Prediction of crop yield, water consumption and water use efficiency with a SVAT-crop growth model using remotely sensed data on the North China Plain

Ecological Modelling 183 (2005) 301–322

Prediction of crop yield, water consumption and water use efficiency with a SVAT-crop growth model using remotely sensed data on the North China Plain X. Moa,∗ , S. Liua , Z. Lina , Y. Xub , Y. Xianga , T.R. McVicarc a

b

Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Datun Road, AnWai, Beijing 100101, PR China Department of Resources and Environmental Geosciences, Peking University, Beijing 100871, PR China c CSIRO Land and Water, GPO Box 1666, Canberra 2601, Australia Received 26 April 2002; received in revised form 25 June 2004; accepted 28 July 2004

Abstract A process-based crop growth model is developed to predict regional crop yield, water consumption and water use efficiency (WUE) using remotely sensed data for a portion of Hebei province (88 779 km2 ), most of which is located on the North China Plain (NCP). Model inputs consist of Geographic Information System (GIS) maps of land use, Digital Elevation Model (DEM), soil texture, crop canopy leaf area index (LAI) which is retrieved from the 10-day maximum composite National Oceanic and Atmospheric Administration (NOAA)–Advanced Very High Resolution Radiometer (AVHRR) data, and daily interpolated meteorological variables. The model is run at 92 367 30 × 30 . resolution grids at an hourly time-step for energy balance and daily time-step for crop growth simulation. Simulated winter wheat and summer maize yields in 1992 and 1993 are compared with both the point samples and the county-level census data. For the 108 counties, the root mean square error (relative deviation) is 1124 kg ha−1 (23%) for winter wheat and 1359 kg ha−1 (33%) for summer maize, respectively. Spatial patterns of simulated crop yield and water use efficiency are strongly influenced by irrigated/rainfed conditions. The modelled grain yield for irrigated winter wheat ranges from 3900 to 7200 kg ha−1 , which is significantly larger than when only rainfed, 100–2600 kg ha−1 . The modelled grain yield for irrigated maize ranges from 5800 to 8600 kg ha−1 , which is significantly larger than when only rainfed, 1400–4800 kg ha−1 . This suggests that the potential exists to increase yield in this region, if sufficient irrigation water is supplied. However, given the over allocation of limited surface water, an increase in irrigation is unlikely, and increasing importance will be placed on maximizing regional WUE over the NCP. The water consumption (defined here as modelled evapotranspiration (ET)) for winter wheat ranges from 330 to 500 and 70 to 280 mm for irrigated and rainfed conditions, respectively. The modelled ET for irrigated maize ranges from 350 to 520, and 140 to 350 mm for rainfed conditions. The simulated WUE

∗

Corresponding author. Tel.: +86 10 64889307; fax: +86 10 64851844. E-mail address: [email protected] (X. Mo).

0304-3800/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2004.07.032

302

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

ranges from 12.25 to 15.75 kg ha−1 mm−1 for irrigated winter wheat and from 0.5 to 8.25 kg ha−1 mm−1 in rainfed areas. The simulated WUE ranges from 11 to 19.25 kg ha−1 mm−1 for irrigated maize and from 5 to 11 kg ha−1 mm−1 in rainfed areas. These, together with the results of simulated crop yields, are comparable with previous studies in the NCP, other parts of China, and internationally, indicating the potential to apply this model to other agricultural regions. This study implicates a new view for regional agricultural and water resources management by assessing regional crop yield, water consumption and WUE consistently based on modelling biophysical processes with the aid of remote sensing. © 2004 Elsevier B.V. All rights reserved. Keywords: Simulation modelling; Spatial pattern; Regional scale; Evapotranspiration; Remote sensing; North China Plain

1. Introduction For many important grain producing regions (defined here as >75 000 km2 ) worldwide, including the North China Plain (NCP), there is increasing pressure to increase agricultural water use efficiency (WUE), i.e., the crop yield per unit of water consumed. Increasing WUE can be achieved by increasing the amount of food per unit of water consumed, or reducing the amount of water consumed to produce the same amount of food. That is, WUE combines two somewhat related processes in agricultural systems: crop yield and water consumption. A key question for managers of these large agricultural systems is how to monitor changes in WUE in a repeatable and reliable manner. Previously, only one study has monitored WUE for a large agricultural region, using an ‘input–output’ Geographic Information System (GIS) method (McVicar et al., 2002). This method was data dependent, relying heavily on county yield statistics, and cannot be transferred to other key global agricultural regions. Hence there is a need to develop a process-based approach to capture the spatial and temporal variability of crop yield and water consumption, hence providing a means to calculate WUE regionally for the many important grain producing regions. This paper addresses that need. Next we introduce two main types of crop growth models, many of which have been developed at points. Then we discuss, and provide examples, of how some of these point models have been extended spatially, using soil parameters and interpolated meteorological variables stored in GIS. Surfaces resulting from the interpolation of meteorological variables are often smooth, and combining these with the inherently high spatial density of remotely sensed data to perform the spatial extension of a point-based model provides greater spatial resolution; the benefits of this logical extension to the GIS approach are discussed. It should be noted

that outputs from spatial–temporal modelling depend on both the model philosophy and the inherent data characteristics used to solve the model. There are many crop growth models (Jørgensen, 1994), that can be broadly divided into two categories to estimate regional crop yield. The first category comprises empirical-statistical models; they relate regional crop yields to climatic variables using methods such as regression analysis (Dor´ea et al., 1997). While this type of models provides a means to estimate crop yield, water consumption cannot be estimated, and most of these models are locally calibrated, making extension to larger areas (or ‘porting’ to other regions) nearly impossible. For these reasons, this class of crop growth models will not be discussed further. The second category is comprised of process-based crop growth simulation models; they simulate the physiological development, growth and yield of a crop on the basis of the interaction between environmental variables and plant physiological processes such as photosynthesis and respiration. They are capable of providing mechanical description for prediction, both spatially and temporally. The mechanics is complex as crop growth is driven by both climatic variables (e.g., solar radiation, temperature and humidity) and soil variables (mainly moisture and nutrients). As leaf stomata are responsible for the intake of CO2 for photosynthesis and water loss through transpiration, its regulatory mechanisms are significant on the exchanges of mass and energy in soil–vegetation–atmosphere system (Lo Seen et al., 1997; Calvet et al., 1998; Cayrol et al., 2000; Ghaffari et al., 2001). It is thus acknowledged that an appropriate process-based model to predict regional agricultural crop yield should be moderately complex (Hansen and Jones, 2000; Lagacherie et al., 2000). The key is to model the main processes of both crop growth dynamics and soil–vegetation–atmosphere transfer (SVAT), and capture the essential mechanisms of crop responses

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

to weather variability and its interaction with management. Many process-based models have been developed at points that adequately take into account the interactions and feedbacks discussed previously. Given that regional land managers need methods to monitor and predict regionally, several models have been extended spatially, see the examples provided in Table 1. In all cases, the driving parameters and variables are determined for each input cell and the model is run repeatedly at the resolution covering the spatial extent of the study area. As meteorological variables are needed to run all process-based crop growth models, meteorological data are often interpolated to surfaces from the isolated stations where measurements are made. This modelling approach is described as ‘interpolate-thencalculate’, abbreviated as IC, see Stein et al. (1991) and McVicar and Jupp (2002) for further discussion. A common feature of these examples is the need to access ground-truth databases over large regions. Gaining access to such databases with adequate resolution, both temporally and spatially, has remained difficult, as collecting field data is time-consuming, expensive and often not spatially explicit (Reynolds et al., 2000). This is the main reason that the models were validated at the entire country scale or the scale of the whole study area, usually with overestimates and underestimates of each of the individual grid cells primarily canceling each other out, hence providing a reasonable model result for these large regions. However, site-specific monitoring, which will enable site-specific management, is lost in this aggregation process. Furthermore, in all the above cases, the crop growth status is a ‘simulated’ response; it is possible that this may diverge from reality. Given this, it is imperative to integrate other primary data that has high spatial and temporal resolution and provides another way to drive the model. Remote sensing offers a solution to these problems. Remote sensing has recently been widely used for the simulation of regional crop yield and water consumption, given its inherent high spatial density and the measurements being repeatable, consistent and synoptic. Additionally, remotely sensed databases are able to immediately generate high priority global data sets and improve follow-on data sets and communication within the land science community (Sellers et al., 1995; Plummer, 2000). Operationally, yield can be predicted using remotely sensed data in three main ways. They

303

are: (1) developing regression models; (2) using the light use efficiency (or Epsilon models) method; and (3) developing process-based models that capture the interactions between the SVAT and crop growth environment, as measured by remote sensing instruments. They are discussed in turn below. Firstly, regressive relationships are developed between the within season cumulated crop spectral index and census crop yield data at the county level to operationally predict the regional or country crop yield (e.g., Hayes and Decker, 1996). Gaining relatively high accuracies with readily available input data has meant that these models are in wide use, particularly in operational environments. However, such models are not process-based, these relationships heavily rely on the determination of empirical coefficients, and model developers cannot easily extend such models to include a water consumption module. Secondly, the light use efficiency method (Monteith, 1972) has been widely used for estimation of terrestrial net primary productivity (NPP) with remotely sensed data input (Kumar and Monteith, 1982; Fung et al., 1987; Maisongrade et al., 1995; Choudhury, 2001; Seaquist et al., 2003). It has also been used to predict crop yield (Clevers, 1997; Willocquet et al., 2002; Lobell et al., 2002, 2003; Bastiaanssen and Ali, 2003). Despite the popularity and simplicity of light use efficiency (or Epsilon models), this approach has several sources of uncertainty and inconsistencies. For example, light use efficiency has been found to vary with crop growth stages, and can be greatly influenced by other environmental factors (Gower et al., 1999). Finally, the third approach involves using processbased models to trace processes of dry mass accumulation and final ecological production, which has shown a promising future in the use of remote sensing to predict regional crop yield (van der Keur et al., 2001); also see Maas (1988), Moulin et al. (1995) and Plummer (2000) for a detailed review. Of all the applications of remote sensing in process-based models, the simplest is to use remotely sensed images to classify the spatial patterns of crops to form a map layer in the GIS database (e.g., Thornton et al., 1997; Basso et al., 2001; Chiesi et al., 2002; Yun, 2003), thus remote sensing is an input to one data-layer, essentially this use of remote sensing is similar to the IC-based examples shown in Table 1. More complex applications assimilate remotely sensed reflectance and thermal radiance

