Support system for decision making in the management of the greenhouse environmental based on growth model for sweet pepper

Agricultural Systems 139 (2015) 144–152

Contents lists available at ScienceDirect

Agricultural Systems journal homepage: www.elsevier.com/locate/agsy

Support system for decision making in the management of the greenhouse environmental based on growth model for sweet pepper J.A. Sánchez-Molina a,b,⁎, N. Pérez a,b, F. Rodríguez a,b, J.L. Guzmán a,b, J.C. López c a b c

Automatic Control, Robotics, and Mechatronics Research Group, Dep. of Informatics, University of Almería, 04120 Almería, Spain The Agrifood Campus of International Excellence (ceiA3), Spain Experimental Station of Cajamar Foundation, Paraje Las Palmerillas, 04710 Sta. M del Águila, El Ejido, Almería, Spain

a r t i c l e

i n f o

Article history: Received 8 March 2015 Received in revised form 20 June 2015 Accepted 29 June 2015 Available online xxxx Keywords: Temperature CO2 Radiation Control C++

a b s t r a c t Over recent years, intensive Mediterranean agriculture has been gradually changing from very low-technology greenhouses to those incorporating intermediate and, in some cases, advanced technology. The Decision Support System (DSS) can help growers, engineers and students learn about and manage the system dynamic and its influence on production. This work shows the calibration and validation of a pepper growth model based on physiological principles and on the works of several authors. The model gives good results on dry matter production estimation as well as partitioning between different plant organs. The promising results obtained in the model validation (tested with real data from the southeast of Spain) allowed us to design and implement the Graphical User Interface (GUI). This software tool, which predicts pepper crop growth using models based on climatic variables, permits the development of an optimum control system. The DSS final version is user-friendly and easily managed by growers. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction Crop growth models can be used to better understand the physiological processes that determine yield and allow strategies to be tested. Many of these processes have their limits genetically determined, but microclimate plays an important role (Rodríguez et al., 2015; Sánchez-Molina et al., 2014a). In this line, a greenhouse is an ideal setting for farming because climate variables can be manipulated to achieve optimal plant growth and development (Ramirez-Arias et al., 2012; Sanchez et al., 2012). In this work, automation of the management of crop growth is carried out by the hierarchical approach comprising subsystems, processes and variables in relation to the crop (Fig. 1): the inputs are variables that can be acted on (windows, irrigation valves, heating, etc.), the disturbances are variables that cannot be manipulated (Berenguel et al., 2006), but can be measured so their effect on the system is taken into account (e.g., weather, pests and diseases) and the outputs are the variables to be controlled (interior temperature, relative humidity, water, nutrients and so forth). Integration of the greenhouse subsystem is an important challenge because of the need to know not only the interactions, but also the possibility of controlling them (Rodriguez et al., 2008). In this architecture, supervision and control are carried out by the grower (Wolf et al., 2002; Ewert et al., 2002). The information necessary for decision-making comes from a diversity of sources, such as ⁎ Corresponding author. E-mail address: [email protected] (J.A. Sánchez-Molina).

http://dx.doi.org/10.1016/j.agsy.2015.06.009 0308-521X/© 2015 Elsevier Ltd. All rights reserved.

common knowledge, books and specialized publications, and the experience of the persons who work in the specific field. The availability of this information would help optimize crop growth (production), lower costs and assist in decision-making at the right time (Rodríguez et al., 2015). Consequently, control system requirements must be determined based on the desired crop performance and accordingly have a crop growth model is therefore important (Marcelis et al., 1998; Heuvelink, 1999). Nonetheless, these models are quite complicated to use and they require considerable knowledge of the crop. Therefore, having a software tool capable of implementing such a model would greatly assist growers and researchers in the decision-making process (Jame and Cutforth, 1996; Hoch and Agabriel, 2004). Decision Support Systems (DSS) and Expert Systems fill this role, they are computer-based systems that assist decision-makers by allowing them to access and use data and growth models (Aubert et al., 2012; Castelán-Ortega et al., 2003; Chevalier et al., 2012). These systems are interactive, utilizing models with internal and external databases. The main characteristics of DSS are flexibility, effectiveness, and adaptability. These characteristics have guided much of the DSS research; however, the potential benefits they offer greenhouse climate control have not as yet been exploited. This work describes a DSS based on a crop growth model for pepper with the aim of optimizing crop growth conditions. These crop growth models are based on climatic variables and are a good support tool for decision-making systems in production management. The model presented in this paper is focused on the models developed by Marcelis

