A Physicoempirical Model for Soil Water Simulation in Crop Root Zone

Pedosphene 21(4): 512- -521, 2011ISSN 1002-0160/CN 32-1315/PPEDOSPHERE回2011 Soil Science Society of ChinaPublished by Elsevier B.V. and Science Presswww .elsevier com/locate/pedosphereA Physicoempirical Model for Soil Water Simulationin Crop Root Zone*lSHANG Song -Hao1,*2 and MAO Xiao-Min21 State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua University, Beijing100084 (China)2 Center for Agricultural Water Research in China, College of Water Conservancy and Civril Engineering, China AgriculturalUniversity, Beijing 100083 (China)(Received September 13, 2010; revised May 5, 2011)ABSTRACTTo predict soil water variation in the crop root zone, a general exponential recession (GER) model was developed todepict the ecession process of soil water storage. Incorporating the GER model into the mass balance model for soil water,a GER-based physicoempirical (PE-GER) model was proposed for simulating soil water variation in the crop root zone. ThePE GER model was calibrated and validated with experimental data of winter wheat in North China. Simulation resultsagreed well with the field experiment results, as well as were consistentdeveloped soil water balance model which required more detailed parameters and inputs. Compared with a previouslydeveloped simple exponential recession (SER) based physicoempirical (PE -SER) model, PE-GER was more suitable forapplication in a broad range of soil texture, from light soil to heavy soil. Practical application of PE-GER showed thatPE-GER could provide a convenient way to simulate and predict the variation of soil water storage in the crop root zone,especially in case of insufficient data for conceptual or hydrodynamic models.Key Words: general exponential recession model, irrigation scheduling, recession process, soil textureCitation: Shang, S. H. and Mao, X. M. 2011. A physicoempirical model for soil water simulation in crop root zone.Pedosphere. 21(4): 512- -521.INTRODUCTIONgement to predict soil water content in the crop rootzone (Panigrahi and Panda, 2003; Ranatunga et al,Soil water is the main source for crop root uptake in2008). The soil water dynamics is mainly influencedthe field and usually taken as a key index for field waterby meteorological, soil and crop factors. Various simu-management. Soil water dynamics also has great infue-lation models have been proposed with different setsnce on nutrients and contaminants transport in soils,of model parameters, inputs and outputs (Ranatungaplant growth, meteorological processes and groundwa-et al, 2008; Shang et al, 2009), which can be classi-ter dynamics (Daly and Porporato, 2005). Therefore,fied into three types, i.e., empirical models, conceptualresearch on soil water has aroused the interest of soilmodels and hydrodynamic models (Shang et al, 2009).scientist, agronomist, hydrologist, meteorologist, etC.,Empirical models are usually data-driven black-boxand greatly facilitated the development of theoreticalbasis and application of soil water models (Raats, 2001;and outputs. Examples of empirical models for soil wa-Bastiaanssen et al, 2007; Shang et al, 2009).ter simulation include regression model (Kang, 1987),Because long-term in situ monitoring of soil watersupport vector machines model (Wu et al, 2008), arti-is time consuming and costly, the simulation-modelingficial neural networks model (Shang et al, 2009), andapproach was widely used in agricultural water mana-time series analysis model (Wu et al, 1997; Oliveira,*lSupported by the National Natural Science Foundation of China (Nos. 50879041 and 50939004) and the Program for NewCentury Excellent Talents in University of the Ministry of Education of China (Nim 7F-9059中国煤化工*2Corresponding author. E mail: shangsh@tsinghua.edu.cn.TYHCNMHGPHYSICOEMPIRICAL MODEL FOR SOIL WATER SIMULATION5132001). These empirical models are not established onmodels require fewer inputs and can be easily usedthe physical mechanism of the soil water flow and relyin practice, but they consider little of the mechanismheavily on field observation results at specified site andgoverning soil water flow. Therefore, a physicormpiri-weather