304

Table 1 Example of process-based crop growth models that have been extended spatially in an IC approach with GIS database Crop(s)

Spatial extent and resolution

Time extent

Accuracy (where reported) and comments on accuracy

Specific point model and the references

Tan and Shibasaki (2003)

Paddy rice, maize, wheat, soybean

1900–2003

Just test results in 2000 with FAO data, the deviations in most countries are small with a visual comparison

EPIC (Williams et al., 1990)

Priya and Shibasaki (2001)

Maize, wheat, rice

1974–1990

Winter, barley silage, maize

Simulated average yield in 1990 compares well with the reported via visual comparison; mean difference is 32% for rice, 15% for maize and 2% for wheat No comparison for specific locations; only for the total state statistically over the sub-areas

EPIC (Williams et al., 1990)

Krysanova et al. (1998) Saarikko (2000)

Spring wheat

Global 80◦ W–180◦ E and 84◦ N–56.5◦ S, 6  × 6 India, 8◦ 4 –37◦ 6 N and 68◦ 7 –97◦ 25 E, 50 km × 50 km Brandenburg, Germany subareas × sub-area Finland 10 km × 10 km

Supit and van der Goot (1999)

Wheat

France 50 km × 50 km

11 years

Rosenthal (1998)

et

al.

Sorghum

1975–1988

Harrison (2000)

et

al.

Wheat

Landau et al. (1998)

Wheat

QLD Shires, Australia, rainfall polygon × rainfall polygon 11W–42E and 35–71.5N, 0.5 × 0.5◦ UK

1981–1992

1961–1996

1961–1990

1976–1993

Making comparison at few sites in Finland with RMSD being 849 kg ha−1 ; over the southern Finland, the mean was overestimated with RMSD being 1970 kg ha−1 Better yield prediction results of total national yield was obtained than over regional or sub regional level Three shires were examined in detail, showing a general tendency for the simulated yields to be greater and more variable than those reported Four issues were addressed, no yield result is available Substantial disagreement was found between the model predictions of both yield and yield loss due to water limitation

EPIC (Williams et al., 1990)

CERES (Ritchie and Otter, 1985; Hodges et al., 1987)

WOFST (van Diepen et al., 1989) QSORG (Hammer and Muchow, 1994)

AFRCWHEAT2 (Porter, 1993) CERES-wheat, AFR CWHEAT2 and SIRIUS (Ritchie and Otter, 1985; Porter, 1993; Jamieson et al., 1998)

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

References

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

data into the crop growth models for validation (e.g., Moulin et al., 1998; Olioso et al., 1999), adjusting some parameters and initial conditions to enhance model performance (Gu´erif and Duke, 2000; Boegh et al., 2004), and retrieving vegetation biophysical variables, such as leaf area index (LAI) and fraction of absorbed photosynthetically active radiation (PAR) from remote sensing data to force the growth model (Liu et al., 1997; Coops et al., 2001; Matsushita and Tamura, 2002). To date, this last approach has been primarily applied to natural vegetation (e.g., forests) and not to the more temporally varying agriculture systems. By forcing a plant growth model with remotely sensed data, this provides a means to conduct site-specific modelling. Previously, no regional research combining remote sensing and a process-based model to predict crop yield and water consumption allowing WUE to be monitored has been reported; though Schuepp et al. (1987) provided remotely sensed, process-oriented estimates of WUE for a small area (about 100 km2 ). Given that the process-based model includes a SVAT, this allows crop water consumption, and hence WUE to be modelled, in addition to simulating crop yield. In this paper we implemented a biophysically process-based model by combining a moderately complex crop growth model with a SVAT scheme to predict crop yield, water consumption and WUE conjunctively under the support of remote sensing over the NCP region. The SVAT scheme used here has been validated previously at our study site, the NCP (Mo and Liu, 2001), and already has been scaled to a catchment (Mo et al., 2004). Remotely sensed 10-day maximum composite Normalized Difference of Vegetation Index (NDVI) data recorded by the Advanced Very High Resolution Radiometer (AVHRR) sensor with 1 km resolution are used to retrieve LAI to drive the model. The paper is organized as follows: the development of a moderately complex process-based crop growth model and model parameters used are provided in Section 2. Both the remotely sensed and GIS data of the focal region – Hebei province portion of the NCP – are introduced in detail in Section 3. In Section 4 detailed information for model running is provided, with Section 5 comparing the estimates of crop yield with the county official data to determine the bias of the model predictions; results for water consumption and WUE are also provided in Section 5. The sensitivities of key model parameters to crop yield, water consumption and

305

WUE are reported in Section 6, and conclusions, including that the model can potentially be applied to other important grain producing areas (based on reasonable results achieved here), are made in Section 7.

2. Model description The energy partitioning of the SVAT model developed by Mo and Liu (2001) is coupled with a moderately complex dynamic crop growth model. Carbon and water fluxes are coupled through the stomatal resistance that directly links photosynthesis, LAI and root zone soil moisture. The model focuses on biochemical mechanism of photosynthesis, crop dry mass formation and water consumption, and scaling up method for regional application. Due to widespread use of fertilizers on the NCP (e.g., McVicar et al., 2002), we do not consider nutrient cycling. Many research documents show that this area is not short of nutrients and the demands are met by current agricultural practices. 2.1. Photosynthesis and evapotranspiration The scheme to predict energy fluxes, photosynthesis and water balance is designed for flat crop fields, in which the crop canopy is idealized as a plane parallel medium with random distributed leaves (Mo and Liu, 2001). Transfer of solar radiation within the crop canopy is simulated with a multi-layer approach that distinguishes the direct and diffuse components of both the visible (300–700 nm) and near infrared (700–1300 nm) fractions. Considering the highly nonlinear response of photosynthesis to incident light, photosynthetic rates for sunlit and shaded leaves are estimated separately, then summed and up-scaled to the canopy using a stomatal conductance/photosynthesis relationship (Leuning et al., 1995). A detailed biochemical approach of photosynthesis for C3 (wheat) and C4 (maize) crops, similar to Farquhar et al. (1980), Collatz et al. (1991) and Sellers et al. (1996), is employed in which the net assimilation rate of CO2 (An , ␮mol C m−2 s−1 ) is expressed as An = min{Av , Ae } − Rd

(1)

where Av and Ae (both with the units ␮mol C m−2 s−1 ) are the gross rates of photosynthesis limited by Ribu-

306

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

lose Bisphosphate Carboxylase-Oxygenase (Rubisco) activity and the rate of Ribulose Bisphosphate (RuBP) regeneration through electron transport, respectively. Rd is the daytime respiration (␮mol C m−2 s−1 ). For C3 crops: Av = Vc max

ci − Γ ci + Kc (1 + o2 /Ko )

J ci − Γ Ae = 4 ci + 2Γ

Ae = αQPAR

ci Vc max P




λEc =

(2)

1 − fw fw + ∆ + γ(1 + rc /rac ) ∆ + γ 
ρCp D0 × ∆Rnc + rac



(6)

and (3) λEs =

and for C4 crops: Av = 2 × 104

equation (Monteith and Unsworth, 1990) as

(4) (5)

where Vc max is the maximum catalytic activity (␮mol C m−2 s−1 ) of Rubisco in the presence of saturating levels of RuBP and CO2 ; ci and o2 represent the intercellular CO2 and O2 concentrations (Pa) in the chloroplasts; Γ is the CO2 compensation point (Pa, equal to 0.5 o2 /S, S is the Rubisco specificity for CO2 relative to O2 (dimensionless)) without dark respiration; Kc and Ko are the Michaelis–Menten coefficients (Pa) for CO2 and O2 , respectively; J is the electron transport rate (␮mol C m−2 s−1 ) for the absorbed photosynthetic photon flux that is related to Vc max ; P is the atmospheric pressure (Pa); QPAR is the PAR (␮mol photon m−2 s−1 ) loaded on the leaf surface; α is the intrinsic quantum efficiency (␮mol C ␮mol−1 photon). Parameters of Vc max , Γ , Kc and Ko are temperature dependent and described in terms of the Q10 function (temperature coefficient, the relative increase of a parameter in response to 10 ◦ C temperature increase (Collatz et al., 1991; Sellers et al., 1992). Energy fluxes are described with a two-source scheme that discerns the canopy and soil surface separately (Shuttleworth and Wallace, 1985). Energy balance equations of the canopy and soil surface are solved analytically with a potential resistance network for homogeneous surfaces under steady conditions. The total above-canopy evapotranspiration (ET) consists of soil evaporation, canopy transpiration and its intercepted water evaporation. The canopy evapotranspiration rate, Ec (kg m−2 s−1 ), which includes both the canopy transpiration and its intercepted water evaporation, and evaporation rate from the soil surface, Es (kg m−2 s−1 ), are expressed in the forms similar to Penman–Monteith