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

145

Fig. 1. Hierarchical architecture (Ramirez-Arias et al., 2012).

et al. (2006), which was modified and improved with the works of Scaife and Jones (1976), Collatz et al. (1991), Rabbinge et al. (1991), Teh (2006), and Gonzalez-Real et al. (2008). The paper has been organized as follows: Section 2 gives the background to the crop growth modeling process. Section 3 gives a general overview of the greenhouse and its main characteristics as well as the experimental data collected. The main results and discussion are summarized in Section 4. The decision-making software developed is shown in Section 5 including an example of its utilization. In Section 6, the main conclusions are drawn. Finally, an Appendix section has been added with information as well as parameter and variable values. 2. Pepper growth model The model presented here is dynamic and deterministic and driven by: photosynthetically active radiation (PAR), temperature and CO2 concentration inside the greenhouse (Fig. 2). In general, such models quantitatively describe the mechanisms and processes that cause crop growth. They are typically hierarchical models of at least two levels of depth with the highest level describing the production of dry matter (growth, WT) and leaf area index (LAI), while the lower level describes leaf photosynthesis (Pleaf) and maintenance respiration (Rm). The calibrated and validated model in this paper is based on the works of Marcelis et al. (2006), who presented a mechanistic dynamic model validated with six experiments in greenhouses in France and The Netherlands. It is built on the generalist model of Gijzen (1994), called INTKAM and is derived from the cucumber growth model developed by Marcelis (1994). Gross leaf photosynthesis is estimated using the model of Farquhar et al. (1980) and stomatal conductance from Nederhoff et al. (1992). Crop photosynthesis is calculated from these leaf photosynthesis calculations by using the Gaussian integration method (Goudriaan and Laar, 1994) and maintenance respiration from the works of Spitters et al. (1989). The model is also based on the assumption that dry matter distribution is regulated by force or a similar uptake ability of the different plant organs. The developed model uses these substructures included in Marcelis et al. (2006) and further improved by the works of other authors. It consists of a set of five ordinary differential equations (ODE), algebraic equations, parameters and constants. The plant dry matter is estimated by the generation of carbohydrates from the photosynthesis process. The model shows the plant as a set of organs competing to access a common pool of assimilates. The remaining carbohydrates are available for

plant growth and development. Each organ growth rate is given by the amount of available carbohydrate and organ growth factor (Gonzalez-Real et al., 2008). As a first step, the leaf photosynthesis is obtained (Collatz et al., 1991; Teh, 2006). Then, the gross photosynthesis is calculated, using the leaf area index (LAI), which is an indicator that represents the leaf surface available for taking up solar energy; it is estimated using the Gompertz function (Scaife and Jones, 1976). The organs dry weight model presented by Gonzalez-Real et al. (2008) was used to address the problem of synchronous production oscillations in the sweet pepper crop. These oscillations are caused by the crop physiology (fruit abortion in periods of heavy fruit load per plant), resulting in an over-supply at market. Pruning, harvesting, aborting fruit and climate control are introduced to modify this pattern in the model behavior. It is a discrete time model describing the crop development in time steps throughout a day.

Fig. 2. Diagram of the proposed growth model.