conditions. However, they benefit from hav-cal model that combines the advantages of conceptualing fewer parameters and requiring fewer inputs thanmodel and empirical model may be more appropriate :hydrodynamic or conceptual models, and are thereforefor modeling soil water variation in case of insuficientmore convenient to be used in specified conditions.data. Shang et al. (2000) proposed a physicoempiri-Conceptual models are based on the mass bala~cal model, which considered the principle of soil waternce of soil water in the crop root zone (e.g, Shangbalance and used a simple exponential recession (SER)and Mao, 2006; Nishat et al, 2007; Jiang et al,curve to depict the recession process of soil water in2008)，where soil water variations were derived fromthe crop root zone. This SER model based physicoem-other measured or estimated water balance compo-pirical (PE-SER) model can avoid the estimation ofnents, such 88 precipitation, irrigation, evapotranspi-evapotranspiration and require fewer inputs to simu-ration, and deep percolation. These models mainlylate the dynamic variation of soil water in the cropconsider the soil water quantity in the root zone, whilezone. It has been validated with field data for lightthe mechanism of soil water flow was not fully consi-soils (Shang et al, 2000; Ma et al, 2002), but its ap-dered. They are more widely used in agricultural waterplicability for heavy soils is still questionable.The main purposes of this study were to extend thequire fewer inputs and can be easily used at the fieldSER model to a general exponential recession (GER)scale (Panigrahi and Panda, 2003).model and to develop a GER model based physicoem-Hydrodynamic models, or physically-based models,pirical (PE-GER) model to simulate the variation ofare based on general laws of the physical mechanismsoil water in the crop root zone.governing soil water fow, e.g., the principle of massconservation and the Darcy's law. Richards' equationMODEL DEVELOPMENTis generally used in these models for the description ofGER model for soil water recession in the crop rootunsaturated soil water fow (e.g., Ji et al, 2007; Shangzoneet al, 2009; Yadav et al, 2009). More detailed hydro-dynamic models consider simultaneous heat and waterThe variation of soil water content in the crop roottransfer in the soil-plant-atmosphere continuum (e.g,zone is infuenced by meteorological factors as well asChoudhury and Monteith, 1988; Braud et al, 1995;crop and soil properties. Soil water storage (W) tendsShang et al, 2009). Hydrodynamic models can pro-to increase with rainfall and irrigation and decrease :vide a profound description of spatial and temporalwith evapotranspiration and deep percolation. Whenvariation of soil water. These outputs can be readilyneglecting lateral soil water fAow, soil water balanceincorporated into transport models for predicting the(OW) in crop root zone can be described by the fol-fate of contaminants, nutrients and/or pesticides in thelowing equation:root zone (Siminek et al., 2008), which is crucial foragricultural environmental impact study. Despite theOW= W2- W1=P+I- R- (ET + q)Ot(1)above advantages, hydrodynamic models require de-tailed information of meteorology, soil and crop, whichwhere W1 and W2 are the initial and final soil wa-restricts their practical applications.ter storage, respectively, in time interval [t1,t2], P, Rand I are the amounts of precipitation, runoff and ir-widely used in the agricultural water managementrigation, respectively, in this time interval, ET is the .(Panigrahi and Panda, 2003). A key challenge of con-evapotranspiration rate, q is the soil water fAux throughceptual models is the calculation of field evapotranspi-the bottom of crop root zone (positive for percolationration (Shang and Mao, 2006), which usually requiresdownwards), and Ot=tz-t1 is the time step.detailed meteorological data, crop information and soilDuring periods without irrigation or precipitation,water regime. The meteorological datasoil water is gradually removed from the crop root zonethe estimation of reference evapotranspiration may bethrough evapotranspiration and deep percolation. Inunavailable under some circumstances, especially forthis