∆(Rns − G) + (ρCp D0 /ras ) ∆ + γ(1 + rs /ras )

(7)

where Rnc and Rns are the net radiation amounts (W m−2 ) absorbed by canopy and soil surface, respectively, fw is the fraction of the wetted canopy after a rain event (dimensionless), ρ is the air density (kg m−3 ), Cp is the specific heat of air at constant pressure (J kg−1 K−1 ), D0 is the water vapor deficit (hPa) at the source height, λ is the latent heat of vaporization of water (J kg−1 ), ∆ is the slope of saturated vapor pressure-temperature curve (Pa K−1 ), γ is the psychrometric constant (hPa K−1 ), rac is the leaf boundary aerodynamic resistance (s m−1 ), ras is the aerodynamic resistance (s m−1 ) from soil surface to source height, rc is the canopy resistance (s m−1 ) and rs is the soil resistance (s m−1 ). The parameterization of resistances and the working variable D0 are described in detail in Mo et al. (2004). G is the soil heat flux (W m−2 ) and can be estimated as a fraction of net radiation above canopy, expressed as (Mo et al., 2004) G = 0.183 exp(−0.299Lt ) Rn

(8)

where Rn is the net radiation (W m−2 ) above canopy, Lt is the total green LAI. The soil moisture budget is estimated with a threelayer scheme in which crop roots only uptake water from the second layer (Sellers et al., 1986). The depth of soil layer is set as 3 m with 0.05, 1.95 and 1.00 m used for each layer, respectively. Soil hydraulic parameters are estimated with the Clapp and Hornberger (1978) empirical formula for sandy loam and loam, the dominant soil textures of the NCP. The summation of Ec and Es , denoted as ET, is for the estimation of water consumption, which is related with canopy height. The canopy height (Hc , m) is esti-

X. Mo et al. / Ecological Modelling 183 (2005) 301–322 Table 2 Parameters (◦ C) for calculating thermal growing variable

Winter wheat Summer maize

A0

A1

A2

B

110 130

1110 1400

860 1400

0 0

mated by a logistic function as follows:  

Hc max × 1.11 , Hc = 1 + exp(5.1342Dv2 − 12.399Dv + 5.2476)  Hc max ,

Dv ≤ 1 Dv > 1

(9) where Hc max is the maximum of canopy height (m), Dv is the thermal growing variable related to phenological development stage (dimensionless), varying from 0 at emergence to 1 at flowering, and finally reaching 2 at maturity. The quantity Dv is calculated as follows:  t1 (Ta −B)  0 < Dv ≤ 1 A1 Dv = (10) t 2 (Ta −B)  1+ 1 < D ≤ 2 v A2 where t1 and t2 are the cumulative days (d) from emergence to flowering and from flowering to maturity, A1 and A2 are the active cumulative temperatures (◦ C) of the above two periods, respectively, B is the minimum effective temperature (◦ C) for crop growth, Ta is the daily average air temperature (◦ C). To calculate Dv , a cumulative temperature from planting to emergence, denoted as A0 (◦ C), which is assigned a value initially, is needed. Values of A1 , A2 , B and A0 used in this paper are listed in Table 2.

307

maintenance respiration Rm (mol C m−2 d−1 ), S is the senescent rate (mol C m−2 d−1 ) taken as a fraction of the green biomass until seed maturation, affected by soil moisture (Cayrol et al., 2000). At seed maturation stage, S increases rapidly due to leaf senescing. The Rc and Rm are calculated by (Amthor, 1989) Rc = (1 − YG )(Ac − Rm )

(12)

Rm = mW

(13)

where YG is the growth conversion efficiency (dimensionless), m is the maintenance coefficient (d−1 ). According to Johnson and Thornley (1983), YG is typically about 0.75 and m is set as 0.01 d−1 under 20 ◦ C condition and corrected with a Q10 -type function with temperature. In this paper crop yield is calculated as the total DM (including the root system) multiplied by the harvest index, HI, which is slightly different from what Bastiaanssen and Ali (2003) used where only aboveground DM was considered. For winter wheat, the HI is taken as 0.48 and for summer maize 0.38. Crop WUE can be defined as the ratio of crop growth amount to ET on physiological aspect (Viets, 1962; Stanhill, 1986). The crop growth amount can be either the net DM or crop yield. There are several concepts for crop WUE which correspond to a range of spatial scales such as leaf – canopy – field – regional scales (McVicar et al., 2002), and time scales, as second – day – growth stage – season – year scales (Sinclair et al., 1984). Our working definition of WUE is the ratio of crop yield to seasonal ET, and has the unit of kg crop yield per mm of growing season ET per hectare, that is, kg ha−1 mm−1 .

2.2. Dynamic crop growth 2.3. Model parameters The daily growth of crop dry mass (DM, mol C m−2 ) denoted as W, is expressed as the balance of gross photosynthesis, respiration and senescence, namely dW = Ac − R − S dt

(11)

where Ac is the daily net photosynthesis (mol C m−2 d−1 ) at canopy level derived by integrating the instantaneous An values over daytime and the canopy profile, R is the total respiration composed of construction respiration Rc (mol C m−2 d−1 ) and

The parameters used in the model were derived from the field experiments and obtained from published references (see Table 1 of Mo and Liu, 2001). It is recognized that the two parameters, Vc max and α, are much more sensitive to the canopy photosynthesis rate than other parameters. In this study, Vc max and α were set as 150 ␮mol C m−2 s−1 and 0.385 ␮mol C ␮mol−1 photon for winter wheat (Wang and Leuning, 1998) and 41.5 ␮mol C m−2 s−1 and 0.062 ␮mol C ␮mol−1 photon for maize (Choudhury, 2001), respectively.

308

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

Fig. 1. Location of the study site (shaded grey); all of which is within Hebei province, and most of which is within the North China Plain. The plain is enclosed by the thick black line in the main map and shaded black in the inset map of China. Shan is a Chinese term for mountain range and has been used in the figure to avoid overcrowding.

3. Study area and key datasets 3.1. Hebei province portion of the NCP The study area is located in Hebei province and covers 88 779 km2 (or 29% of the NCP) consisting of 108 counties (Fig. 1). To the west are the foothills of the Taihang Shan mountain range, with the NCP gently sloping to the Bohai Sea in the east. Physiognomically, the study area is divided into three parts from west to east: the piedmont, lowland and seashore plains (Fig. 2). Traditionally high crop yields are reported from the piedmont plain where soil and climatic conditions and irrigation from groundwater recharged by Taihang Shan mountain range are favorable for agriculture. In the lowland plain, deep groundwater is insufficient and the shallow groundwater is contaminated by secondary salinity, which is not so suitable for irrigation. The seashore plain is affected by the seawater intrusion, and hence low crop yields are reported from this area. The land use in the area is dominated by the intensive dual-cropping system based on winter wheat and summer crops, including maize, millet, soybean, cotton and sorghum. Despite a variety of crops in summer, maize is the dominant summer crop in most counties

(e.g., McVicar et al., 2002). According to the traditional tillage practice, winter wheat is sown in early October, harvested in early or mid June next year, and summer maize is planted in early to mid June and harvested at the end of September, from where the cycle is repeated. Annual precipitation is about 600 mm, more than 50% of which arrives during the summer monsoon between July and September. Due to the limited and variable precipitation, in spring the winter wheat productivity is only guaranteed by irrigation; otherwise the pro-

Fig. 2. Irrigation/rainfed map of the study area including the seashore, lowland and piedmont plains and part of the Taihang Shan Mountain range.

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

ductivity will be low, with no grain being yielded in extreme drought conditions. The intensive cropping systems have increased the irrigation water requirement from 100 mm year−1 in the 1950s to 300 mm year−1 in the 1980s for winter wheat (Wang et al., 2001). Because of the low annual precipitation and shortage of surface water, the intensive agricultural systems mainly rely on irrigation water extracted from ground aquifers. Continuous over-extraction of regional groundwater has resulted in the aquifer levels decreasing by as much as a 1 m year−1 over a prolonged 40-year period in some areas (Liu et al., 2001).

for 1992 and 1993, and then interpolated linearly to the daily time-step. The relationship between LAI and the fraction of Absorbed PAR, denoted as FPAR , by the vegetation at nearly noon for crop is expressed as (Monteith and Unsworth, 1990) Lt = Lt max

The GIS datasets include: (1) land use/cover digital data and land use map; (2) an irrigated/rainfed distribution map; (3) Digital Elevation Model (DEM); and (4) soil texture map. All were resampled to 30 × 30 grid cell (each cell being approximately 926 m × 725 m at 38.5◦ N) and the data source for each is as follows. The land use/cover digital data was derived from Landsat Thematic Mapper (TM) images and was used to identify arable fields. In our simulation, only winter wheat and summer maize were considered. The 1:1 000 000 land use map (Wu, 1990) was digitized to identify irrigated and rainfed fields; this is illustrated in Fig. 2. The topographical elevation above sea level is described with a 30 × 30 resolution DEM (from USGS, 1998), most of which is below 300 m. Soil texture was digitized from a 1:14 000 000 scale map (Institute of Soil Science, 1986). 3.3. Remotely sensed data Data recorded by the National Oceanic and Atmospheric Administration (NOAA) series of satellites with the AVHRR sensor have a 1.1 km resolution at nadir and daily coverage. The two reflective bands (Red (580–680 nm) and near infrared (NIR 725–1100 nm)) are useful for monitoring vegetation development. The Normalized Difference Vegetation Index (NDVI, defined as (NIR − Red)/(NIR + Red)), and the Simple Ratio (SR, defined as NIR/Red) are two common transformations of AVHRR data that allow the prolonged effects of water stresses that reduce LAI to be detected (Maisongrande et al., 1995). We obtained ‘10day’ composites from http://edcdaac.usgs.gov/1km/

ln(1 − FPAR ) ln(1 − FPAR max )

