scientists becauseof its potential threat to rice productivity and theassociated impacts on the socio-economic struc-ture of rice growing countries. A number of modelsimulations and field experiments have been car-ried out to show the impacts of climatic variabilityand climatic change on the rice productivity(Baker et al., 1990; Jensen, 1990; Rosenzweig andParry, 1994). In the present paper, performance oftwo rice growth simulation models YIELD andCERES-Rice are discussed. Both models wereapplied to sites in Bangladesh under the samethermal climate change scenarios. The YIELDmodel was applied to 12 meteorological stationslocated in the major boro rice growing regions toestimate yield, length of growing season, andevapotranspiration (Mahmood and Hayes, 1995),phenology and irrigation requirements (Mah-mood, 1996, 1997), and changes in the croppingpattern (Mahmood, 1998) under abnormal ther-mal climatic conditions. The CERES-Rice modelwas applied to two of these 12 meteorologicalstations to estimate the length of boro rice grow-ing season and yield (Karim et al., 1994) under thesame thermal climate scenarios. This paper pre-sents results for the two regions (Mymensingh andBarisal) where both of the models were applied.These results are accompanied by discussions onthe performance of the YIELD and CERES-Ricemodels.

2. Model descriptions

2.1. The YIELD model

The YIELD model is able to simulate andpredict seasonal crop yield, crop water use, lengthof growing season and related growth characteris-tics for 11 crops including rice. Its data require-ments are modest (Table 1) and it allows flexibility(via user-selected options from programmed de-fault scenarios) with many built-in cause and ef-fect links. The model evolved from a crop waterbalance model developed by Burt et al. (Burt et al.1980; Burt et al., 1981) and has been validated and

sensitivity-tested for several major agriculturalcrops including rice (Burt et al., 1980, 1981; Hayeset al., 1982a,b; Terjung et al., 1984b,c). It wassuccessfully applied to several regional studies, forexample, in China (Terjung et al., 1983,1984a,b,c,d,e,f,g, 1985, 1989; Todhunter et al.,1989), in Korea (Terjung et al., 1985), in Califor-nia (Hayes, 1986), and in Bangladesh (Mahmood,1993, 1996, 1997, 1998; Mahmood and Hayes,1995) to estimate rice productivity and its relation-ship with various growth related parameters.

The YIELD model is based on the methods ofDoorenbos and Pruitt (1977) and Doorenbos andKassam (1979); Doorenbos and Kassam suggestedthat it is possible to determine relationships be-tween crop yield and water supply if crop waterdeficits and actual yield data are simultaneouslyderived. Water stress adversely affects both evapo-transpiration (ET) and yield. A ratio of actual ETto potential ET, called relative ET deficit, is

Table 1Selected data requirements for the YIELD model (modifiedfrom Hayes et al., 1982a)

Input data

ClimatologicalMean monthly maximum and minimum air temperatureMean monthly maximum and minimum relativehumidityMean monthly daily solar radiationMean monthly daily wind speedMean monthly precipitationMean monthly cloud cover

Pedological and hydrologicalSoil typePercolation rateDepth of ground water table

AgronomicField sizeBase temperatureSlope of the fieldIrrigation efficiencyRoot depthHarvest indexMaximum tolerable soil salinity

Station informationLatitudeLongitudeElevation

R. Mahmood / Ecological Modelling 106 (1998) 201–212 203

used to calculate this stress. This ratio is related toanother index known as relative yield decrease,which is the ratio of actual yield (YA) to maxi-mum yield (YM), through an empirical yield re-sponse factor.

The methodology for derivation of relative ETdeficit has been discussed extensively (Burt et al.,1980, 1981). Penman’s (1948) combination equa-tion for calculation of ET constitutes the basis ofrelative ET deficit. Doorenbos and Kassam (1979)successfully modified this equation for a referencecrop (an extended surface covered with 8–15 cmtall green grass of uniform height, actively grow-ing, completely shading the ground, and an opti-mum supply of water) by sets of tables and graphsto include the effects of crop type, crop-growth-stage, selected site factors, influence of unusualclimatic conditions, crop coefficients that adjustET for specific crops and specific crop growthstages, and soil moisture budget conditions. Theessential contribution of Burt et al. (1980, 1981)and Hayes et al. (1982a,b) was to present a math-ematical model as a research tool for geographic,agronomic, and climatic/environmental changestudies of the human-managed soil–crop–atmo-sphere system. In YIELD, the calculation of ETfor the reference and actual crops include a dailyweather-generating submodel developed byAkima (1970) to interpolate daily values frommonthly climate data.