case, soil water balance in the crop root zone canlong- term predictions. On the other hand, empirical be simplified to中国煤化工MHCNMH G514s. H. SHANG AND x. M. MAOOW = -(ET + q)Ot(2)which is close to the wilting point.Supposing the initial soil water storage at time toThe differential form of the above soil water ba-is Wo, and assuming that the recession coefficient re-lance equation ismains constant in a short time interval, the integrationof Eq.5 results in the general exponential recession pat-d=-ET-q3)tern of soil water storage, which iswhere t is time.W(t) = Ws + (Wo - W,)exp[-k(t - to)](6)From Eq- 3, the recession rate (dW /dt) of soil wateris mainly determined by ET and q. In general condi-where W(t) is the soil water storage at time t.tions, ET has more infuence on the recession rate thanEq.6 indicates that soil water decreases exponen-q does. .tially with time in the period from to till a precipitationET is mainly infuenced by meteorological condi-event or an irrigation application. If no irrigation ortions, crop factors and soil water regime and can beprecipitation apply for a long time, soil water storageestimated by the equation described by Allen et al.tends to approach Ws.(1998): .PE- GER model for simulating soil water variation inthe crop root zoneET = KgETmConsidering the recession process of soil water de-where Kg is the cofficient of soil water stress, andscribed by the GER model and the increase of soil wa-ETm is the crop water requirement or crop evapotran-ter due to effective precipitation and irrigation, PE-spiration without water stress. Kg is mainly infuencedGER model for the simulation of soil water storage inby soil water regime and soil properties. It is close tothe root zone on a daily basis can be written as0 when soil water content is lower than the wiltingpoint, and increases linearly to 1 until the soil waterWt+1= Ws+ (Wt- Wz)exp(-kOt)+ P+Ii7)content reaches the critical value of soil water contentshowing no stress to evapotranspiration (Allen et al,where Wt is the soil water storage of the tth day, Pt1998). Therefore, for a given ETm, ET is mainly de-and It are the effective precipitation and irrigation intermined by the soil water storage in the root zone.the tth day, and the time step Ot= 1d.On the other hand, q is infuenced by soil hy-When neglecting Ws, PE-GER model (Eq.7) can .draulic conductivity and potential gradient at thbe simplified to PE-SER model (Shang et al, 2000),lower boundary and can be calculated from the Darcy'swhich is .Law. However, the soil water potential near the bot-Wr+1 = Wtexp(-koOt)+ P + It8)tom of root zone is usually not monitored. As an al-ternative method, q can be estimated from empiricalwhere ko, similar to k in PE-GER model, is the re-formula related with the soil water storage in the rootcession cofficient in PE SER model. PE-SER modelzone (Wang, 1996; Shang and Mao, 2006). Based on(Eq. 8) has shown to be applicable for light soil (Shangthis empirical estimation, downward percolation oCet al, 2000; Ma et al, 2002), while not validated forcurs if the soil water storage in the root zone is largerheavy soil yet.than the critical value of soil water storage for soil wa-ter fux after irrigation or precipitation (Wang, 1996).Determination of the model parametersConsidering the above mentioned relationshipsamong ET, q, W and dW /dt, the recession rate of soilParameters of PE-GER model include the steadywater storage is assumed to be linearly related withsoil water storage in the root zone (W3) and the timethe soil water storage at the specified ETm, i.e,dependent recession coefficient (k). In time interval[t1, t2] without irrigation and precipitation, the average .dW=-k(W- W.)(5)recession coefficient can be estimated from initial andfinal soil water storage if Ws is known. The followingformula can be obtained from Eq.6.which is mainly infuenced by ETm and soil properties,中国煤化工and Ws is the steady soil water storage in the root zoneh-t9)TYHCNMH GPHYSICOEMPIRICAL MODEL FOR SOIL WATER SIMULATION515where W1 and W2 are the soil water storage in the cropwhere M is the number of data used in calibration,zone at time t1 and t2, respectively. However, measureWmk and Wek (k = 1, 2...,M) are the measured andment error of soil water content may lead to signifcantsimulated values of soil water storage in the crop rootuncertainty for the estimation of k. Considering thatzone, respectively. The calibration was processed withW。