(14)

where FPAR is calculated as a linear relation of vegetation index (dimensionless, Sellers et al., 1996): FPAR =

3.2. GIS data

309

(SVI − SVImin )(FPAR max − FPAR min ) (15) SVImax − SVImin

where SVI is the spectral vegetation index. SVImax and SVImin are the maximum and minimum SVI, respectively. FPAR max and FPAR min are the maximum and minimum FPAR corresponding to SVImax and SVImin , set as 0.95 and 0.001. Both the SR and NDVI were substituted into Eq. (15) as the SVI in this study; differences were examined and are discussed in Section 6. The above empirical relationship is sensitive to soil reflectance and differences in sun/sensor geometry (e.g., Hatfield et al., 1984; Myneni and Williams, 1994; Sellers et al., 1996). A generalized relationship between FPAR and NDVI must be seen as an average model for a variety of canopy conditions. 3.4. Climatic data Daily meteorological data were recorded at 52 climatological stations in and around the focus area, including mean, minimum and maximum air temperature, mean vapor pressure, atmospheric pressure, daily sunshine duration and rainfall. Except sunshine duration, all variables above were corrected with elevation above sea level. The daily variation of air temperature was approximately expressed as Fourier harmonics (Kondo and Xu, 1997) and solar radiation was estimated with clear sky radiation and relative sunshine duration. All variables above were interpolated into hourly data according to temperature. As spatial interpolation of climatic data was required to obtain the driving variables over the grid cells, the thin plate spline method was employed for rainfall interpolation, whereas the inverse distance squared method was used for other variables.

310

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

3.5. Crop yield data To validate the processes incorporated in the model for both winter wheat and summer maize, field experiments were conducted at the Yucheng Agro-Ecosystem Experimental Station (37◦ 53 N, 115◦ 41 E) from October 1991 to September 1992 for the measurements of total DM. The wheat and maize were fertilized with 300 and 480 kg N ha−1 , respectively. Each irrigation event was 60 mm; applied six times in a year, four for wheat and two for maize; typical of irrigation management on the NCP. County-level crop yield data were available from the Agricultural Bureau of Hebei Province, which records the total grain yield and its planting area of a crop. County-level crop yield relies on the aggregating yield estimates from village to township, then to the county, using a Chinese National standard (e.g., McVicar et al., 2002). The county-level fraction of maize area of the total summer planted crops in 1993 is 0.666 with standard deviation around the mean reported for the 108 counties being 0.145, and the fraction of winter wheat of autumn planted crops is 0.990 with standard deviation being 0.041.

the water stress is the only factor limiting crop growth and grain formation. Often on the NCP fertilizers are applied simultaneously with irrigation water, termed ‘fertigate’, this assists in dissolving the fertilizer so plant can effectively access to the nutrients.

5. Results 5.1. Validation with the field data The simulated DM of winter wheat and maize were validated with Yucheng field data. The growing processes of the live bio-DM for winter wheat and maize are shown in Fig. 3a and b, respectively. It is shown that the simulated and the observed DM agree quite well over the whole growing period, with Pearson correlation coefficient approaching 1. Data collected at the Yucheng field treatments had an above ground bio-DM at harvest in the range of 14 500–17 500 kg ha−1 for winter wheat with 300 mm of irrigation, similar results have been presented by Kang et al. (2002) and Zhang et al. (1998). The above ground bio-DM for summer maize at harvest ranged from 17 000 to 21 000 kg ha−1 .

4. Model running We simulated the winter wheat–summer maize rotation from 1 October 1991 to 30 September 1993 hourly at 30 × 30 grid resolution. The sowing dates of winter wheat and summer maize were artificially set as 1 October and 15 June, respectively, and harvest dates were determined, as Dv was equal to 2. Model initialization period was performed from 1 October 1991 to 14 June 1992, with simulation results from the 1992 summer maize, and 1993 winter wheat and summer maize being used for analysis. During each dual-crop (one winter wheat followed by one summer maize) rotation (from 1 October to 30 September the following year), the dates of irrigation were determined according to the agronomical conditions. Since the exact irrigation dates were unavailable over the study areas, the uniform irrigate dates were set as 11 October, 30 March, 24 April, 14 May, 18 July and 17 August according to the most popular and traditional irrigation systems from the local averaged information collected. We assume that fertilizers were applied in time so that

Fig. 3. Simulated and observed dry mass (DM) for (a) 1993 winter wheat and (b) 1992 maize over their respective growing seasons with Pearson correlation coefficient (R2 ) and the number of the observations.

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

The root system DM is usually estimated by its ratio to the above ground part; values from 0.15 to 0.20 for winter wheat and 0.10 to 0.15 for summer maize are common in this region. 5.2. Validation with the county census yield data While we recognize that the county yield statistics are derived by estimating yields under different productivity levels and their relevant planting sizes, and hence are potentially prone to errors (McVicar et al., 2002), these data, to some extent, represent a regional ‘ground truth’, and offer a validation of the prediction. Fig. 4 shows the county records and simulated yields

311

for wheat and maize. For winter wheat, the absolute deviation is 908 kg ha−1 , relative deviation is 23% and root mean square error is 1124 kg ha−1 . The model performance for the 2 years of maize is a bit lower, the absolute deviation is 1282 kg ha−1 , relative deviation is 33% and root mean square error is 1359 kg ha−1 , in spite that the Pearson correlation coefficient of maize 1992 is higher than that for wheat 1993. These relative deviations are quite similar to those reported by Bastiaanssen and Ali (2003) for the Indus Basin in Pakistan. The variability of county-level crop yields shows a wide range, coinciding with heterogeneity of soil, water resources and climatic conditions over the study site. In the NCP there are a variety of summer

Fig. 4. Cross plots of county-level and simulated yields for (a) wheat in 1993; (b) maize in 1992 and (c) maize in 1993 with Pearson correlation coefficient R2 and the number of samples; and (d) the relation between the ratio of the area of maize in 1992 to the total county cropping area with the difference between the simulated and observed. The empty circles in (b) show the counties with the ratio over 0.6 and the stars show the counties with the ratio less than 0.4.

312

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

crops, such as soybean and cotton that are cultivated broadly in the southeast of the study site, however, as the summer crops have a similar NDVI time series it is not possible to identify these crops directly from AVHRR NDVI images. The differences between the field size (some 10 000 m2 ) and the 1.1 km resolution of the AVHRR data exacerbate this problem. As a consequence, mixed crop types within an AVHRR pixel may significantly interfere with the inversion of crop characteristics, and in some cases will enhance model uncertainty. On the other hand, the bias of planting area estimates may also result in uncertainty of yield. It is found that the maize yields are better modelled in counties where maize dominants the summer planting (Fig. 4d). As shown in Fig. 4b, the maize area at those counties with small difference between the simulation and observation is over 60% of the total county cropping area (empty circles) and that with large difference is less than 40% (stars), approximately. 5.3. Spatial patterns of crop yield The potential crop yield (defined as crop yield without water or nutrition stress) was calculated as 7500 kg ha−1 for winter wheat and 8800 kg ha−1 for summer maize, with LAI estimated by logarithmic functions shown in Appendix A. There is considerable spatial structure presented in both winter wheat and maize yields, see Fig. 5. The wheat yields (Fig. 5a) are over 4800 kg ha−1 in the northern, western and southern areas where irrigation water quality is high, and irrigation demand is guaranteed. With more than 200 mm of irrigation over the growing period, winter wheat

plot-level yields range from 5000 to 7000 kg ha−1 in the NCP (Zhang et al., 1999; Jin et al., 1999). As expected, due to no irrigation infrastructure and poorer soils, lower winter wheat yields were modelled for the mid and eastern portions of the study area. During the wheat-growing season, total precipitation is usually scant and erratic (less than 150 mm), resulting in severe water stress, consequently low harvest grains in these areas. Prior to the 1980s soil salinity was the main factor limiting crop yield in the lowland plains, the southwest to northeast belt in the mid of the study area. These soil conditions were greatly improved after a 10-year effort in the 1980s by: (1) increasing soil organic matter; (2) lowering groundwater table by using many motor-pumped wells to withdraw groundwater; (3) building the drainage engineering; (4) flattening the soil hampers in the field where salt was accumulated; and (5) afforesting the landscape. Afforestation assists the regional environment in two ways: firstly increasing transpiration from forests with deep roots reduces the groundwater table over the region and secondly reducing the wind velocity, lowering the air temperature and increasing air humidity over the fields, which subsequently decrease the field evaporation and is useful to hinder the salt accumulation in the surface layer of the fields. Currently, when sufficient water is available, crop productivity from the lowland plain is as high as that achieved in the traditionally high-yielding western piedmont plain (e.g., Jin et al., 1999). However, limited surface water resources often results in low productivities being attained over much of the lower plain. Limited surface water resources is due to precipitation in

Fig. 5. Maps of simulated crop yield for (a) winter wheat in 1993 and (b) summer maize in 1992.