For each crop and season, a soil water budgetis calculated as a function of several variablesincluding growth-stage-specific estimates of ET,soil water storage, effective precipitation, ground-water contributions, variable root depths, percola-tion losses, and irrigation frequency and amount.The average soil water potential for the root zoneis determined from this calculation. The crop rootzone soil water potential is unique for each sce-nario. At this point, the soil water potential isused for the determination of the relative ETdeficit for each crop.

The yield response factor varies with crop spe-cies, crop growth-stage and temporal sequencingof crop water stress. Doorenbos and Kassam(1979) estimated this factor for many crops andclimates. These empirical relationships are appli-cable to high yielding crop varieties grown in

large fields and under optimal conditions (suffi-cient supply of fertilizer, herbicides, pesticides,and trace minerals). The yield response factor canbe applied to relatively evenly distributed waterstress conditions during a complete growing sea-son of a crop, as well as to intermittent droughtand resultant water stress situations. In the lattersituations, water stress during early stages caninfluence the yield response factors of later stages.YIELD allows previous stresses to provide anameliorating effect on later crop-growth-stages byphysiologically hardening the crop against futurewater stress. The most appropriate yield responsefactor and its associated growth-stage-specific ETdeficit is automatically chosen by the program.

The methodology for calculating relative yielddecrease (YA/YM) has been discussed in Dooren-bos and Kassam (1979). They defined YM as theharvested yield of a high yielding variety (HYV)crop that is well-adapted to the growing condi-tions and grown under optimum management.The agro-ecological zone method of Kassam(1977) is applied for the calculation of maximumor potential yield of rice. It should be noted thata method developed by de Wit (1965) has beenused to compute the average gross dry matterproduction of a standard crop.

The YIELD model adopts a phenology sub-model partially developed from Doorenbos andPruitt (1977) and it assumes four stages of riceplant growth, namely: initial, vegetative, flower-ing/maturing, and harvesting stage. Burt et al.(1980) presented a detailed description of thisprocedure. In short, it is based upon the conceptof cumulative degree-days above a crop-specificthreshold temperature, which is a function oflatitude. The accumulation of a certain amount ofdegree-days of heat marks the end of a growthstage. Thus, in the YIELD model, variations inthermal conditions can influence the estimatedlength of each growth-stages and hence the lengthof the growing season.

2.2. The CERES-Rice model

The CERES-Rice model was developed underthe International Benchmark Sites for Agrotech-nology Transfer (IBSNAT) project (Ritchie et al.,1987; Tsuji et al., 1994) and has been successfully


Table 2Selected data requirements for the CERES-Rice model(modified from Ritchie et al., 1987 and Tsuji et al., 1994)

Input data

Weather dataDaily maximum and minimum air temperatureDaily PrecipitationDaily solar radiation

Pedological-hydrological dataSoil classificationTextureNumber of layers in soil profileSlopePermeabilityDrainageSoil layer depthSoil horizonClay, silt, and sand contentBulk densitySaturated hydraulic conductivity for each soil layerTotal nitrogen for each layerpH of the soil in water for each layerRoot quantity for each layer

AgronomicSowing and transplanting dateRow spacing/seeding depthNumber of plants per hillNumber of plants per square meterAge of seedlingBase temperature to estimate phenological stagesIrrigation amount and frequencyBund heightFloodwater depthFertilizer application dates, amounts

Station information:LatitudeLongitude

the CERES-Rice model requires a quite detailedinput data set (Table 2).

Ritchie et al. (1987) and Tsuji et al. (1994) haveprovided a detailed description of the model. Insummary, the CERES-Rice model assumes ninestages of rice plant growth: pre-sowing, germina-tion, emergence, juvenile, floral induction, head-ing, flowering, grain filling, and harvesting.Completion of these growth stages is determinedby accumulation of degree-days. Potential drymatter production is a function of photosyntheti-cally active radiation (PAR) absorbed by plantcommunities. Beer’s Law has been used to esti-mate this solar radiation absorption. TheCERES-Rice model assumes PAR=50% of in-coming solar radiation and to estimate actual drymatter production it adjusts potential dry matterproduction for thermal stress, water and nitrogendeficiency. Dry matter partitioning depends onthe phenological stage and can be modified bywater and nutrient stress. Allocation of biomassto roots influences root density and the efficiencyof roots in supplying nutrients to the shoot. It hasbeen assumed that the allocation of biomass toroots decreases as the growing season progressesand rice plant becomes mature. It is also pre-sumed that partitioning to root will increase un-der water or nitrogen stress during all of thegrowth stages except during grain filling.