is also unknown, direct estimation of ht is imprac-the Microsoft Excel Solver, which is a Microsoft Ex-ticable.cel add-in that can be used to determine the optimumFrom Eqs.3to5, k= (K,ETm +q)/(W- Ws). Invalue for an objective function in a particular targetcases of deep groundwater table and less rainfall, q iscell on a Microsoft Excel worksheet (Winston, 2004).far less than ET. Considering that Ks varies linearlywith the difference of W and the wilting point (Wp)SITE DESCRIPTION AND FIELD EXPERIMENTfor moderate water content and Ws is close to Wp, thevariation of k is similar to ETm, i.e., increasing in earlySite descriptiongrowing stages and decreasing in late stages (Allen etal, 1998). Here the temporal variation of k after theField experiment was carried out at the Xiaohegreening of winter wheat is described by the same equa-Irrigation Experiment Station located at 112° 40' Etion as the recession coefficient ko for PE SER modeland 37° 38' N in Shanxi Province of North China.(Shang et al, 2000).The elevation above sea level is 782.6 m. This rek= kmexp[-(t' - )2/c2](10)gion belongs to semiarid area with less precipitationand higher evaporation potential. During the periodwhere km is the maximum value of recession coefficient,of 1978 to 2003, the mean annual precipitation is 430t' = t/T is the normalized growing time defined as themm, and the mean annual reference evapotranspirationratio of growing time (t) to the whole growing periodestimated from the Penman-Monteith equation (Allen(T), tm is the normalized growing time correspondinget al, 1998) is 986 mm. Groundwater level in this re-to krm, and C is a shape cofficient.gion was deeper than 8 m in recent years.In this case, the model parameter vector is p =Table I shows the pbysical and chemical properties[Ws, kem, tm,c]T. If these parameters, as well as the ini-in different soil layers. For the main root zone of wintertial soil water storage, irrigation and precipitation datawheat (0-1 m soil layer), soil is cassified as silty clayare known, we can simulate the soil water storage inloam or clay loam (Table I). The average bulk densitythe rest growing period using Eq.7 on a daily basis.is 1.42 g cm-3, and volumetric water contents at fieldSimulation results and field measured data are com-capacity and saturation are 0.396 and 0.480 cm3 cm-3 ,pared to calibrate model parameters, where the objec-respectively.tive function is the minimization of the mean squarederror (MSE) between simulation and measurements.Experiment design and measurementsWinter wheat ( Triticum aestivum L.) is one of theMinimum MSE =一S [Wek(p) - Wmk]2(11)main crops in this region. It grows in dry scasons fromk=1TABLE ISoil pbysical and chemical propertiesSoilSoil depthParticle size distributionSoil textureOrganicTotal Na)Total pa) Total Ka)layermatter2)SandSiltClaym%g kg~0-2029.353.717.0Silty clay loam 0.8980.90.418.16I20- 3525.058.7Silty clay loam 0.5350.53.60II35-6046.338.6 15.1Clay loam0.430.480.393.63V60 -10029.752.3 18.0Silty clay loam100-20015.371.2 13.5Silt loam)Soil samples for particle size and nutrient analysis were taken at different soil layer:中国煤化工were takenon September 10, 2002 from soil layers of 020, 20 -50 and 50 -100 cm, respectively.TYHCNMHG516s. H. SHANG AND X. M. MAOearly October to next June. During soil frozen periodfrom late November to next February, wheat growscount ratio of the neutron probe defined as the ratiovery slowly or even stops for about 100 d. There-of count to standard count, R is the correlation coeffi-fore, the growing period from greening (usually earlycient, *** indicates the statistical significance level ofMarch) to harvest (usually late June) will be our majorP≤0.001, and n is the number of data pairs for linearconcern. .regression.In two growing seasons of 2003 and 2004, experi-The soil water content was usually measured everyments were carried out at 30 experiment plots to study10 d, with additional measurerment before and