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

313

Fig. 6. Histograms of simulated yield for (a) winter wheat in 1993 and (b) summer maize in 1992.

this area being the lowest in the NCP, with the Tai Yi Shan mountains in the south (see Fig. 1) acting as a barrier to the moist from southern air movement. Fig. 5b shows that the spatial pattern of summer maize yield corresponds to the location of irrigation and rainfed fields shown in Fig. 2. Areas of high productivity were located in the north, south and piedmont areas; and areas of low productivity were found in the lowland and seashore plains where rainfall during the summer of 1992 was only about 200 mm; usually in rainfed areas the productivity can reach 4800 kg ha−1 . Although there is enough summer-dominated rainfall to meet most of the maize’s water demand, the temporal distribution of rainfall is often irregular. This may cause severe crop water stress in one or two of maize growth and development stages in rainfed condition. In addition, the dates of fertilizer application may be non-optimum, due to waiting for adequate rainfall; consequently maize yield may be distinctively lower in rainfed areas when compared with irrigated areas. The yield data over each grid of the study area show an interesting bimodal distribution (Fig. 6) corresponding to rainfed field and irrigated field, respectively. The yield of wheat ranges from 3900 to 7200 kg ha−1 in the irrigated field with the maximum frequency at 5900 kg ha−1 , whereas rainfed fields produce from 100 to 2600 kg ha−1 of wheat with the maximum frequency at 600 kg ha−1 . For summer maize, the yield ranges from 5800 to 8600 kg ha−1 with the maximum frequency at 7400 kg ha−1 in the irrigated field, whereas that of rainfed field is from 1400 to 4800 kg ha−1 with a bimodal distribution, with local maxima located at 2800 and 4100 kg ha−1 . The bimodal distribution in rainfed maize fields shows the strong spatial variability of crop yield in the focus region. The great differ-

ence between the rainfed and irrigated yields indicates that there is a great potential for yield increment in this region if sufficient irrigation water is supplied. However, given that water is currently over allocated, and that competition for water resources, from economic and population growth, and for environmental flows, is increasing, it seems unlikely this water will become readily available for the agricultural producers of the NCP. 5.4. Water consumption (ET) Both wheat and maize growing seasonal ET patterns, i.e., the water consumption pattern, show noticeably spatial variability (Fig. 7). For irrigated wheat, seasonal ET mostly ranges from 400 to 450 mm; 70% is from irrigation. For rainfed wheat, most seasonal ET is less than 250 mm, most is from rainfall, with limited residual soil moisture carried over from the previous maize growing season. The reason for low ET for rainfed wheat is that summer rainfall is almost entirely consumed by the previous maize crop; hence the available soil moisture is low when wheat is seeded in October and rainfall during the winter wheat growing season is low. In maize fields, seasonal ET mostly ranges from 350 to 450 mm; about 20% is from irrigation. Seasonal ET ranges widely from 140 to 350 mm for rainfed maize, all of which is from summer rain during the maize growing period. Our field experiments show that following the wheat-growing period there is some residual moisture in the soil layer for root uptake at the start of the maize season. This part of soil water is usually deposited under the soil plough pan layer, a compacted layer 0.2–0.35 m below the soil surface on the NCP as a result of ploughing or cultivation activity, across which soil water moves very slowly. For both

314

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

Fig. 7. Maps of simulated growing season ET for (a) winter wheat in 1993 and (b) summer maize in 1992.

irrigated wheat and maize fields, Es is about 30–40% of growing seasonal ET. Potential ET (ET0 ) calculated using a Penman–Monteith formula (Monteith, 1965) in the wheat and maize growing periods are about 560 and 500 mm, respectively. Our simulation results, see Fig. 7, are comparable with those presented in other studies in the NCP. Liu et al. (2002) presented the observed seasonal ET values of 400 to 480 mm for wheat and 395–450 mm for maize measured by a large scale weighing lysimeter. Zhang et al. (1999) reported seasonally observed ET values ranging from 305 to 455 mm at four winter wheat sites with different irrigation volumes applied. The frequency distribution of ET (Fig. 8) is consistent with the distribution of crop yield, showing also the obvious differences between the irrigated and rainfed fields. However the distribution of ET is more normally distributed than that for wheat yield (compare Fig. 8a with Fig. 6a). The ET of wheat ranges from 330 to 500 mm in the irrigated field with the maxi-

mum frequency at 430 mm, whereas ET from rainfed fields ranges from 70 to 280 mm with the maximum frequency at 130 mm. There is a high agreement between the low ET and low wheat yield; according to Zhang et al. (1999), if there is less than 110–120 mm of seasonal ET then there is no wheat yield in the rainfed areas. The distribution of ET for summer maize is different from that of the crop yield in that there is not an obvious difference between the irrigated field and rainfed field; or we say that the two populations for ET from summer maize overlap (Fig. 8b), with the ET primarily being above 140 mm and less than 520 mm for both the rainfed and irrigated fields. From Fig. 8b it seems that the maximum frequency for irrigated fields is about 420 mm and the maximum frequency for rainfed field is also 420 mm. However, comparing with the spatial distribution of ET in maize fields shown in the map of Fig. 7b, the ET from summer maize in irrigated field mostly ranges from 350 to 520 and 140 to 350 mm in rainfed conditions.

Fig. 8. Histograms of simulated ET for (a) winter wheat in 1993 and (b) summer maize in 1992.

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

315

Fig. 9. Maps of simulated WUE for (a) winter wheat in 1993 and (b) summer maize in 1992.

5.5. WUE The spatial patterns of WUE for winter wheat and summer maize are presented in Fig. 9a and b, respectively. For winter wheat (Fig. 9a) the pattern of WUE is similar to that of yield, namely high WUE (greater than 11 kg ha−1 mm−1 ) occurred in areas irrigated and low WUE (less than 8.5 kg ha−1 mm−1 ) was experienced in rainfed areas. For summer maize, most WUE is greater than 16 kg ha−1 mm−1 with irrigation and less than 8.5 kg ha−1 mm−1 under rainfed conditions. The simulated WUE mostly is less than 18 kg ha−1 mm−1 for wheat and less than 21 kg ha−1 mm−1 for maize. In the same region, McVicar et al. (2002) developed a countylevel ‘input–output’ GIS approach, and reported winter wheat WUEs from 1.2 to 21.5 kg ha−1 mm−1 and summer maize WUEs from 1.7 to 39 kg ha−1 mm−1 and over 85% of their observations for winter wheat (summer maize) WUE less than 15 (20) kg ha−1 mm−1 . The high values (summer maize WUE > 30 kg ha−1 mm−1 ) reported by McVicar et al. (2002) were attributed to errors in China’s agricultural statistics; providing weight to the use of a remote sensing-based approach, as developed here. In the 49 counties, for the winter wheat and the two-summer maize growing seasons, Fig. 10 presents the comparison of the WUEs developed here with those reported by McVicar et al. (2002). It is found that the WUE values are agreeable in most of counties, except those with very high WUEs in McVicar et al. (2002). Additionally, our simulated WUE values are also comparable with those reported in the literature for the NCP, other areas in China as well as some international sites, see Table 3.

The frequency distribution of WUE (Fig. 11) for wheat is consistent with the distribution of crop yield, showing the obvious differences between the irrigated and rainfed fields. The WUE of wheat ranges from 12.25 to 15.75 kg ha−1 mm−1 where there is irrigation with the maximum frequency at 13.75 kg ha−1 mm−1 , whereas for rainfed wheat WUE ranges from 0.5 to 8.25 kg ha−1 mm−1 with the maximum frequency at 3.25 kg ha−1 mm−1 . Comparing Fig. 11a and b, it is easy to see a clear division of WUE for irrigated wheat and rainfed wheat. However the division of WUE for irrigated and rainfed maize is unclear; suggesting that WUE is not uniquely dependent on irrigation. Maize grows during the rainy season, during which the effect of irrigation on crop yield is not noticeable as rainfall can nearly satisfy the crop demand, as it did in 1992 and 1993. By a joint analysis with Fig. 9b, we divide the irrigated and rainfed fields for maize in Fig. 11b. It is shown that the WUE for irrigated summer maize is from 11 to 20 kg ha−1 mm−1 with a bimodal distribution, with local maxima located at 13.25 and 17.25 kg ha−1 mm−1 . The bimodal distribution in irrigated maize fields shows the strong spatial variability of WUE in the focus region whereas for rainfed conditions WUE is from 5 to 11 kg ha−1 mm−1 with the maximum frequency at 7 kg ha−1 mm−1 .

6. Discussions There are several sources of uncertainty in the crop yield simulation developed here. For example, we have used a constant HI for each crop (0.48 for winter wheat

316

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

Fig. 10. Cross plots of simulated WUE from this study with those reported by McVicar et al. (2002) for (a) wheat in 1993; (b) maize in 1992; and (c) maize in 1993. On (a) three dubious points are identified by hollow circles and have been removed for some statistical analysis.

and 0.38 for summer maize), however Bastiaanssen and Ali (2003), summarizing previous studies, reported that the HI varies (ranging from 0.10 to 0.65 for winter wheat and 0.30 to 0.47 for summer maize). Consequently changes from the constant used here will affect the accuracy of the predicted yield; a spatially varying HI is detailed information currently not available. Management factors, such as cultivar’s genetic characteristics, tillage, use of fertilizers, and irrigation timing and amount will also affect crop yield significantly. Exceptional events such as pests, unseasonably

hot winds, frosts or flooding are site-specific factors that may greatly reduce the crop yields. These events, especially if they occur late in the growing season, when a large biomass has already been observed using remotely sensed data, can significantly reduce final grain yield, causing large discrepancies in modelled and measured county yields. The retrieval of crop canopy LAI may also result in prediction uncertainty. LAI can be either calculated from logistic functions, as shown in Eqs. (A.1) and (A.2), or retrieved from remotely sensed SVI, such as the SR and NDVI. Practically, logistic function can represent approximately the LAI curve, whereas the relationship between LAI and SVI is non-linear as the LAI retrieved from remote sensing is insensitive to fully canopy cover; hence decreasing the relationship reliabilities when LAI is greater than 3. Lu et al. (2003), using validation data and a review of previous research, found that a linear calibration was not supported between LAI and NDVI. Although the linear relationships were found between LAI and SR, however they were not identical (assuming a linear relationship in one case implies a slightly non-linear relationship in the other). To date, no commonly accepted SVI–LAI relation has been developed. Choudhury (1987) reported that using NDVI in Eq. (15) resulted in more accurate estimates of FPAR (hence LAI) than using SR to estimate LAI. Based on the simulated ET and assimilation amounts by using the Dv –LAI logistic functions (Eqs. (A.1) and (A.2)) with Lt max being 6.0 (4.0) for wheat (maize) under irrigated condition, we compared the simulated ET and assimilation amounts by using NDVI and SR data in Eq. (15) respectively. We found that the predictions using NDVI were very close to that by using logistic relations (less than 2%), whereas that by using SR was 16% lower than that by the logistic relations. This is why in this paper we used the NDVI to estimate LAI; the method to calculate LAI from logistic functions is only for the estimation of potential crop yield. Five model parameters (Vc max , α, A1 , A2 and Lt max ) are important to crop yield and ET (and hence WUE) and were selected to assess the model sensitivity. For both wheat and maize, each parameter was perturbed, in turn, by ±10% from the initial value; percentage differences in yield, ET and WUE from the results using the initial value are presented in Table 4. This shows that α is the most sensitive parameter, with the model being moderately sensitive to Vc max , Lt max and

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

317

Fig. 11. Histograms of simulated WUE for (a) winter wheat in 1993 and (b) summer maize in 1992.