The soil water balance and nitrogen componentof the model can be bypassed when a user as-sumes a non-limiting condition. The model calcu-lates infiltration, runoff, drainage andevpotranspiration to estimate soil water balance.CERES-Rice estimates runoff by using USDA(1972) modified Soil Conservation Service CurveNumber Technique. The difference between dailyprecipitation and runoff provides estimates ofinfiltration. If irrigation is included as an input tothe crop, the model does not estimate runoff butallows all water to infiltrate. To estimate potentialET the model offers options of using the Priestlyand Taylor (1972) method and the FAO-Penmanmethod (Doorenbos and Pruitt, 1977). The FAO-Penman method requires additional data com-pared with the Priestly and Taylor method. Toestimate actual ET the method of Ritchie (1972)has been incorporated in the model.

applied to a number of regional studies (Tongyai,1994; Escano and Buendia, 1994; Seino, 1994;Baer et al., 1994; Jin et al., 1995). The CERES-Rice model estimates yield of irrigated and rain-fed rice, determines duration of growth stages, drymatter production and partitioning, root systemdynamics, effect of soil water and soil nitrogencontent on photosynthesis and photosynthate par-titioning, carbon balance, and water balance. Toestimate these detailed characteristics of growth


The nitrogen submodel estimates the nitrogenrequirement of rice, its supply and uptake. Themodel assumes that nitrogen deficiency adverselyaffects leaf expansion, photosynthesis, and itsconcentration in grain. The final yield, as calcu-lated by the model, depends on the grain weightwhich is a function of grain growth rate andlength of filling period. In the CERES-Ricemodel grain growth rate is assigned a valuecharacteristic of the size classes of rice plant,namely, long, medium, and short. The modelalso assumes that yield is directly proportionalto panicle weight.

3. Model applications and results

3.1. Selection of the scenarios for the modelsapplications

To determine the effect of higher than normalair temperature on the BR3 boro rice develop-ment and production, the YIELD and theCERES-Rice models were applied for normal(Tair) and 2 and 4°C above normal conditions(Tair+2 and Tair+4). These models were ap-plied to scenarios based on historic climatic con-ditions and to changed conditions predicted byGeneral Circulation Models (GCMs) (Mitchell etal., 1990; Karim et al., 1994). Fig. 1(a) and (b)shows the recorded monthly mean maximumand minimum temperatures and extreme maxi-mum and minimum temperatures for My-mensingh and Barisal during the boro ricegrowing season. The Goddard Institute of SpaceStudies (GISS), the Geophysical Fluid DynamicsLaboratory (GFDL), and the UK Meteorologi-cal Office (UKMO) GCMs estimate 3–4°C, 4–6°C, 2–4°C increases in air temperature underclimate change scenarios (Mitchell et al., 1990;Karim et al., 1994) (since various GCMs predict2–6°C increase in temperature, 2–4°C increasesare considered as typical). The scenario buildingprocess for the YIELD model applications con-sider both the recorded extreme thermal condi-tions during boro rice growing season (Fig. 1(a)and (b)) and the increasing air temperature sim-

ulated by the GCMs. The CERES-Rice modelapplications only consider the GCM predic-tions for air temperature changes. The inclusionof the recorded extreme thermal condi-tions along with the GCM-predicted increasein air temperature provided a stronger reason-ing for selection of certain air temperature sce-narios.

It is important to note that the boro is a dryseason rice. During the initial growth stage bororice plants grow under cool and dry conditionswhile the vegetative, flowering, and harvestingperiods are hot and dry. As a result supply ofwater through irrigation is essential for thehealthy growth of boro rice plants. Furthermore,from the meteorological point of view, the bororice growing season is relatively uneventful ex-cept for a few localized thundershowers. How-ever, it is found that temperature stress andresultant crop damage is a significant reason forreductions in boro rice yield. Moreover, supplyof water through irrigation can be optimum andfulfil the crop-water requirements and can elimi-nate water-stress (in this paper the YIELD andthe CERES-Rice models assumed optimum sup-ply of water) while air temperature fluctuationscan notably influence crop productivity. As aresult, this paper investigates the impacts of airtemperature variations (because it is a criticalvariable) on the boro rice.