afterthe field water balance. Adjacent plots were separatedirrigation and heavy rain. Since the maximum rootby a brick wall of 1.5 m deep. Each plot is 3 m widedepth of winter wheat is about 1 m, soil water stora-and 6.67 m long with an area of 20 m2. Guard rows ofge in the upper 1 m soil layer was considered in the3 m wide around experiment area were used to reducefollowing analysis.the edge efect.Meteorological factors, including precipitation, airIn each growing season, winter wheat was plantedtemperature, humidity， wind speed and sunshinein early October and harvest in late June. Five ir-hours, were also monitored. Total precipitations in therigation treatments (Table II for 2003) with differentgrowing period after the greening of winter wheat inirrigation frequency, timing and amount after greening2003 and 2004 were 111 and 74 mm, respectively. Theof winter wheat were tested for 6 replications. Amountmaximum amount of a precipitation event was 32 mm,of irrigation ranged from 0 to 225 mm. Irrigation wa-and no runof occurred in the field. Therefore, runoffter was pumped from groundwater and gauged by flowwas not considered in the modeling. Although unne-:essary for PE-GER model, the detailed meteorologicalmeter.Soil water content was measured every 20 cm in thedata were used to calculate the reference evapotranspi-upper 2 m soil layer using an L520 neutron moistureration (Allen et al, 1998), which was necessary for theprobe (manufactured by the Jiangsu Academy of Agri-water balance simulation to validate PE-GER model.cultural Sciences, Nanjing, China). Considering thatMODEL CALIBRATION AND VALIDATIONthe measurement of neutron probe is inaccurate nearthe soil surface (Mohamed et al, 1997), the surfaceModel calibrationlayer and deep soil layer were calibrated separately u-PE-GER model was calibrated with measured datasing gravimetric water content measurement. The cali-bration curves of the neutron probe were both linearfrom five experiment plots in 2003, i.e., Nos.1,4, 7, 10and 13. To simplify the calibration procedure, we firstfor surface and deep soil layers, which are:took different values of W。in the range of 0 to 200θ = 0.8637CR - 0.0016 (R = 0.879***,mm. For each Ws, three parameters in Eq. 10 were op-timized by the Microsoft Excel Solver, and parametersn= 29, the surface soil layer)(12)corresponding to minimum value of MSE were selectedas the calibrated model parameters.θ = 0.6332CR + 0.0123 (R = 0.915***,Fig. 1 shows the least mean squared errors (LMSE)n= 174, the deep soil layer)(13)of model simulated and measured soil water storageTABLE IIIrrigation treatmentsin 2003 after the greening of winter wheatIrrigationExperiment plots forIrrigation quota at different stagestreatmentthe 6 replicationsJointingHeadingMilkingTotalmm1,2,3, 16, 17, 1875I4,5, 6, 19, 20, 215045II7,8, 9, 22, 23, 2401510, 11, 12, 25, 26, 275v13, 14, 15, 28, 29, 30中国煤化工MHCNMH GPHYSICOEMPIRICAL MODEL FOR SOIL WATER SIMULATION17further compared with those from PE-SER model.Fig. 2 shows the comparisons of the measured soil waterstorage and the simulated values using PE SER model700and PE-GER model for different irrigation treatmentsin 2003. The results indicate that PE-GER model ispreferable to PE SER model. The differences between600simulation results of these two models are very small inthe early growing stages (March to April) when soil wa-500ter storage is high and the recession coefficient is low.For late growing period (May to June), PE SER modelshow higher, similar or lower simulation values of soil4005100150200water storage for higher (Fig. 2a), medium (Fig. 2b)W, (mm)or lower (Fig. 2c, d) irrigation treatments, respecti-Fig. 1 Relationship between the least mean squared errorvely. It indicates that PE-SER model tends to over-(LMSE) of simulated and measured soil water storage inestimate or underestimate the soil water storage forthe root zone and the steady soil water storage (W).higher and lower irrigation treatments in the late grow-ing stage. Compared with PE SER model, PE GERfor diferent steady soil water storage (W,). LMSE de-model is more accurate and applicable for heavy soilscreases with Wg when Wg is lower than 160 mm andin the present