Table 3 Documented results of WUE in NCP, other areas of China and international sites References

Site

WUE values (kg ha−1 mm−1 )

Notes

Zhu et al. (1994) Zhang et al. (1999) Jin et al. (1999)

NCP NCP NCP

14.8 11.8–14.0 14.9–23.0

McVicar et al. (2002) This paper This paper Zhang et al. (1998) Li et al. (2000) Kang et al. (2002) Regan et al. (1997) Zhang and Oweis (1999) Howell et al. (1995) Musick et al. (1994) Corbeels et al. (1998) Zhu et al. (1994) Jin et al. (1999)

Hebei portion of NCP Hebei portion of NCP Hebei portion of NCP Beijing, NCP Northwest China Northwest China Australia North America USA USA Morocco NCP NCP

1.2–21.5 12.25–15.75 0.5–8.25 9.3–15.5 6.5–12.1 7.7–14.6 5.5–16.5 2.5–16.0 4–8.8 0.84 0–11.5 19.4 20.6–23.4

McVicar et al. (2002) This paper This paper Li et al. (2000) Huang et al. (2003) Stone et al. (1996)

Hebei portion of NCP Hebei portion of NCP Hebei portion of NCP Northwest China Northwest China Semi-arid central and southern high plain of USA As above Lebanon USA USA Turkey

1.7–39 11.1–19.25 5–11 12.6–23.1 13.6 8.3–15.7

Wheat Irrigated wheat Wheat mulch, fertilizer application and irrigation treatment Wheat Irrigated wheat Rainfed wheat Wheat Wheat Irrigated wheat Wheat Wheat Wheat Wheat Wheat Maize Maize, with mulching and irrigated treatments Maize Irrigated maize Rainfed maize Maize Maize Maize

12.6–15.8 13.4–18.8 16.5 12.5–14.6 12.4–13.8

Maize Maize Maize Maize Maize

Tolk et al. (1999) Karam et al. (2003) Howell et al. (1998) Musick and Dusek (1980) Oktem et al. (2003)

318

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

Table 4 Percent sensitivity (calculated as reference minus perturbed) of yield, ET and WUE to perturbing the parameters listed by ±10% Yield

ET

WUE

−10%

+10%

−10%

+10%

−10%

Wheat Vc max α A1 A2 Lt max

−3 −9 −1 0 −3

4 −9 1 0 3

0 −2 −1 0 −2

1 2 1 0 2

−2 −7 −2 0 −2

3 6 2 0 1

Maize Vc max α A1 A2 Lt max

−2 −8 2 0 −3

3 8 0 0 3

−2 −3 1 0 −3

2 3 −2 0 3

−2 −8 2 0 −3

3 8 0 0 3

+10%

A1 , with no sensitivity to A2 . Overall, perturbing the parameters Vc max and α changes the crop yield response greater than ET. The reason for this is that ET consists of canopy transpiration, intercepted water evaporation and soil evaporation; Vc max and α are only indirectly related with the latter; Transpiration is controlled not only by canopy physiological factors, but also by available energy. The crop yield, however, is closely related with these two parameters.

7. Conclusions A biophysically process-based model, combining a moderately complex crop growth model with a SVAT scheme, has been used to predict crop yield, water consumption ET and WUE using NDVI data from NOAA–AVHRR for C3 and C4 crops. The model was calibrated and validated for a portion of Hebei province, most of which located within the NCP. The purpose of the model is to trace crop yield and water consumption consistently and to present the spatial variations of crop yield, ET, and WUE regionally, which are strongly desirable for agriculture and water resources management. Generally, the performance of the model for winter wheat and summer maize is successful with comparable deviations from previous international results. The low predicted deviation of maize is perhaps because 66% of the summer crop is maize. The dispersed pres-

ence of small-scale maize, cotton and sorghum fields within AVHRR pixels may introduce significant error in the summer integration of biomass. Additionally, the presence of varying amounts of water vapor in the atmosphere during the wetter summers, may also result in remnant noise in the remotely sensed data after atmosphere correction. The simulated spatial patterns of crop yield, ET and WUE correspond to the rainfed/irrigated pattern. The predicted crop yield for irrigated winter wheat ranges from 3900 to 7200 kg ha−1 , and ranges from 100 to 2600 kg ha−1 for rainfed conditions. The crop yield for irrigated summer maize is 5800–8600 and 1400–4800 kg ha−1 for rainfed areas. The predicted ET ranges from 330 to 500 mm for irrigated winter wheat, and from 70 to 280 mm for rainfed conditions. The ET for irrigated summer maize is primarily from 350 to 520 and 140 to 350 mm in the rainfed fields. The WUE is from 12.25 to 15.75 kg ha−1 mm−1 for irrigated winter wheat and from 0.5 to 8.25 kg ha−1 mm−1 for rainfed fields. The WUE for summer maize mostly is from 11 to 20 kg ha−1 mm−1 when irrigated and from 5 to 11 kg ha−1 mm−1 in rainfed areas. The substantial gap between the irrigated and rainfed field reveals that the crop yield of this area can be significantly enhanced if irrigation efficiency is noticeably improved, more farms are irrigated and water resources and other resources are managed better. To improve the accuracy and reliability of the crop yield prediction, more detail and timely agronomic information and higher resolution satellite images such as MODIS should be cheaply available, preferably free on-line access, globally.

Acknowledgements This research was supported by the Chinese National Natural Science Foundation project 90211007 and the Chinese National Basic Research Key Projects 2002CB412500. Many thanks to Dr. Huiyi Zhu at Institute of Geographical Sciences and Natural Resources Research for providing the TM land use data. Dr. Tim McVicar’s involvement was funded by CSIRO Land and Water and the Australian Centre for Agricultural Research (ACIAR), specifically projects LWR1/1995/007 and LWR1/2002/018. We thank the anonymous reviewers and the editors of the journal for

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

their comments that improved an earlier draft of this paper. Appendix A A.1. Using logistic functions to estimate Lt For estimation of potential crop yield, Lt is estimated by logarithmic functions with the thermal growing variable, Dv , for wheat Lt =

1.2011Lt max 1 + exp(−1.1601 − 4.1194(Dv − 0.8661) +7.9603(Dv − 0.8661)2 ) (A.1)

for maize Lt =

1.11Lt max 1 + exp(5.2476 − 12.3990Dv + 5.1342Dv2 ) (A.2)

where Lt max is the maximum Lt during the growing period (assumed to be 6.0 for winter wheat and 4.0 for summer maize). References Amthor, J.S., 1989. Respiration and Crop Productivity. SpringerVerlag, Berlin, 205. Basso, B., Ritchie, J.T., Pierce, F.J., Braga, R.P., Jones, J.W., 2001. Spatial validation of crop models for precision agriculture. Agric. Syst. 68, 97–112. Bastiaanssen, W.G.M., Ali, S., 2003. A new crop yield forecasting model based on satellite measurements applied across the Indus Basin, Pakistan. Agric. Ecosyst. Environ. 94, 321–340. Boegh, E., Thorsen, M., Butts, M.B., Hansen, S., Christiansen, J.S., Abrahamsen, P., Hasager, C.B., Jensen, N.O., van der Keur, P., Refsgaard, J.C., Schelde, K., Soegaard, H., Thomsen, A., 2004. Incorporating remote sensing data in physically based distributed agro-hydrological modeling. J. Hydrol. 287, 279–299. Calvet, J., Noilhan, J., Roujean, J., Bessemoulin, P., Cabelguenne, M., Olioso, A., Wigneron, J., 1998. An interactive vegetation SVAT model tested against data from six contrasting sites. Agric. For. Meteorol. 92, 73–95. Cayrol, P., Kergoat, L., Moulin, S., Dedieu, G., 2000. Calibrating a coupled SVAT-vegetation growth model with remotely sensed reflectance and surface temperature—a case study for the HAPEXSahel grassland sites. J. Appl. Meteorol. 39, 2452–2472. Chiesi, M., Maselli, F., Bindi, M., Fibbi, L., Bonora, L., Raschi, A., Tognetti, R., Cermak, J., Nadezhdina, N., 2002. Calibration and application of FOREST-BGC in a Mediterranean area by the