Furthermore, based on our current under-standing of the monsoonal circulation, it can beinferred that a few degrees of higher air temper-ature would not change the monsoonal circula-tion and make dry boro rice growing seasoninto a wet season. Moreover, the GCMs predic-tions of climate change do not indicate anychange in the timing of the onset of the mon-soon. Zhao et al. (1988) indicated drier soilmoisture conditions will exist in South Asia (in-cluding Bangladesh) during the boro rice grow-ing season under global warming conditions. Insummary, since air temperature is the criticalvariable that influences the dry season boro riceproductivity (when water supply is not a majorissue), inclusion of the variations in the otherweather variables is not necessary.


Fig. 1. Mean monthly maximum and minimum and extreme maximum and minimum temperatures for (a) Mymensingh and (b)Barisal.

4. Results and discussions

The YIELD model’s estimated length of theboro rice growing season for Mymensingh and

Barisal is shorter than the CERES-Rice’s estima-tion (Table 3). Key reasons for differences in theestimations of the length of the growing seasonare due to the assumptions of the threshold tem-


Table 3The CERES-Rice (CR) and the YIELD model predicted boro rice productivity and length of growing season due to air temperaturevariations

BarisalMymensinghTair+

Productivity (ton ha−1) Length of growing seasonProductivity (ton ha−1)Length of growing season(days)(days)

YIELDCRCRCR YIELDYIELD CR YIELD

6.81 7.09 138 1350 6.39 7.86 156 1451201316.232 5.77 6.316.75 150 125

129 1104 4.94 6.16 147 115 5.65 5.66

perature for these two models. The CERES-Rice(Karim et al., 1994) and the YIELD (Mahmoodand Hayes, 1995) assumed 9 and 15°C thresholdtemperature for their model applications, respec-tively. The lower threshold temperature assumedby Karim et al. (1994) resulted in longer growingseasons as compared with the YIELD model esti-mations. Its is important to note that the relation-ship between the length of the growing season andaccumulation of heat is relatively linear for theYIELD model (section 2.1) as compared with theCERES-Rice model. The CERES-Rice modeluses a detailed set of crop-specific genetic coeffi-cients which allows the model to respond to vari-ous non-linear geophysical conditions in a morerealistic manner.

The results show that the YIELD model’s esti-mated boro rice productivity for Mymensingh andBarisal is higher than the CERES-Rice’s estima-tion (Table 3). The fundamental reason is that theCERES-Rice model uses a management inputscenario for boro rice which closely resembles themanagement practices by Bangladeshi farmers.The YIELD model assumes an optimum manage-ment practice scenario which eliminates the effectof irregularities in the supply of inputs such asirrigation water or fertilizer, and this resulted inhigher yield estimations. Moreover, this modelassumes a near-linear relationship between therice productivity and supply of water. TheYIELD model also assumes a single layer of soilwhose depth at any time is determined by thedepth of roots. The model assumptions include ageneralized soil type (Mahmood and Hayes (1995)

assumed loamy soil), geotropic root growth, andan absence of soil compaction. This simplifiedscenario of soil condition and root growth cou-pled with an optimum supply of water (whichallowed rice plants to extract water from the soilrelatively easily to maintain physiological pro-cesses including evapotranspiration) produced anideal soil water balance and resulted in the higherproductivity estimates by the YIELD model. Onthe other hand, the CERES-Rice model was ap-plied to silty clay loam and silty soil for Barisaland Mymensingh, respectively. Furthermore,soils of Barisal and Mymensingh have been di-vided into five and six layers, respectively. Datafor various soil parameters for each of theselayers were used as inputs to the CERES-Ricemodel. These parameters include depth of layers,root growth factors, saturated hydraulic con-ductivity, bulk density (moist), percent of or-ganic carbon, percent of clay, silt, sand, andnitrogen, pH in water, and cation exchange capac-ity. The inclusion of these parameters in theCERES-Rice model created a more realistic soiland soil-water scenario. Moreover, theCERES-Rice model does not assume a near-linearrelationship between availability of water andyield. Thus, a more complex and realistic replica-tion of soil characteristics by the CERES-Ricemodel resulted in lower yields (compared with theYIELD model estimates) in Mymensingh andBarisal.

