You are on page 1of 7

AGRICULTURAL AND FOREST METEOROLOGY

Agricultural and Forest Meteorology 86 (1997) 291-297

Predicting th.e microclimate inside a greenhouse: an application of a one-dime,nsional numerical model in an unheated greenhouse
, Yinsuo Zhang Yitzhak Mahrer *, Meir Margolin
The Seagram Center For Soil and Water Sciences, Faculty of Agriculture, The Hebrew University of Jerusalem, Rehouot 76100, Israel Received 16 January 1996; accepted 3 July 1996

Abstract A one-dimensional numerical model has been adapted to predict the microclimate inside an unheated commercial greenhouse during a continuous period of 51 days. The main outputs of the model (hourly air and leaf temperatures and relative humidity) were used to derive the leaf wetness duration (LWLI) each day. Measuredmicroclimateparameters inside the greenhouse were used to test the model performance during the corresponding period. Reasonable agreement was found between the predicted and measured parameters for the entire period. The root mean square differences between the predicted and actual air and leaf temperatures, relative humidity and LWD were 1.2 and 1.8C, 5.8% and 1.9 h d- , respectively. 0 1997 Elsevier Science B.V.
Keywords: Greenhouse; Microclimate; Leaf wetness duration: Relative humidity; Leaf temperature

1. Introduction

Microclimate parameters inside a greenhouse such as air and leaf temperatures, relative humidity and leaf wetness duration (LWD), not only influence the growth of the vegetation inside the greenhouse, but can also be critical factors affecting the spread of epidemics of certain crop diseases such as Botrytis (Yunis et al., 1990) and Septoriu upiicola (Lacy, 1994). Thus microclimate variability is of considerable interest not only to greenhouse growers but also to horticultural advisers. Direct on-site microclimate monitoring can provide the most useful data, but because of the high cost and sophistication of the
* Corresponding author. Present address: Meteorological
Huhhot 010051, China. 016%1923/97/$17.00 0 1997 Elsevier Science B.V. All rights reserved. PII SO168-1923(96)02422-7

Institute of Inner Mongolia

measuring equipment, most greenhouse growers are reluctant to monitor the elements themselves. As an alternative, estimates of the indoor microclimate are made using external weather data from local meteorological stations. Several studies have been performed in order to understand the relationship between the outdoor and indoor microclimates of greenhouses. Following the first detailed analysis of the energy balance in a greenhouse (Businger, 19631, a number of dynamic models have been developed (Takakura et al., 1971; Takakura, 1989; Kimball, 1973; Avissar and Mahrer, 1982; Van Bavel et al., 1985). Even though most of these models produced reasonable estimates, very few numerical models have ever been used in practice for applied greenhouse microclimate predictions over a long continuous period. In this study, we have adapted the one-dimensional numerical model devel-

292 Table 1 List of measurements Elements Air temperature Loaf temperature Relative humidity Leaf wetness duration Wind speed Solar radiation Data collection

Y. Zhang et al. /Agricultural

and Forest Meteorology 86 (1997) 291-297

taken, the instruments

used and their location Model Copper-constantan Everest 4000A Rotronic YA-100 Campbell 237 MET1 Li-200SB c21x, Campbell (K) Location Outside and inside Inside Outside and inside Inside Outside Outside

Instruments Thermocouple Infrared thermometer Humidity sensor Leaf wetness sensor Cup anemometer Pyranometer Data logger

oped by Avissar and Mahrer (1982) to predict the microclimate elements in an unheated commercial greenhouse for tomatoes. The main outputs of the model for a 51&y period (hourly air and leaf temperatures, relative humidity inside the greenhouse) were used to derive the leaf wetness duration (LWD) for each day.

house were measured hourly from 6 February to 28 March 1995. All the instruments used and their locations are listed in Table 1. Characteristics of the greenhouses, as well as the plant and soil parameters used as model inputs, whether measured or estimated from the literature, are listed in Table 2.