study. Since PE-SER model is a specialincreases otherwise. It reaches the minimum of 490case of PE-GER model and is applicable to light soilsmm2 when Ws =160 mm. As expected, this value of(Shang et al, 2000), PE-GER. model can be appliedWs is very close to the wilting point of 158 mm (Shangover a wider range of soil textures.and Mao, 2006). In this case, the calibrated parametersResults from PE-GER model were also comparedfor PE-GER model arewith those from a soil water balance model (Shang.and Mao, 2006), which had more parameters (inclu-[W,km,t"m,qT = [160 mm, 0.0645,0.692,0.398]”(14)ding parameters for crop cofficient, water stress -For PE SER model without considering the steadyefficient and lower boundary fux) and required de-soil water storage, the value of LMSE is 718 mm2, 47%tailed information of meteorology (including precipi-higher than that of PE-GER model with Wg = : 160 mm.tation and meteorological factors to estimate referenceThis indicates that Ws is indispensable for proper de-evapotranspiration) to estimate field evapotranspira-scription of soil water recession process, especially fortion from Eq. 4. Fig. 3 shows that simulation resultsby these two models are very close for different irriga-heavy soils which have higher wilting point.tion treatments. Therefore, PE-GER model providesPerformance of PE- GER model in simulating soil wa-in alternative to simulate the variation of soil waterter variationstorage in the crop root zone, especially in cases ofinsuficient data for conceptual or hydrodynamic mo-Using the calibrated parameters, temporal varia-dels because PE GER model requires fewer inputs fortion of soil water storage in the crop root zone wassimulation.simulated with PE-GER models for all 30 experimentPE-GER mode was also validated by experimentplots (including 5 plots for model calibration) in 2003.data in the other year, i.e., 2004. The model parame-The simulated and measured values of soil water stora-ters were taken directly from the above simulationsge for different experiment plots were all well corre-without further calibration. As shown in Fig. 4, simula-lated, with the correlation coefficients between 0.726tion results with different irrigation scheduling agreedand 0.961. Linear regression of simulated and meareasonably well with the measured ones. This indi-sured values of soil water storage with zero interceptcates that the recession cofficient is relatively stableresulted in the slope ranging from 0.965 to 1.066 forin different years, although it varies significantly withindiferent experiment plots, which are all close to thethe growing period. Therefore, PE-GER model can beperfect value of 1. These indicate that PE-GER modelused for predicting the soil water storage in various cir-and calibrated parameters are applicable to simulatingcumstances even without detailed knowledge of meteo-the variation of soil water storage in the crop root zone.rology and croP gu中国煤化工cofficientThe simulated results from PE-GER model werecalibrated from hi|YHCNMHG518s. H. SHANG AND x. M. MAO350ab奸300250小200150100Mar. 1Mar.31 Apr. 30 May30 Jun. 29Mar. 1 Mar. 31Apr.30 May30 Jun. 29d甄“小山00Mar.1 Mar. 31pr.30 May30 Jun. 29Fig. 2 Comparisons of soil water storage (W) between the measured values (dot, with the error bar indicating the standardderivation of measurement) and the simulated values using PE SER model (dashed line) and PE GER model (solid line)for rrigation quota of 225 mm (a), 105 mm (b), 45 mm (c) and 0 (d) after grcening in 2003.50p30300车E 250h程纸ξ 20050 t44+100 LMar. 1Mar. 31Apr. 30.May30 Jun. 29Mar.31 Apr. 30DateFig. 3 Comparisons of soil water storage (W) between the measured values (dot, with the error bar indicating the standardderivation of measurement) and simulated values using the water balance model (dashed line) and PE-GER model (solidline) for irrigation quota of 225 mm (a) and 0 (b) after greening in 2003.DEMONSTRATION OF THE APPLICATION 0and time -dependent recession coficient (k) which isPE-GER MODELexpressed as a function of time (Eq. 10). These parame-ters are infuenced by meteorology, crop and soil chara-cteristics, and can be determined from historical dataeasily simulatcd using PE-GER model. The model pa-iputs includerameters include the steady soil water storage (W.)initial soil waterYHCNMHGPHYSICOEMPIRICAL MODEL FOR SOIL WATER SIMULATION519350ab300250Re.200weeeeair150 .100Mar. 1Mar. 31Apr.30 May 30Jun. 29.Mar.'Apr. 30.May30 Jun. 29s350(350 [c250 .15010Jun.29 Mar. 1Mar.31Apr. 30May 30Jun. 29DateFig. 