146

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

Fig. 3 shows the process for obtaining the dry weight of the stem, leaves, roots and fruit from the greenhouse climate data. All the model equations are shown in the following sections: crop photosynthesis (Section 2.1) is estimated from leaf photosynthesis (Section 2.1.1) and leaf area index (Section 2.1.2). Maintenance respiration (Section 2.2) uses the different organs' dry weights to calculate the carbohydrate losses. Crop dry weight (Section 2.3) is calculated from photosynthesis and respiration, while the various organs' dry weight estimations (Section 2.4) are estimated from fractions of the total daily dry weight. In this work, the time step used for calculating crop photosynthesis and maintenance respiration is 1 min, while the step time for calculating dry matter production is one day. The dry matter is divided between the different plant organs according to growth factors. 2.1. Crop photosynthesis Crop photosynthesis (Pcrop) is obtained by integrating the leaf photosynthesis (Pleaf) rate per canopy leaf area index (LAI) over the day as follows: P crop ¼ P leaf LAI:

ð1Þ

This equation is carried out using the Gaussian integration method (Marcelis et al., 2006). It is a simple, rapid method that specifies the discrete points at which the value of the function to be integrated has to be calculated, along with the weighting factors that needs to be applied to these values to achieve minimum deviation from the analytical solution (Rabbinge et al., 1991). 2.1.1. Leaf photosynthesis Leaf photosynthesis (mol m−2 s−1) is calculated with the LAI and climatic variables. The equation of Marcelis et al. (2006), based in the works of Farquhar et al. (1980), is used to determine the CO2 assimilation by the Rubisco enzyme (ribulose-1,5-bisphosphate carboxylase oxygenase):
P leaf ¼ min V c ; V q ; V s :

ð2Þ

Therefore, three equations are shown, one for each variable (Vc, Vq, and Vs). The leaf photosynthesis will take the minimum value; this is determined by the most limiting factor at each moment (Eq. (2)). This enzyme Rubisco exhibits dual behavior, catalyzing two opposing processes. The first process, called carboxylation (Vc), is the fixation of CO2 to an organic form. In contrast, the plant O2 fixation process (photorespiration), is that through which a part of the assimilated carbon is lost. In Eq. (3), the assimilation of CO2 by the Rubisco enzyme is determined as: V c ¼ V cmax

C−Γ 
C þ K c 1 þ KO

ð3Þ

0

where Vcmax is the maximum carboxylation velocity of the enzyme (μmol m− 2 s− 1); C, represents intercellular concentration of CO2 (μmol mol−1); O is the partial pressure of O2 (μmol mol−1); and Γ⁎ is CO2 compensation point without dark respiration (μmol mol−1). The parameters Kc and Ko are the Michaelis–Menten constants for CO2 and O2 respectively, both in μmol mol−1. According to Collatz et al. (1991), in low light levels, CO2 assimilation is not limited by the Rubisco enzyme, and another (Eq. (4)) should be added to consider these cases (Vq), representing the photosynthesis process limitations caused by low-light levels: V q ¼ eq αI f

C−Γ C þ 2Γ

ð4Þ

where eq. is the light quantum efficiency (μmol mol−1); α is the leaf absorption for PAR (−); If is PAR flux incident on a unit leaf area (μmol m−2 s−1); and the intercellular CO2 concentration, C, is calculated from the CO2 concentration in the greenhouse. Besides the light limitations and the ability of the Rubisco enzyme, Collatz et al. (1991) included another limiting factor — the accumulation of photosynthetic products. Collatz et al. (1991) asserted that the production rate of a product decreases when the product concentration increases. This limitation was assumed simply by halving the capacity of the Rubisco enzyme (Vs, Teh, 2006): Vs ¼

Vc

max

2

:

ð5Þ

2.1.2. Leaf area index Leaf area index (LAI) is defined in the Eq. (6), the one-sided green leaf area per unit of ground area in broadleaf canopies; thus, it is dimensionless (m2[leaves] m−2 [ground]), typically ranging from 0 for bare ground to 6 for a dense area in greenhouses. LAI is employed to predict photosynthetic primary production and for use as a crop growth reference tool. Its calculation was made based on thermal time using the Gompertz growth function (Scaife and Jones, 1976). This is a sigmoid function to model time series where growth is slow at the beginning and at the end of the period:  X LAI ¼ X LAImax