It is important to note that the CERES-Ricemodel calculates genetic coefficients based onthreshold temperature. The genetic coefficients in-


Table 4Percent changes in the boro rice productivity and in the length of growing season under above normal thermal conditions asestimated by the CERES-Rice (CR) and the YIELD model

BarisalMymensinghTair+

Change in productivityChange in length of grow- Change in length of growingChange in productivity(%) season (%)ing season (%)(%)

YIELDCRCRCR YIELDYIELD CR YIELD

7.30 12.10 5.102 9.70 14.10 11.103.80 13.8017.00 20.20 6.504 22.70 21.60 18.505.80 20.70

clude thermal time required to complete the juve-nile stage, rate of photoinduction, optimum pho-toperiod, thermal time for grain filling, conversionefficiency from sunlight to assimilates, tilleringrate, and grain size (Ritchie et al., 1987). Thesecoefficients significantly influence rice productivityestimates. As a result, a separate model sensitivitystudy should be conducted to determine the ef-fects of the variable threshold temperature on thecalculation of genetic coefficients and yield. Fur-thermore, it is very likely that the different meth-ods for estimating dry matter production by theCERES-Rice and the YIELD models contributedto the disagreement in estimates of rice productiv-ity at Mymensingh and Barisal. The YIELDmodel adopted a method developed by de Wit(1965). Average gross dry matter production is afunction of average observed incoming short waveradiation, average maximum photosyntheticallyactive incoming short wave radiation (PAR) onclear days, and fraction of cloud cover (Hayes etal., 1982a,b). The CERES-Rice model assumesdry matter production is a function of PAR andleaf area index (LAI) and can be affected by thenitrogen deficiency and extreme temperatures.Also, this model explicitly includes Beer’s law oflight absorption by plant community during itsestimation of dry matter production. However,the YIELD model includes LAI during its calcu-lation of the final crop productivity through acorrection parameter for crop development whichis a function of LAI during the middle of the totalgrowing season (Hayes et al., 1982a). Since theamount of light interception is critical for photo-synthesis and dry matter production, the use of

different methods to estimate this parameter bythese two models obviously influenced calculationof the yield under variable thermal conditions.

A further analysis of the results show that theCERES-Rice estimates a productivity decrease of9.7 and 22.7 for 2 and 4°C increase of air temper-ature in Mymensingh, respectively (Table 4). Onthe other hand, the YIELD model estimates 14.1and 21.6% decrease in productivity due to 2 and4°C increase of air temperature in Mymensingh.The models applications to Barisal provided simi-lar trends in estimates (Table 4). Thus, it is clearthat the productivity estimates of the CERES-Rice (YIELD) is relatively less (more) sensitive tosmall increase in temperature, while both of themodels are nearly equally sensitive to large in-crease. Furthermore, it is clear that both modelsagree that the productivity decreases with temper-ature increase.

The YIELD model’s predicted percent changesin the length of growing season for Mymensinghand Barisal are nearly similar and shows a near-linear relationship with increasing air tempera-tures (Table 4). On the other hand, theCERES-Rice model’s predicted changes in thelength of growing season are less dramatic com-pared with the YIELD model predictions (Table4). In addition, the CERES-Rice and the YIELDmodel predicted percent changes in the yield andthe length of growing season for Barisal is lessthan Mymensingh. It is important to note thatboth the YIELD and the CERES-Rice modelpredicts very similar boro rice productivity for allthree scenarios for Barisal. This is probably be-cause the coastal location of Barisal causes signifi-


Fig. 2. Variations in the range of diurnal air temperature during boro rice growing season.

cantly less variations in diurnal temperatures ascompared with Mymensingh (Fig. 2). Moreover,this thermal environment allows rice plants togrow under a relatively less temperature-stressenvironment in Barisal.

5. Conclusions and final remarks

The present study finds that YIELD modelestimated lengths of the boro rice growing seasonfrom Mymensingh and Barisal are longer than theCERES-Rice model estimates. Threshold temper-ature inputs played a key role in the disagreementbetween the two models. The CERES-Rice andYIELD model applications assumed 9 and 15°Cthreshold temperatures, respectively. As a result,the CERES-Rice model estimated longer bororice growing seasons as compared with theYIELD model. Furthermore, the estimates fromthe YIELD model for the boro rice productivityfor Barisal and Mymensingh are higher than theCERES-Rice model’s estimates. The CERES-Ricemodel’s assumption of near-optimum manage-ment practice (the YIELD model assumes opti-mum management practice), more realistic

replication of soil characteristics, methods for es-timation of dry matter production influencedthese calculations. This study finds that productiv-ity estimates of the CERES-Rice model is moresensitive to a small increase of temperature ascompared with those of the YIELD model. Bothof these models are nearly equally sensitive tolarge increase of temperature. Moreover, theYIELD model estimated length of the growingseason is more sensitive to the increasing air tem-perature as compared with the CERES-Ricemodel.