3. The model 2. Experimental procedure 3.1. Model concept and equations To test the model performance and the accuracy of the predicted results, we conducted an experiment in an unheated greenhouse in Uza, Israel. The microclimate parameters both inside and outside the greenTable 2 Characteristics Greenhouse:

In general, the model equations are based on the energy balance method (Avissar and Mahrer, 1982) applied to four vertical layers in the greenhouse (soil,

of the greenhouse,

soil, vegetation

and cover Vegetation: Crop Shading factor Solar radiation Transitivity Reflectivity Absorptivity Thermal radiation Reflectivity Emissivity Cover: Material Solar radiation Transmissivity Reflectivity Absorptivity Thermal radiation Transmissivity Reflectivity Emissivity Tomato 75% 8% 21% 71%

Length
Width Height at eaves Height at ridge Orientation Heaters Ventilation soik Organic matter Sand Clay Porosity Specific density Solar radiation Reflectivity Absorptivity Thermal radiation Emissivity Reflectivity

143 m 28 m 3.5 m 4.5 m East-west None Natural ventilation

by opening side cover

2% 70% 28% 40% 1.6 g cm- 20% 80% 94% 20%

5% 96%

PE (IR absorbent) 60% 20% 20% 45% 15% 40%

Y. Bang

et al. /Agriculhual

and Forest Meteorology 86 (1997) 291-297

293

air, vegetation and cover). With this approach it is assumed that the heat storage in the plants, internal air and cover can be neglected, since the heat capacities of these elements are small compared with existing heat fluxes. The heat and mass fluxes are modeled at each layer to generate balance equations, which have the following general form: for energy,
R,+R,+E+A+S=O, (1)

This means that the model is even more applicable to almost all types of commercial greenhouses. 3.2. Model parameters

3.2.1. Constant parameters

and for moisture


Mi, - M,,,, = 0,

(2)

where R, is the net flux of solar radiation, R, is the net flux of long-wavelength radiation, E, A and S are the latent, sensible and conductive heat fluxes, and Mi, and M,,, are the total moisture fluxes entering or leaving a specific layer. Since the moisture and temperature vary with depth in the soil layer, the model divided the soil layer into 10 homogeneous sublayers. The thermal diffusion and moisture transfer equations applied to each layer were

All the constant parameters, such as characteristics of the greenhouse, cover and soil properties as well as heating and ventilation settings, are given at the beginning of the model run and do not change during the modeling period. All of these parameters are defined either by estimation or by measurements. The constant parameters for the greenhouse in Uza are listed in Table 2. 3.2.2. Time-dependent parameters To apply the model to long time periods (such as an entire season), several timedependent factors have to be taken into account: the growth of the crop (indicated in the model by the leaf area in&x, LAI, and the root density distribution with depth, RD), the amount of irrigation applied to the soil (which also varies during the crop growing season), and the lower boundary conditions (at the deepest soil layer) of temperature and moisture. For most crops, the LAI and RD change from 0 (before germination) to maximum values (when the plants are fully grown) at which they remain constant for some time period and then begin to decline. In the case of modeling an entire growing season, time-dependent functions (or measured values) of the LAI and RD should be introduced into the model even though accurate measurements are very laborious and difficult. However, for certain crops or varieties good estimates of these parameters can be achieved by less frequent measurements during the growing period. For example, Milthorpe (1956) used only limited observations to derive the LAI for an entire growing season. In this study, the tomato plants in the greenhouse in Uza were already fully grown. The plant canopy was about 2 m high with a LAI of approximately 5.5. This variable changed very little during the modeling period. The variation in the amount of irrigation, IR( z, t), during the growing season is incorporated as scheduled by the grower.

(3)
-= at

ae

az at az+ D paz + - v,( ZJ) + IR( ZJ)

ae

aq

aK,

(4)

where t is time, T, is the soil temperature, z is depth, K,(z) is thermal diffusivity, 8 is soil volumetric moisture, De is soil moisture diffisivity, D, is soil moisture thermal diffusivity, V,(z,t> is evapotranspiration, and IR(z,t) is the amount of irrigation. Eqs. (3) and (4) are solved numerically from the surface to a depth of 1 m at which the diurnal temperature variation is virtually eliminated (T, = constant), and the moisture gradient is negligible

(ae/az = 0).
Detailed descriptions of the equations applied to all the other layers in the model were presented in Avissar and Mahrer (1982). The energy and mass balance equations are then solved by an iterative procedure to obtain the unknown temperatures and humidity of the different layers. The model also takes into account ventilation, artificial cooling and heating, as well as the air exchange through the walls and roof due to leaks.