 X LAIini e−C pLAI X TS X LAImax

ð6Þ

where X LAImax is the maximum value that LAI can take; X LAIini is the initial value of LAI; CpLAI is a tuning parameter of the Gompertz function; and finally, XTS, is the thermal time taken by the plant (°C day−1). This thermal time is calculated as the accumulative sum of the daily average temperature minus a base temperature of 10°. 2.2. Maintenance respiration

Fig. 3. Growth model outline.

Plant maintenance respiration (Rm,r) consumes part of the carbohydrates produced by photosynthesis in order to provide energy to the existing biostructures. The value of this respiration depends on the

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

temperature and the dry weights of different plant organs as set out by Marcelis et al. (2006): Rm;r ¼ 0:03W l þ 0:015W s þ 0:015W r þ 0:01W f

ð7Þ

where Rm,r is the maintenance respiration rate at a reference temperature (Tr) of 25 °C; and W is the dry weight of the organ in g m− 2, whose subscripts refer to leaves (Wl), stems (Ws), roots (Wr) and storage organs such as fruits (Wf) respectively. This equation is carried out hourly. Therefore, the maintenance respiration to temperature T is: Rm ðT Þ ¼ Rm;r 2ðT−T r Þ=10

ð8Þ

where Rm is the maintenance respiration (g CH2O m−2 h−1) at a reference temperature (Tr) of 25 °C. Moreover, maintenance respiration is assumed to be proportional to the fraction of the accumulated leaf weight. 2.3. Crop dry weight Once the crop photosynthesis and maintenance respiration are obtained, the total dry weight per m−2 is calculated using the sum of this hourly data this is done to obtain daily data (g CH2O m−2 day−1). Maintenance respiration is assumed to have priority over growth. The amount of assimilates needed for maintenance respiration (Rm) is subtracted from the assimilates produced by photosynthesis (Pcrop). The remaining assimilates are available for dry weight increment (dW/dt). A conversion factor for grams of CH2O to grams dry weight (g DW) is used. Eq. (9) describes the temporal variation of dry weight in each sample (Gonzalez-Real et al., 2008):
dW T ¼ f c P crop −Rm : dt

ð9Þ

Once the total dry weight is estimated, we can obtain the weight for each of the different plant organs. The daily total dry weight will be multiplied by the daily fraction of assimilates directed to each organ. Therefore, the sum of the four fractions always has to take a value of 1. 2.4. Organs dry weight Daily fruit dry weight (Wf; gDW m− 2 day− 1) and daily root dry weight (Wr; gDW m−2 day−1) are estimated by multiplying the daily total dry weight (dWT/dt) by the daily fraction of dry weight for fruits, and roots. This is represented by Eq. (10) for fruits dry weight and Eq. (11) for roots dry weight (Gonzalez-Real et al., 2008): dW f dt

¼ ff

dW T dt

ð10Þ

dWr dt

¼ fr

dW T dt

ð11Þ

under constant environmental conditions. These temporal variations in production can be approximated by a sinusoidal function, f, to determine the daily fraction of the total dry weight that is directed to the fruits and also to the roots. The values for this function for fruits and roots are different thus obtaining two functions which fit satisfactorily, following an opposite trend but a similar amplitude: f ¼ a þ b sin





2π ðdays þ cÞ d

¼ fl

dW T dt

ð12Þ

dW s dt

¼ fs

dW T : dt

ð13Þ

Gonzalez-Real et al. (2008, 2009) observed that the sweet pepper crop is characterized by periodic fluctuations in fruit production, even