319

use of conventional and remote sensing data. Ecol. Model. 154, 251–262. Choudhury, B.J., 1987. Relationships between vegetation indices, radiation absorption, and net photosynthesis evaluated by a sensitivity analysis. Remote Sens. Environ. 22, 209–233. Choudhury, B.J., 2001. Estimating gross photosynthesis with remote sensing and ancillary data: approach and preliminary results. Remote Sens. Environ. 75, 1–21. Clapp, R.B., Hornberger, G.M., 1978. Empirical equations for some soil hydraulic properties. Water Resour. Res. 144, 601–604. Clevers, J.G.P.W., 1997. A simplified approach for yield prediction of sugar beet based on optical remote sensing. Remote Sens. Environ. 61, 221–228. Collatz, G.J., Ball, J.T., Grivet, C., Berry, J.A., 1991. Physiological and environmental regulation of stomatal conductance, photosynthesis and transipiration: a model that includes a laminar boundary layer. Agric. For. Meteorol. 54, 107–136. Coops, N.C., Waring, R.H., Brown, S.R., Running, S.W., 2001. Comparisons of predictions of net primary production and seasonal patterns in water use derived with two forest growth models in southwestern Oregon. Ecol. Model. 142, 61–81. Corbeels, M., Hofman, G., Van Cleemput, O., 1998. Analysis of water use by wheat grown on a cracking clay soil in a semi-arid Mediterranean environment: weather and nitrogen effects. Agric. Water Manage. 38, 147–167. Dor´ea, T., Sebillottea, M., Meynard, J.M., 1997. A diagnostic method for assessing regional variations in crop yield. Agric. Syst. 54, 169–188. Farquhar, G.D., Von, C.C., Berry, J.A., 1980. A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta 149, 78–99. Fung, I.Y., Tucker, C.J., Prentice, K.C., 1987. Application of advanced very high resolution radiometer vegetation index to study atmosphere–biosphere exchange of CO2 . J. Geophys. Res. 92, 2999–3015. Ghaffari, A., Cook, H.F., Lee, H.C., 2001. Simulating winter wheat yields under temperate conditions: exploring different management scenarios. Eur. J. Agron. 15, 231–240. Gower, S.T., Kucharik, C.J., Norman, J.M., 1999. Direct and indirect estimation of leaf area index, fAPAR , and net primary production of terrestrial ecosystems. Remote Sens. Environ. 70, 29–51. Gu´erif, M., Duke, C.L., 2000. Adjustment procedures of a crop model to the site specific characteristics of soil and crop using remote sensing data assimilation. Agric. Ecosyst. Environ. 81, 57–69. Hammer, G.L., Muchow, R.C., 1994. Assessing climatic risk to sorghum production in water-limited subtropical environments. 1. Development and testing of a simulation model. Field Crops Res. 36, 221–234. Hansen, J.W., Jones, J.W., 2000. Scaling-up crop models for climate variability applications. Agric. Syst. 65, 43–72. Harrison, P.A., Porter, J.R., Downing, T.E., 2000. Scaling-up the AFRCWHEAT2 model to assess phenological development for wheat in Europe. Agric. For. Meteorol. 101, 167–186. Hatfield, J.L., Asrar, G., Kanemasu, E.T., 1984. Intercepted photosynthetically active radiation estimated by spectral reflectance. Remote Sens. Environ. 14, 65–75.

320

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

Hayes, M.J., Decker, W.L., 1996. Using NOAA AVHRR data to estimate maize production in the United States corn belt. Int. J. Remote Sens. 17, 3189–3200. Hodges, T.D., Botner, D., Sakamoto, C.M., Haug, J.H., 1987. Using the CERES–MAIZE model to estimate production for the U.S. Cornbelt. Agric. For. Meteorol. 40, 293–303. Howell, T.A., Tolk, J.A., Arland, D.S., Evertt, R., 1998. Evapotranspiration, yield and water use efficiency of corn hybrids differing in maturity. Agron. J. 90, 3–9. Howell, T.A., Steiner, J.L., Schneider, A.D., Evett, S.R., 1995. Evapotranspiration of irrigated winter wheat—southern high plains. Trans. ASAE 38, 745–759. Huang, M., Shao, M., Zhang, L., Li, Y., 2003. Water use efficiency and sustainability of different long-term crop rotation systems in the Loess Plateau of China. Soil Till. Res. 72, 95–104. Institute of Soil Science, Academia Sinica (compiled), 1986. The Soil Atlas of China. Cartographic Publishing House, Beijing, China. Jamieson, P.D., Scmenov, M.A., Brooking, I.R., Francis, G.S., 1998. Sirius: a mechanistic model of wheat response to environmental variation. Eur. J. Agron. 8, 161–179. Jin, M., Zhang, R., Sun, L., Gao, Y., 1999. Temporal and spatial soil water management: a case study in the Heilonggang region, PR China. Agric. Water Manage. 42, 173–187. Johnson, I.R., Thornley, J.H.M., 1983. Vegetative crop growth model incorporating leaf area expansion and senescence, and applied to grass. Plant Cell Environ. 6, 721–729. Jørgensen, S.E., 1994. Models as instruments for combination of ecological theory and environmental practice. Ecol. Model. 75–76, 5–20. Kang, S., Zhang, L., Liang, Y., Hu, X., Cai, H., Gu, B., 2002. Effects of limited irrigation on yield and water use efficiency of winter wheat in the Loess Plateau of China. Agric. Water Manage. 55, 203–216. Karam, F., Breidy, J., Stephan, C., Rouphael, J., 2003. Evapotranspiration, yield and water use efficiency of drip irrigated corn in the Bekaa Valley of Lebanon. Agric. Water Manage. 63, 125– 137. Kondo, J., Xu, J., 1997. Seasonal variations in the heat and water balances for nonvegetated surfaces. J. Appl. Meteorol. 36, 1676–1695. Krysanova, V., M¨uller-Wohlfei, D., Becker, A., 1998. Development and test of a spatially distributed hydrological/water quality model for mesoscale watersheds. Ecol. Model. 106 (2–3), 261–289. Kumar, M., Monteith, J.L., 1982. Remote sensing of crop growth. In: Smith, H. (Ed.), Plants and the Daylight Spectrum. Academic Press, London, pp. 133–144. Landau, S., Mitchell, R.A.C., Barnett, V., Coils, J.J., Craigon, J., Moore, K.L., Payne, R.W., 1998. Testing winter wheat simulation models’ predictions against observed UK grain yields. Agric. For. Meteorol. 89, 85–99. Lagacherie, P., Cazemier, D.R., Martin-Clousire, R., Wassenaar, T., 2000. A spatial approach using imprecise soil data for modeling crop yields over vast areas. Agric. Ecosyst. Environ. 81, 5–16. Leuning, R., Kelliher, F.M., De Pury, D.G., Schulze, E.D., 1995. Leaf nitrogen, photosynthesis, conductance and transpiration:

scaling from leaves to canopies. Plant Cell Environ. 18, 1183– 1200. Li, F., Zhao, S., Geballe, G.T., 2000. Water use patterns and agronomic performance for some cropping systems with and without fallow crops in a semi-arid environment of northwest China. Agric. Ecosyst. Environ. 79, 129–142. Liu, C., Yu, J., Kendy, E., 2001. Grounderwater exploitation and its impacts on the environment in the North China. Water Int. 26, 265–272. Liu, C., Zhang, X., Zhang, Y., 2002. Determination of daily evaporation and evapotranspiration of winter wheat and maize by large-scale weighing lysimeter and micro-lysimeter. Agric. For. Meteorol. 11, 109–120. Liu, J., Chen, J.M., Cihlar, J., Park, W.M., 1997. A process-based boreal ecosystem productivity simulator using remote sensing input. Remote Sens. Environ. 62, 158–175. Lobell, D.B., Hicke, J.A., Asner, G.P., Field, C.B., Tucker, C.J., Los, S.O., 2002. Satellite estimate of productivity and light use efficiency in United States agriculture, 1982–1998. Global. Change Biol. 8, 722–735. Lobell, D.B., Asner, G.P., Ortiz-Monasterio, J.I., Benning, T.L., 2003. Remote sensing of regional crop production in the Yaqui Valley, Mexico: estimates and uncertainties. Agric. Ecosyst. Environ. 94, 205–220. Lo Seen, D., Chehbouni, A., Njoku, E., Saatchi, S., Mougin, E., Monteny, B., 1997. An approach to couple vegetation functioning and soil–vegetation–atmosphere-transfer models for semiarid grasslands during the HAPEX-Sahel experiment. Agric. For. Meteorol. 83, 49–74. Lu, H., Raupach, M.R., McVicar, T.R., Barrett, D.J., 2003. Decomposition of vegetation cover into woody and herbaceous components using AVHRR NDVI time series. Remote Sens. Environ. 86, 1–18. Maisongrade, P., Ruimy, A., Dedieu, G., Saugier, B., 1995. Monitoring seasonal and interannual variations of gross primary productivity, net primary productivity and net ecosystem productivity using a diagnostic model and remotely sensed data. Tellus 47B, 178–190. Maas, S.J., 1988. Use of remotely sensed information in agricultural crop growth models. Ecol. Model. 41, 247–268. Matsushita, B., Tamura, M., 2002. Integrating remotely sensed data with an ecosystem model to estimate net primary productivity in East Asia. Remote Sens. Environ. 81 (1), 56–66. McVicar, T.R., Jupp, D.L.B., 2002. Using covariates to spatially interpolate moisture availability in the Murray-Darling Basin: a novel use of remotely sensed data. Remote Sens. Environ. 79, 199–212. McVicar, T.R., Zhang, G., Bradford, A.S., Wang, H., Dawes, W.R., Zhang, L., Li, L., 2002. Monitoring regional agricultural water use efficiency for Hebei province on the North China Plain. Aust. J. Agric. Res. 53, 55–76. Mo, X., Liu, S., 2001. Simulating evapotranspiration and photosynthesis of winter wheat over the growing season. Agric. For. Meteorol. 109, 203–222. Mo, X., Liu, S., Lin, Z., Zhao, W., 2004. Simulating temporal and spatial variation of evapotranspiration over the Lushi Basin. J. Hydrol. 285 (1–4), 125–142.