Due to the limited application (only in twostations) of the CERES-Rice model inBangladesh, it is difficult to conduct a comprehen-sive comparative study of the performance of theYIELD (originally applied to 12 stations) and theCERES-Rice models. However, it is clear thatboth the CERES-Rice and the YIELD model canbe used to estimate crop productivity and relatedparameters. Data requirements for the CERES-Rice model are significantly complex as comparedwith the YIELD model. Despite the superiority inthe replication of the actual bio-chemical-physicalprocesses by the CERES-Rice model, thedifficulties in the model calibration, validation,


and subsequent applications due to data un-availability in many cases is an impediment toextensive application of this model. This problemis especially acute for application in developingcountries. The data required to characterize phys-ical and chemical properties of the soils for theCERES-Rice model is daunting, even for manysites in the developed nations. For example, datafor percent of organic carbon and nitrogen in aparticular layer of the soil or data for pH in aparticular layer of the soil in a particular site or aregion is very difficult to acquire. It was noted byscientists from the developing countries thatsometimes a separate soil experiment projectneeds to be conducted to fulfil the soil data re-quirements of the CERES-Rice model (Hussain,S.G., 1997. Personal communication. TheBangladesh Agricultural Research Council.Dhaka, Bangladesh). The Bangladesh Agricul-tural Research Council (BARC) recommendedthat all the (Bangladesh) government funded soilsurveys should be required to fulfil the data needsfor the CERES-Rice model (Hussain, S.G., 1997.Personal communication. The Bangladesh Agri-cultural Research Council. Dhaka, Bangladesh).On the other hand, data requirements for theYIELD model are relatively simple. For example,a simplified soil classification scheme can satisfythe model requirement and yet successfully repli-cate some of the growth parameters. In otherwords, although cause-and-effect loops have beensimplified in the YIELD model, the physical basisand performance of the model is satisfactory. As aresult, it appears that inclusion of a CO2 assimila-tion procedure in the YIELD model would con-vert it to an effective tool for estimating impactsof climatic change on the crop physiology andcrop productivity due to doubling of CO2. Also, itis clear that where availability of the data islimited for the CERES-Rice model, the YIELDmodel can be a logical choice for various analysesof crop–environment relationships and crop pro-ductivity.

The future goals include extensive applicationof the CERES-Rice model so that a more detailedcomparative study can be conducted for theYIELD and the CERES-Rice models. Amongothers, this planned study will carry out sensitivity

analysis of various components of the CERES-Rice model including the affects of threshold tem-perature on genetic coefficients and its subsequentimpacts on yield calculation. Moreover, impactsof variations in the range of diurnal air tempera-ture on the yield and related growth parameterswill also be investigated for both models.

Acknowledgements

The author would like to thank two anony-mous reviewers, the editor, and Prof. David R.Legates for valuable comments and suggestions.Thanks also go to Rebecca W. Scott for readingan earlier version of the manuscript.

References

Akima, H., 1970. A new method of interpolation and smoothcurve fitting based on local procedures. J. Assoc. Comp.Mach. 17, 589–602.

Baer, B.D., Meyer, W.S., Erskine, D., 1994. Possible effects ofglobal climate change on wheat and rice production inAustralia. In: Rosenzweig, C., Iglesias, A. (Eds.), Implica-tions of Climate Change for International Agriculture:Crop Modeling Study. United States Environmental Pro-tection Agency, Australia, pp. 1–14.

Baker, J.T., Allen, L.H. Jr, Boote., K.J., 1990. Growth andyield responses of rice to CO2 concentration. J. Agric. Sci.115, 313–320.

Burt, J.E., Hayes, J.T., O’Rourke, P.A., Terjung, W.H., Tod-hunter, P.E., 1980. WATER: A Model of Water Require-ments for Irrigated and Rainfed Agriculture.

Burt, J.E., Hayes, J.T., O’Rourke, P.A., Terjung, W.H., Tod-hunter, P.E., 1981. A Parametric Crop Water Use Model.Water Resourc. Res. 17, 1095–1108.

de Wit, C.T., 1965. Photosynthesis of Leaf Canopies. Pudoc,Wageningen, Netherlands.