ð14Þ

where the a parameter represents the average dry weight fraction in fruit for the entire production cycle. The b parameter is the amplitude of variation for the fraction values devoted to fruits and roots. The c parameter represents the sine phase (in days). Parameter d can be approximated as the fruit lifespan. Finally, the variable days corresponds to the number of days after planting. By assuming sinusoidal functions that are identical but opposite in phase to the fruits and roots implies that the sum of the fractions involved in stems and leaves is constant, and therefore, the sum of the four fractions (fruits, leaves, stems and roots) must be equal to 1, to fully distribute the weight between plant organs. Once the total dry weight of the plants is obtained, the organ dry weights are obtained by the fraction. Fig. 4 shows the dry weight fractions of the different organs involved. One can see that the fractions of fruits and roots are opposite yet equal in amplitude, regulated by the sinusoidal function f. In the initial period of the plants vegetative growth, dry matter is distributed only among the roots, stems, and leaves, due to the fruits growth being delayed. Fig. 5 shows the fractions for roots, leaves and stems ( fr, fl and fs). These fractions have been estimated as linear correlations dependent on the fraction of the fruits, omitting the plants initial phase of vegetative growth. As was expected, there is a strong negative correlation between the fraction devoted to fruit and roots. The fractions involved in leaves and stems approach linearly in the correct way, since, as noted above, data have to be adjusted so that the sum of these two fractions is constant. 3. Greenhouse and crop data The research data used in this work was located in the Experimental Station of Cajamar Foundation (El Ejido) at the province of Almeria in SE Spain (2° 43′ W, 36° 48′ N, and 151 m elevation). The crops were grown in two multitunnel greenhouses. Each greenhouse was 800 m2, with a crop area of 616 m2 with polyethylene cover, each had automated ventilation with lateral windows in the northern and southern walls, a flap roof window in each span, and mesh-protected anti-trip bionets, 20 × 10 in thickness. The orientation of the greenhouses was E–W; with crop rows aligned N–S. Cropping conditions, and crop management were very similar to those in commercial greenhouses. A meteorological station was installed outside the greenhouses, in which air temperature and relative humidity (Vaisala HMP60), solar

where WT is the total dry weight of the plant, and fr and ff are the fractions of the total daily dry weight to fruits and roots, respectively. The leaves (Wl) and stem (Ws) dry weights are calculated in the same way as fruits and roots. The fractions of total dry weight for leaves and stems (fl and fs) have been calculated as linear correlations, based on the fruit fraction, omitting the plants initial phase of vegetative growth, as follows: dW l dt

147

Fig. 4. Dry weight fraction for each plant organ.

148

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

4. Model validation The validation process included data taken in a third greenhouse used in the same time period, also located at the Experimental Station of the Cajamar Foundation, and comprising the same characteristics. 4.1. Leaf area index

Fig. 5. Linear correlation of the dry weight fraction for each plant organ and the fruit dry weight.