294

Y. Zhang et al. /Agricultural

and Forest Meteorology 86 (1997) 291-297

3.2.3. Transfer coefficients

It should be noted that the uncertainty involved in the choice of expressions for the transfer coefficients between the different layers inside the greenhouse constitutes a critical sensitivity for most greenhouse models (Kimball, 1973). The simulations made by this model also indicate the necessity for accurate evaluation of these parameters. The original model (Avissar and Mahrer, 1982) suggested three altemative expressions based on the work of Jakob (1949), Sears and Zemansky (1960) and Kimball (1973). For the case study presented in this paper, the following transfer coefficient (h) gave the best results (Sears and Zemansky, 1960): h=

yJATJ0.25,

where AT is the temperature difference between any two vertically spaced layers, and y is an empirical constant (y = 0.21 for AT > 0; y = 0.11 for AT < 0). 3.3. The application scheme A flow chart indicating the model application procedure is shown in Fig. 1. The initial input data required include general characteristics of the greenhouse; the thermal and optical properties of the cover, soil and plants; the plant shading factor; and the leaf area index (LAI). The external climatic conditions (air temperature, relative humidity, wind speed and solar radiation) are prescribed at each time step.

El
RUN HODEL THE
l4RNE.SDUMTION

4.

Results
and RH inside the

4.1. Air and leaf temperature greenhouse

Fig. 1. Flow chart of the model application scheme.

Figs. 2 and 3 show the predicted and measured leaf temperature and relative humidity values for two typical periods of 10 consecutive days chosen from the measuring period 6 February to 28 March 1995. Table 3 provides the root mean square error (RMSE) between predicted and observed values of air and leaf temperatures and relative humidity based on an analysis of 1224 values (51 days times 24 hours per day). Fig. 4 shows the hourly distribution of the RMSE. In general, good agreement between the

Y. Zhang et al./Agriculmral

and Forest Meteorology

86 (1997) 291-297

295

Table 3 Error analysis of the model results (1224 hour data set) Elements Root mean square error 1.2 1.8 5.8 Maximum absolute error 5.7 6.2 28.1

measured -

predicted

It

Air temperature Leaf temperature Relative humidity

measured

--

OkI
17/2 18/2 19/2 20/2 2112 22/2 23/2 2412 25/2 26/2

Date (1995)
Fig. 2. Diurnal cycles of predicted and measured air and leaf temperatures and relative humidity, inside an unheated greenhouse in Uza (Israel) for 10 consecutive days (17-26 February 1995).

predicted and measured values for all three elements was achieved during the entire modeling period, although the night-time simulations were in better overall agreement with observations than the daytime estimates. This is most likely due to the fact that more complicated energy and water exchange and transfer processes existed during the daytime when the solar radiation and air exchange through the greenhouse walls and roof reached their maxima. Throughout most of the measuring period the temperatures during daylight hours inside the greenhouse (both measured and predicted) were lower then the external temperature even though no cooling systems were used. This is due mainly to the large LA1 of the plants (5.51, which caused intensive evaporative cooling during the day, and thus acted like an evaporative cooling system. 4.2. Leaf wetness duration (LWD) Several methods have been developed to predict the leaf wetness duration (LWD) because of its

measured -

predicted

3.0 0D 2.5 l? 2 2.0 E $ E 1.5 g : 2 Ok&


lW3 19/3 2013 2113 22l3 2313 24l3 2513 26l3 27l3

2.0 0.0 .o .o .O .O I 4 I 6 1 0.0 I I / I I 10 12 14 16 18 20 22 ;: 0 2 z 3

measured __

predicted

1.0 0.5 0.01 0

Date (1995)
Fig. 3. Diurnal cycles of predicted and measured air and leaf temperatures and relative humidity, inside a non heated greenhouse in Uza (Israel) for 10 consecutive days (18-27 March 19951.

Time (hours LST)


Fig. 4. Hourly distribution of the root mean square error of air and leaf temperatures and relative humidity during the entire measuring period (6 February to 28 March 1995).

296

Y. Zhang et al. /Agricultural

and Forest Meteorology 86 (1997) 291-297

$j

16.0

5 E 12.0

4.0-

Y I 1.1 I i
1712 2512

0.0-n I I
912

I x
513

,
1313

1, >
2513

Date (1995)
Fig. 5. Daily variation of measured leaf wetness duration (LWD) and predicted LWD using the thresholds of (1) predicted leaf temperature being equal or less than the dew point temperature (LWD,); (2) predicted leaf temperature being less than the dew point temperature plus 15C (LWD,,,); and (3) predicted relative humidity is greater than 90% (LWD,).

air layer, and may be slightly different from that adjacent to the vegetation layer where dew appears. This seemingly trivial difference may cause large errors during the periods when dew begins to condense or evaporate. Another reason for the differences might be that condensed water remains on the sensor (or leaf) for some time before it completely evaporates. This is not taken into account in the model. As the difference between the leaf temperature and the dew point is also small and within the range of the measuring error itself, a criterion slightly larger than 0C (such as 15C in this case) could be more reasonable. The more common simple criterion (RH > 90%) used by Walin (1963) also gave reasonable good results (RMSE = 2.5 h dd >, although an error as high as 10 h was found on one day.