Doorenbos, J., Kassam, A.H., 1979. Yield Responses to Wa-ter. Irrigation and Drainage Paper, vol. 33. Food andAgricultural Organization, Rome.

Doorenbos, J., Pruitt, W.O., 1977. Irrigation and DrainagePaper, vol. 24. Food and Agricultural Organization,Rome.

Escano, C.R., Buendia, L., 1994. Climate impact assessmentfor agriculture in the Philippines: simulation of rice yieldunder climate change scenarios. In: Rosenzweig, C., Igle-sias, A. (Eds.), Implications of Climate Change for Inter-national Agriculture: Crop Modeling Study. United StatesEnvironmental Protection Agency, Philippines, pp. 1–13.


Hayes, J.T., 1986. Climatic Change and Water and YieldResponses for Rice in California. Ph.D. dissertation, Uni-versity of California, LA.

Hayes, J.T., O’Rourke, P.A., Terjung, W.H., Todhunter, P.E.,1982a. A feasible crop yield model for worldwide interna-tional food production. Int. J. Biometeorol. 26, 239–257.

Hayes, J.T., O’Rourke, P.A., Terjung, W.H., Todhunter, P.E.,1982b. YIELD: A Numerical Crop Yield Model of Irri-gated and Rainfed Agriculture.

Jensen, D.M., 1990. Potential rice yields in future weatherconditions in different parts of Asia. Neth. J. Agric. Sci.38, 661–680.

Jin, Z., Ge, D., Chen, H., Zheng, X., 1995. Assessing impactsof climate change on rice production: strategies for adapta-tion in southern China. In: Peng, S., Ingram, K.T., Neue,H.U., Ziska, L.H. (Eds.), Climate Change and Rice. Inter-national Rice Research Institute, Manilla, Philippines, pp.303–313.

Kassam, A.H., 1977. Net Biomass Production and Yield ofCrops: Present and Potential Land Use by Agro-ecologicalZones Project. Food and Agricultural Organization,Rome.

Karim, Z., Ahmed, M., Hussain, S.G., Rashid, Kh.B., 1994.Impact of climate change on the production of modern ricein Bangladesh. In: Rosenzweig, C., Iglesias, A. (Eds.),Implications of Climate Change for International Agricul-ture: Crop Modeling Study. United States EnvironmentalProtection Agency, Bangladesh, pp. 1–11.

Mahmood, R., 1993. Climatic Change and Boro Rice Yield inBangladesh: Application of a Parametric Crop YieldModel. Master’s thesis. Department of Geography andPlanning, State University of New York, Albany, NY.

Mahmood, R., 1996. Model-based estimates of air tempera-ture variations on the boro rice phenology in Bangladesh.Preprints, 22nd Annual Conference on Agricultural andForest Meteorology. The American Meteorological Soci-ety, Boston, pp. 367–368.

Mahmood, R., 1997. Impacts of air temperature variations onthe boro rice phenology in Bangladesh: implications forirrigation requirements. Agric. For. Meteorol. 84, 233–247.

Mahmood, R., 1998. Thermal climate variations and potentialmodification of the cropping pattern in Bangladesh. Theor.Appl. Clim. (in press).

Mahmood, R., Hayes, J.T., 1995. A model-based assessmentof impacts of climate change on boro rice yield inBangladesh. Phys. Geogr. 16, 463–486.

Mitchell, J.F.B., Manabe, S., Tokioka, T., Meleshko, V., 1990.Equilibrium climate change. In: Houghton, J.T., Jenkins,G.J., Ephraums, J.J. (Eds.), Climate Change: The IPCCScientific Assessment. Cambridge University Press, NY,pp. 131–172.

Penman, H.L., 1948. Natural evaporation from open water,bare soil and grass. Proc. R. Soc. Ser. A 193, 120–145.

Priestly, C., Taylor, R., 1972. On the assessment of surfaceheat and evaporation using large-scale parameters. Mon.Weather Rev. 100, 81–92.

Ritchie, J.T., 1972. Model for predicting evaporation from arow crop with complete cover. Water Resourc. Res. 8,1204–1213.