4 Comparisons of the soil water storage (W) between the measured values (dot) and simulated values using thePE- GER model (solid line) for irrigation quota of 0 (a), 45 mm (b), 120 mm (c) and 180 mm (d) after greening in 2004.pitation and artificially controlled irrigation. Whenin the studying area. Without efective rainfall in theprecipitation forecasting is available, the model can befollowing days, soil water storage will decrease to 201used for irrigation management, i.e., to predict if irri-mm on April 28 (point B in Fig. 5). Therefore, irriga-gation is necessary for a given irrigation standard.tion is required on that day. 3) The effective precipi-In periods without precipitation and irrigation,tation or planned irrigation volume (60 mm for thethe recession process can be calculated from PE-GERexample) is added to the soil water, and soil water sto-model for various initial soil water storages. Fig. 5360shows a series of such recession curves of soil water340storage in the crop root zone for the growing periodof winter wheat from early March to late June in the320studying area. When such figure has been drawn out280for a particular region and crop, it can be used easilyfor field water management even by a farmer knowing260240little about the model.Procedures guiding irrigation based on Fig.5 are220illustrated below: 1) Estimate the initial soil water180storage through observation or other approaches in a160given date during the growing period. For example, itwas estimated to be 300 mm on March 15, correspon-ding to point A in Fig.5. 2) Plot the recession processfrom point A through interpolation of two adjacent reFig.5 Recession curves of soil water storage (W) in thecrop root zone and practice demonstration for winter wheatsoil water reaches its lower limit. For the latter case,in the studying site. A indicates the specific point of ini-the corresponding date is the time for irrigation. Fortial soil water storage, and B and C indicate the specificexample, 200 mm, approximately 50% of the field ca-points of soil water starage hof中国煤化工irigationpacity, is taken as the lower limit of soil water storageapplication, respectivYHCNMH G520s. H. SHANG AND x. M. MAOrage rises to a higher level (point C in Fig. 5). And 4)ter Manage. 92: 111-125.the recession process continues. Steps (2) and (3) areBraud, I, Dantas Antonino, A. C., Vauclin, M., Thony, J.repeated for the following growing period.L. and Ruelle, P. 1995. A simple soil-plant atmospheretransfer model (SiSPAT) development and field verifi-CONCLUSIONScation. J. Hydrol. 166: 213- 250.Choudhury, B. J. and Monteith, J. L. 1988. A four-layerBy analyzing main factors affecting the recessionmodel for the heat budget of homogeneous land sur-process of soil water, we found that the recession ratefaces. Q. J. Roy. Meteor. Soc. 114: 373- 398.of soil water storage was approximately linearly relatedDaly, E. and Porporato, A. 2005. A review of soil mois-ture dynamics: from rainfall infiltration to ecosystemwith the soil water storage itself. Based on this linearresponse. Environ. Eng. Sci. 22: 9 -24.relationship, a general exponential recession (GER)Ji, X. B, Kang, E. S, Chen, R. S., Zhao, W. Z, Zhang,model was deduced to depict this recession process.Considering the water balance and the GER modelsimulating water balances in cropped sandy soil withfor soil water recession, a physicoempirical (PE-GER)conventional flood irrigation applied. Agr. Water Man-model was proposed to simulate the variation of soilage. 87: 337- -346.water storage in the crop root zone on daily basis.Jiang, J., Zhang, Y., Wegehenkel, M., Yu, Q. and Xia, J. .Compared with conceptual models for soil water, this2008. Estimation of soil water content and evapotran-model used the GER model to depict the recession pro-spiration from irrigated cropland on the North Chinacess of soil water, and thus avoided the estimation ofPlain. J. Plant Nutr. Soil Sci. 171: 751-761.evapotranspiration and soil water fAux through the bot-Kang, S. Z. 1987. Preliminary study on dynamic simula-tion of soil water in dry land. Agric. Meteorol. (intom of crop root zone which require detailed meteoro-Chinese). 