5. Conclusions
The good agreement between the measured and predicted temperatures and humidity during a 5 l-day period indicate that the model can be used in practice to predict greenhouse microclimates using external weather data from local meteorological stations. The model outputs can be used to predict the leaf wetness duration within an acceptable range of accuracy. This could be of considerable value in helping to prevent the onset of certain kinds of diseases within greenhouse crops. When local weather forecasts are available, the model can also be applied to forecast microclimate parameters and leaf wetness duration in greenhouses. Such information would be valuable for predicting possible outbreaks of certain climate-related diseases (e.g. Botrytis) and can be used by growers to help them decide on the optimal precautions to take to prevent possible epidemics.

critical effect on the development of certain crop diseases. Theoretically, dew appears on leaves when the leaf temperature reaches or is lower than the dew point (Yunis et al., 1990). However, because of the diffkulty in obtaining suffkiently accurate leaf temperature data, most researchers instead try to use other elements (i.e. relative humidity, wind speed, minimum temperature, dew point depression) to estimate the LWD empirically (Walin, 1963; Gillespie and Barr, 1984; Huber and Gillespie, 1992; Gleason et al., 1994). The outputs of the model gave reasonably accurate estimates of the air and leaf temperature and relative humidity, especially during the night, when most leaf wetness occurs. The occurrence of leaf wetness was calculated using the model output data for the following criteria: (a) T,eaf- Tdew Tdew <O"C;(b) TIeilr- < 1.5C; and (c) RH > 90%. The predicted and measured results for the 51 days of the simulation are given in Fig. 5. The results show that the predicted LWDs according to criterion (a) (T,eaf < O'C)were much lower - Tdew than those measured (RMSE = 13.3 h d-l), and that much better agreement between predicted and measured LWDs (RMSE = 1.9 h d- ) was obtained using criterion (b). This apparent contradiction with the theory might be explained by the fact that the model calculated dew point is the dew point in the

References
Avissar, R. and Mahrer, Y., 1982. Verification study of a numerical greenhouse microclimate model. Trans. ASAE, 25: 171 l1720. Businger, J.A., 1963, The greenhouse (glasshouse) climate. In: W.R. Van Wijk (Ed.), Physics of Plant Environment. NorthHolland, Amsterdam, pp. 273-318.

Y. Zhang et al. /Agricultural and Forest Meteorology 86 (I 997) 291-297 Gillespie, T.J. and Barr, A.. 1984. Adaptation of a dew estimation scheme to a new crop and site. Agric. Meteorol., 31: 289-295. Gleason, M.L., Taylor, SE., Loughin, T.M. and Koehler, K.J., 1994. Development and validation of an empirical model to estimate the duration of dew periods. Plant Disease, 78(10): 1011-1016. Huber, L. and Gillespie, T.J., 1992. Modeling leaf wetness in relation to plant disease epidemiology. Ann. Rev. Phytopathol., 30: 553-571. Jakob, M., 1949. Heat Transfer. Wiley, New York, p. 758. Kimball, B.A., 1973. Simulation of the energy balance. of a greenhouse environment. Agric. Meteorol., 11: 243-260. Lacy, M.L., 1994. Influence of wetness periods on infection of celery by Septoria apiicola and use in timing spray for control. Plant Disease, 78(10): 975-979. Milthorpe, F.L., 1956. The growth of the leaves. Nottingham University Easter School in Agricultural Science. Butterworths. London.

291

Sears, F.W. and Zemansky, M., 1960. College Physics, AddisonWesley, Reading, MA. Takakura, T., 1989. Technical models of the greenhouse environment. Acta Hort., 248: 49-53. Takakura, T., Jordan, K.A. and Boyd, L.L., 1971. Dynamic simulation of plant growth and environment in the greenhouse. Trans. ASAE, 14: 964-971. Van Bavel, C.H.M., Takakura, T., and Bot, G.R.A., 1985. Global comparison of three greenhouse climate models. Acta Hort., 174: 21-33. Walin, J.R., 1963. Dew: Its significance and measurement in phytopathology. Phytopathology, 53: 1210-1216. Yunis, H., Elad, Y. and Mahrer, Y., 1990. Effects of air temperature, relative humidity and canopy wetness on gray mold of cucumbers in unheated greenhouses. Phytoparasitica, 18(3): 203-215.

You might also like