Ritchie, J.T., Alocilja, E.C., Singh, U., Uehara, G., 1987.IBSNAT and the CERES-Rice model. Weather and Rice,Proceedings of the International Workshop on the Impactof Weather Parameters on Growth and Yield of Rice,7–10 April, 1986. International Rice Research Institute,Manilla, Philippines, pp. 271–281.

Rosenzweig, C., Parry, M.L., 1994. Potential impact of climatechange on world food supply. Nature 367, 133–138.

Seino, H., 1994. Implications of climate change for Japaneseagriculture: evaluation by simulation of rice, wheat, andmaize growth. In: Rosenzweig, C., Iglesias, A. (Eds.),Implications of Climate Change for International Agricul-ture: Crop Modeling Study. United States EnvironmentalProtection Agency, Japan, pp. 1–18.

Terjung, W.H., Hayes, J.T., Ji, H-Y., O’Rourke, P.A., Tod-hunter, P.E., 1984a. Crop water requirements for rainfedand irrigated rice (paddy) in China. Arch. Meteorol.Geophys. Bioclimatol. Ser. B. 34, 181–202.

Terjung, W.H., Hayes, J.T., Ji, H-Y., Todhunter, P.E.,O’Rourke, P.A., 1985. Potential paddy rice yields for rain-fed and irrigated agriculture in China and Korea. Ann.Assoc. Am. Geogr. 75, 83–101.

Terjung, W.H., Hayes, J.T., O’Rourke, P.A., Todhunter, P.E.,1984b. Yield response of crops to change in environmentand management practices: model sensitivity analysis. I.Maize. Int. J. Biometeorol. 28, 261–278.

Terjung, W.H., Hayes, J.T., O’Rourke, P.A., Todhunter, P.E.,1984c. Yield response of crops to changes in environmentand management practices: model sensitivity analysis. II.Rice, Wheat, and Potato. Int. J. Biometeorol. 28, 279–292.

Terjung, W.H., Ji, H-Y., Hayes, J.T., O’Rourke, P.A., Tod-hunter, P.E., 1983. Crop water requirements for rainfedand irrigated grain corn in China. Agric. Water Manag. 6,43–64.

Terjung, W.H., Ji, H-Y., Hayes, J.T., O’Rourke, P.A., Tod-hunter, P.E., 1984d. Actual and potential yield for rainfedand irrigated maize in China. Int. J. Biometeorol. 28,115–135.

Terjung, W.H., Ji, H-Y., Hayes, J.T., O’Rourke, P.A., Tod-hunter, P.E., 1984e. Crop water requirements for rainfedand irrigated wheat in China and Korea. Agric. WaterManag. 8, 411–427.

Terjung, W.H., Ji, H-Y., Hayes, J.T., O’Rourke, P.A., Tod-hunter, P.E., 1984f. Actual and potential yield for rainfedand irrigated wheat in China. Agric. For. Meteorol. 31,1–23.

Terjung, W.H., Liverman, D.M., Hayes, J.T., et al., 1984g.Climatic change and water requirements for grain corn inthe North American Great Plains. Clim. Change 6, 193–220.

Terjung, W.H., Mearns, L.O., Todhunter, P.E., Hayes, J.T.,Ji, H-Y., 1989. Effects of monsoonal fluctuations on grainyield in China. Part II: Crop water requirements. J. Clim.2, 19–37.


Todhunter, P.E., Mearns, L.O., Terjung, W.H., Hayes, J.T.,Ji, H-Y., 1989. Effects of monsoonal fluctuations on grainyield in China. Part I: Climatic conditions for 1961–1975.J. Clim. 2, 5–17.

Tongyai, C., 1994. Impact of climate change on simulated riceproduction in Thailand. In: Rosenzweig, C., Iglesias, A.(Eds.), Implications of Climate Change for InternationalAgriculture: Crop Modeling Study. United States Environ-mental Protection Agency, Thailand, pp. 1–13.

Tsuji, G.Y., Uehara, G, Balas, G., 1994. DSSAT version 3,vols. 1–3. University of Hawaii, Hawaii.

USDA, Soil National Engineering Handbook, 1972.National Engineering Handbook, Hydrology, Section4.

Zhao, Zong-ci, Kellogg, W.W., 1988. Sensitivity of soil mois-ture to doubling of carbon dioxide in climate model exper-iments. Part II: the Asian monsoon region. J. Clim. 1,367–378.

.

Air Temperature Variation and Productivity in Bangladesh CERES Rice Model

Documents

Transcript of Air Temperature Variation and Productivity in Bangladesh CERES Rice Model