and photosynthetic active radiation (PAR, Par-lite), rain detection (Vaisala DRD11A), wind direction (Young 03002), and velocity measurements (Young 12102) were taken. During the experiments, some indoor climate variables were also taken: air temperature, relative humidity, solar radiation (LP PYRA 02), photosynthetic active radiation (PAR), CO2 monitoring (Vaisala GMP220), soil temperature at 3 cm and 40 cm (PT100), leaf and substrate temperature (Thermocouple), as well as electrical conductivity and pH monitoring (Model A1005) in irrigation and drainage water. All sensors were located in the center of the greenhouse; the psychrometer and CO2 sensor at a height of 2.5 m which was immediately above the mature crop, and the PAR and solar radiation sensors at a height of 3.5 m. All the actuators are driven by relays designed for this task. All climatic data was recorded every minute with a personal computer. The evolution of the leaf area index was determined using leaf area measurements of each plant removed for the biomass task; the dry matter of leaves and secondary stems pruning were also taken into account. The weight of biomass was measured using destructive sampling of five randomly selected plants every 21 days. In this process, the following measurements were taken: the number of nodes, the leaf area, the number of fruits per bunch, as well as the fresh and dry weight of leaves, stem, and fruits. The plant material and fruits were introduced into a drying oven where they remained for 24–48 h (depending on their phenological state) at a temperature of 65 °C. Based on this, the dry matter of the leaves, stems, and fruits were determined using an analytical balance. The leaf matter and secondary stem pruning used for biomass measurements came from the selected plants while remaining in production. Once removed from the plant, the fresh and dry weight was taken, in the same way as for the biomass. In this case, however, the stems and leaves were measured separately. Both the secondary stems and the leaves areas pruning were taken for the leaf area index measurements; they were carried out in the same way as for the biomass, using an electronic planimeter. Model parameters calibration was performed using data collected over one season of sweet pepper (Capsicum annuum L. var. Melchor) in two greenhouses at the Las Palmerillas experimental station of the Cajamar Foundation. The cycle began on the transplant date, 21/07/ 2010, and ended on 07/03/2011, thus lasting 229 days — with a total of 121, 287 samples per variable. The data from a third greenhouse were used for validation. The destructive measurements in the greenhouses were taken on days 28, 57, 113, 162, 196 and 229. On each day, four measurements for LAI and total dry weight were obtained. Hence, for the calibrating process, the arithmetic mean of the daily samples was calculated, and subsequently, a linear interpolation between the means obtained in the previous step was performed. The work space of these parameters was determined using an iterative sequential algorithm to minimize the least square error criterion between the real and the estimated dry weight results (Monte Carlo algorithms, Sbert et al., 1995). The second phase of the calibration process was based on genetic algorithms to fix the final parameters (Whitley et al., 1990).

Fig. 6(a) represents the LAI obtained by the model and the leaf area index data sampled from measurements taken in the greenhouse for six days during the season. The results correctly described the increasing LAI but included an over-estimation of 5%. The use of more calibration data could produce better adjustment of the Gompertz function parameters and thus better results. 4.2. Dry weight The model estimates the gross photosynthesis using the climate data, the LAI and the maintenance respiration in order to obtain the total dry weight for the day. Fig. 6(b) shows the total dry weight (g m− 2) results obtained for the validation season. Actual measurements were taken for four groups of greenhouse plants; these are shown along with the average value for these measurements. The results properly described the plant growth although presented an underestimation, especially in the central–end period of the season. This was expected, as it was also observed in the two calibration seasons. The fruit dry weight estimated by simulation is shown in Fig. 6(c). The figure represents the real measurements (four samples and the average value) and the estimated dry weight. Fig. 6(c) shows that an underestimation is obtained in the middle–end of the season, although the final simulated results fit to the real data. Using the fruit dry weight makes it possible to obtain an approximate value for greenhouse performance which serves as valuable data for growers. The leaf and stem dry weight results have not been validated, since no real measurements were available; only the final value, 350 g m−2. The estimated value for leaf dry weight obtained for the season was around 330 g m− 2, so this value was slightly underestimated. The final stem dry weight was 348 g m−2, and this was accurately estimated by the model (350 g m−2). 4.3. Fresh weight The fresh weight is the actual weight of the crop or fruit. It is the sum of the dry weight plus the water content of the element in question. Therefore, this is the main result for determining crop profitability. The water content in the plants and fruit is not constant. It will vary depending on the maturity, or stage of development, of the crop growth. The content of total fresh weight and fruit fresh weight based on the dry weight content throughout the season has been obtained empirically. Fig. 6(d) shows the results obtained by the simulation for the total fresh weight during the validation season. We observed an underestimation of the values for much of the season. Correct fresh weight values were only obtained at the first and last stages of the season. The same behavior occurred for the calculation of the fruit fresh weight (Fig. 6(e)). Table 1 shows a full errors review for each model. The goodness of fit for the leaf area index, the total and fruit dry weight, and the total and fruit fresh weight was obtained. This goodness of fit for a data series is calculated using the R2. An average error between 4 and 9% for fruit dry weight, fresh weight, LAI and total dry weight shows that the model calibration was successful. However, it also indicated certain problems such as an underestimation in the middle–end of the season. These errors could be caused by the lack of a bigger and richer group of calibration data.