X. Mo et al. / Ecological Modelling 183 (2005) 301–322 Moulin, S., Fischer, A., Dedieu, G., Dele’colle, R., 1995. Temporal variations in satellite reflectances at field and regional scales compared with values simulated by linking crop growth and SAIL models. Remote Sens. Environ. 54, 261–272. Moulin, S., Bondeau, A., Dele’colle, R., 1998. Combining agricultural crop models and satellite observations: from field to regional scales. Int. J. Remote Sens. 19, 1021–1036. Monteith, J.L., 1965. Evaporation and the environment. Symp. Soc. Exp. Biol. 19, 205–234. Monteith, J.L., 1972. Solar radiation and productivity in tropical ecosystems. J. Appl. Ecol. 9, 744–766. Monteith, J.L., Unsworth, M.H., 1990. Principle of Environmental Physics. Edward Armold, London, 291. Musick, J.T., Dusek, D.A., 1980. Irrigated corn yield response to water. Trans. ASAE 23, 92–98. Musick, J.T., Jones, O.R., Stemart, B.A., Dusek, D.A., 1994. Water–yield relationships for irrigated and dryland wheat in the US southern plains. Agron. J. 86, 980–986. Myneni, R.B., Williams, D.L., 1994. On the relationship between fAPAR and NDVI. Remote Sens. Environ. 49, 200–211. Olioso, A., Chauki, H., Courault, D., Wigneron, J., 1999. Estimation of evapotranspiration and photosynthesis by assimilation of remote sensing data into SVAT models. Remote Sens. Environ. 68, 341–356. Oktem, A., Simsek, M., Oktem, A.G., 2003. Deficit irrigation effects on sweet corn (Zea mays saccharata Sturt) with drip irrigation system in a semi-arid region I. Water-yield relationship. Agric. Water Manage. 61, 63–74. Plummer, S.E., 2000. Perspectives on combining ecological process models and remotely sensed data. Ecol. Model. 129, 169– 186. Priya, S., Shibasaki, R., 2001. National spatial crop yield simulation using GIS-based crop production model. Ecol. Model. 135, 113–129. Porter, J.R., 1993. AFRCWHEAT2: a model of the growth and development of wheat incorporating responses to water and nitrogen. Eur. J. Agron. 2, 69–82. Regan, K.L., Siddique, K.H.M., Tennant, D., Abrecht, D.G., 1997. Grain yield and water use efficiency of early maturing wheat in low rainfall Mediterranean environments. Aust. J. Agric. Res. 48, 595–603. Reynolds, C.A., Yitayew, M., Slack, D.C., Hutchinson, C.F., Huete, A., Peterson, M.S., 2000. Estimating crop yields and production by integrating the FAO crop specific water balance model with real-time satellite data and ground-based ancillary data. Int. J. Remote Sens. 21, 3487–3508. Ritchie, J.T., Otter, S., 1985. Description and performance of CERES-wheat: a user-oriented wheat yield model. In: Willis, W.O. (Ed.), ARS Wheat Yield Project. ARS-38, pp. 159– 175. Rosenthal, W.D., Hammer, G.L., Butler, D., 1998. Predicting regional grain sorghum production in Australia using spatial data and crop simulation modelling. Agric. For. Meteorol. 91, 263–274. Saarikko, R.A., 2000. Applying a site based crop model to estimate regional yields under current and changed climates. Ecol. Model. 131, 191–206.

321

Schuepp, P.H., Desjardins, R.L., Macpherson, J.I., Boisvert, J., Austin, L.B., 1987. Airborne determination of regional water use efficiency and evapotranspiration: present capabilities and initial field tests. Agric. For. Meteorol. 41, 1–19. Seaquist, J.W., Olsson, L., Ard¨o, J., 2003. A remote sensing-based primary production model for grassland biomes. Ecol. Model. 169, 131–155. Sellers, P.J., Mintz, Y., Sud, Y.C., Dalcher, A., 1986. A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci. 43, 305–331. Sellers, P.J., Berry, J.A., Collatz, G.J., Field, C.B., Hall, F.G., 1992. Canopy reflectance, photosynthesis and transpiration. Part III. A re-analysis using improved leaf models and a new canopy integration scheme. Remote Sens. Environ. 42, 187– 216. Sellers, P.J., Meeson, B.W., Hall, F.G., Asrar, G., Murphy, R.E., Schiffer, R.A., Bretherton, F.P., Dickinson, R.E., Ellingson, R.G., Field, C.B., Huemmrich, K.F., Justice, C.O., Melack, J.M., Roulet, N.T., Schimel, D.S., Try, P.D., 1995. Remote sensing of the land surface for studies of global change: models–algorithms–experiments. Remote Sens. Environ. 51, 3–26. Sellers, P.J., Los, S.O., Tucker, C.J., Justice, C.O., Dazlich, D.A., Collatz, G.J., Randall, D., 1996. A revised land surface parameterization (SiB2) for atmospheric GCMs. Part II. The generation of global fields of terrestrial biophysical parameters from satellite data. J. Climate 9, 706–737. Shuttleworth, W.J., Wallace, J.S., 1985. Evaporation from sparse crops—an energy combination theory. Q. J. R. Meteorol. Soc. 111, 839–855. Sinclair, T.R., Tanner, C.B., Bennett, J.M., 1984. Water-use efficiency in crop production. BioScience 34, 36–40. Stein, A., Staritsky, I.G., Bouma, J., Van Eijsbergen, A.C., Bregt, A.K., 1991. Simulation of moisture deficits and areal interpolation by universal cokriging. Water Resour. Res. 27, 1963–1973. Stanhill, G., 1986. Water use efficiency. Adv. Agron. 39, 53–85. Stone, L.R., Schlegel, A.J., Gwin Jr., R.E., Khan, A.H., 1996. Response of corn, grain sorghum, and sunflower to irrigation in the High Plains of Kansas. Agric. Water Manage. 30, 251–259. Supit, I., van der Goot, E., 1999. National wheat yield prediction of France as affected by the prediction level. Ecol. Model. 116, 203–223. Tan, G., Shibasaki, R., 2003. Global estimation of crop productivity and the impacts of global warming by GIS and EPIC integration. Ecol. Model. 168, 357–370. Thornton, P.K., Bowen, W.T., Ravelo, A.C., Wilkens, P.W., Farmer, G., Brock, J., Brink, J.E., 1997. Estimating millet production for famine early warning: an application of crop simulation modeling using satellite and ground-based data in Burkina Faso. Agric. For. Meteorol. 83, 95–112. Tolk, J.A., Howell, T.A., Evett, S.R., 1999. Effect of mulch, irrigation, and soil type on water use and yield of maize. Soil Till. Res. 50, 137–147. van der Keur, P., Hansen, S., Schelde, K., Thomsen, A., 2001. Modification of DAISY SVAT model for potential use of remotely sensed data. Agric. For. Meteorol. 106 (3), 215– 231.

322

X. Mo et al. / Ecological Modelling 183 (2005) 301–322

van Diepen, C.A., Wolf, J., van Keulen, H., 1989. WOFOST: a simulation model of crop production. Soil Use Manage. 5, 16–24. Viets, F.G., 1962. Fertilizers and the efficient use of water. Adv. Agron. 14, 223–264. Wang, H., Zhang, L., Dawes, W.R., Liu, C., 2001. Improving water use efficiency of irrigated crops in the North China Plain—measurements and modelling. Agric. Water Manage. 48, 151–167. Wang, Y.P., Leuning, R., 1998. A two-leaf model canopy conductance, photosynthesis and partitioning of available energy. I. Model description. Agric. For. Meteorol. 79, 89–111. Williams, J.R., Jones, C.A., Dyke, P.T., 1990. The EPIC model. EPIC—erosion/productivity impact calculator. 1. Model documentation. Tech. Bull. No. 1768, USDA-ARS, Washington, DC, pp. 3–92. Willocquet, L., Savarya, S., Fernandez, L., Elazegui, F.A., Castilla, N., Zhu, D., Tang, Q., Huang, S., Lin, X., Singhd, H.M., Srivastava, R.K., 2002. Structure and validation of RICEPEST, a production situation-driven, crop growth model simulating rice

yield response to multiple pest injuries for tropical Asia. Ecol. Model. 153, 247–268. Wu, C., 1990. The 1:1,000,000 Land Use Map of China. Science Press, Beijing. Yun, J.I., 2003. Predicting regional rice production in South Korea using spatial data and crop-growth modeling. Agric. Syst. 77, 23–38. Zhang, J., Sui, X., Li, B., Su, B., Li, J., Zhou, D., 1998. An improved water use efficiency of winter wheat grown under reduced irrigation. Field Crop Res. 59, 91–98. Zhang, H., Wang, X., You, M., Liu, C., 1999. Water–yield relations and water-use efficiency of winter wheat in the North China Plain. Irrig. Sci. 19, 37–45. Zhang, H., Oweis, T., 1999. Water-yield relations and optimal irrigation scheduling of wheat in Mediterranean region. Agric. Water Manage. 38, 195–211. Zhu, Z., Stewart, B.A., Fu, X., 1994. Double cropping wheat and corn in a sub-humid region of China. Field Crops Res. 36, 175– 183.