8(2): 38- 41.logical and crop information and soil hydraulic pro-perties. The model was calibrated and validated withMa, X. Y., Wang, J. Q. and Li, Z. J.2002. Study on fieldsoil moisture forecast based on soil moisture exponen-experiment data in winter wheat field in North China.tial depletion relations. Res. Soil Water Conserv. (inIt was found that the simulation results agreed wellChinese). 9(2): 93 -96.with the measured data and were very close to a soilMohamed, s. O., Bertuzzi, P., Bruand, A., Raison, L. andwater balance model which requires more parametersBruckler, L.1997. Field evaluation and error analysis ofand inputs. Compared with a previously developedsoil water content measurerment using the capacitancePE SER model, PE -GER model is more accurate andprobe method. Soil Sci. Soc. Am. J. 61: 399 408.Nishat, S, Guo, Y. and Baetz, B. W. 2007. Development ofapplicable over a wider range of soil textures.a simplified continuous simulation model for investiga-Based on PE-GER model, the procedure was pro-ting long-term soil moisture fluctuations. Agr. Waterposed on irrigation scheduling by referring to the reces-Manage. 92: 53 -63.sion curves of soil water. Therefore, PE-GER modelOliveira, M. T.2001. Modeling water content of a vineyard. provides a convenient and practical way for irrigationsoil in the Douro Region, Portugal. Plant Soil. 233:management.213- 221.However, PE-GER model was established and in-Panigrahi, B. and Panda, s. N. 2003. Field test of a soilvestigated based on winter wheat in North China withwater balance simulation model. Agr. Water Manage.deep ground water table, and only for two kinds of58: 223- -240.Raats, P. A. C.2001. Developments in soil-water physicssince the mid 1960s. Geodrma. 100: 355 -387.mulation models, model parameters are site and cropdependent. The applicability of PE-GER model forRanatunga, K., Nation, E. R. and Barratt, D. G. 2008.Review of soil water models and their applications inother situations still requires further study.Australia. Environ. Model. Softw. 23: 1182 - 1206.Shang, S. H. and Mao, x. M.2006. Application of a simula-REFERENCEStion based optimization model for winter wheat irriga-tion scheduling in North China. Agr. Water Manage.Allen, R. G., Pereira, L. S., Raes, D. and Smith, M. 1998.85: 314- -322.Crop Evapotranspiration: Guidelines for ComputingShang, S. H,Lei, Z. D. and Yang, S. X. 2000. EmpiricalCrop Water Requirements. FAO, Rome.model for soil moisture forecast in winter wheat field.Bastiaanssen, W. G. M., Allen, R. G., Droogers, P., D'Urso,T. Chin. Soc. Agr. Eng. (in Chinese). 16(5): 31- -33.G. and Steduto, P. 2007. Twenty-five years modelingShang, s. H, M中国煤化工Yang, S. Xirrigated and drained soils: State of the art. Agr. Wa-2009. Soil WeApplicationsfYHCNMH GPHYSICOEMPIRICAL MODEL FOR soIl WATER SIMULATION21(in Chinese). Science Press, Beijing.997. Time series analysis of field-measured water con-Siminek, J, van Genuchten, M. Th. and Sejna, M. 2008.tent of a sandy soil. Soil Sci. Soc. Am. J. 61: 736-742.Development and applications of the HYDRUS andWu, W., Wang, X., Xie, D. and Liu, H. 2008. Soil waterSTANMOD software packages and related codes.↓content forecasting by support vector machine in purpledose Zone J. 7: 587- 600.hilly region. In Li, D. (ed.) International FederationWang, Y. R.1996. Study on the transformation of waterfor Information Processing, Computer and Computingin the field and water production function for winterTechnologies in Agriculture. Vol. 1. Springer, Boston.wheat (in Chinese). Master Thesis, Tsinghua Univer-pp. 223- 230.sity, Beijing.Yadav, B. K, Mathur, S. and Siebel, M. A. 2009. SoilWinston, W. L. 2004. Microsoft Excel Data Analysis andmoisture dynamics modeling considering the root com-Business Modeling. Microsoft Press, Redmond.pensation mechanism for water uptake by plants. J.Wu, L, Jury, W. A., Chang, A. C. and Allmaras, R. R.Hydrol. Eng. 14: 913 -922.中国煤化工MHCNMH G