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

149

Fig. 6. Model validation results.

Table 1 presents promising fruit dry weight results, which are the most important from an economic point of view. The dry weight values for leaves and stems are not very meaningful for the grower in deciding the greenhouse climate conditions.

Table 1 Model error validation. Model −2 (m[leaves] −2

LAI WT (g m ) Wf (g m−2)

−2 m[soil] )

Max error

Mean error

σ

R2

Interval

0.61 217.3 108.3

0.201 90.12 44.8

0.164 64.67 39.6

94.7 93.9 93.5

[0, 4.64] [0.65, 1542] [0, 913.6]

5. Decision-making software: utilization example The results obtained by the crop growth model can be useful in supporting growers to make decisions regarding greenhouse climate management in order to maintain optimal crop growth. Growers are starting to see the advantages of installing climate control systems. These tools are tedious and complex and growers have complained about the difficulty and amount of time it takes to carry out the tasks. The development of a decision support tool was realized to choose the right climate control set points to apply (Fig. 7(a)). The DSS is provided to help growers better understand the effect that different variables have on growth (Fig. 7(a)). The various options provided to introduce climate data make this DSS an interesting and valid tool

150

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

Fig. 7. Functionality of software tool.

for growers and researchers (Fig. 7(b)). These options allow users to try out different data combinations and to compare results (Fig. 7(c)). The tool displays valuable information for growers through graphics and result reports (Fig. 7(d)). The first step for the DSS is to define the simulation. In this stage, the main information is introduced about the greenhouse location and size, the crop season duration and density. For this example, four different strategies were tested (Table 2). The climate data was introduced from greenhouse data (GD), and the constant day and night temperature values or CO2 concentrations were modified in the greenhouse data file (Strategies 2, 3, and 4). The simulation is the last process in obtaining the growth data. Fig. 8(a), (b), (d), (c) shows the fruit fresh weight in each strategy tried. With these results, the dynamics of each strategies change in Table 2 Different strategies tried out. Variable

Strategy 1

Strategy 2

Strategy 3

Strategy 4

Temperature (°C) Radiation (w m−2) CO2 (ppm)

GD GD GD

GD GD 700

22/14 GD GD

22/14 GD 700

slope can be drawn even though they show similar behavior. In the end, Strategy 3 gave the best production results due to more favorable climate conditions for crop growth. The DSS can be implemented for a number of greenhouse decision problems such as crop growth management and harvest scheduling for assessing the environmental impact of the various climate regimes. Table 3 shows how modifying the climate conditions (temperature or CO2) leads to changes in final production. Moreover, greenhouses with CO2 enrichment, heating and natural ventilation exhibit better fruit-producing performance, reducing water consumption due to lower leaf transpiration and leaching fraction (Sánchez-Molina et al., 2014b). This translates into a closer relationship between the fruit weight and the total plant weight; even leaf growth (LAI) is comparatively less. Growers can study the effects of temperature or CO2 on fruit weight, as well as LAI and other crop development issues — so the need for balanced growth over the long run suggests that the desired integral value should be coupled to that of available light. Finally, it is possible to generate a summary report, in pdf format, which contains an abstract of the numeric values at the end of the season to determine the decision analysis and highlight the most significant factors in choosing the different alternatives. Following this, it is possible to export simulation data to file.

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

151

Fig. 8. Fruits fresh weight (g m−2).

6. Conclusion Research on the development of growth models for the sweet pepper crop remains scarce, when compared to other greenhouse crops such as tomatoes. In addition, a problem in production fluctuation was found. The model results presented an underestimation in the central periods of the simulated seasons. Nonetheless, the data produced by the model compared satisfactorily to available data. The model can correctly simulate dry matter production based on photosynthesis and partitioning between different plant organs. Similarly, the calibration and validation seasons belong to the same time period. This means that the climate data are very similar to each other, and therefore the models response to different climatic conditions is not known. The tool provides production results as a function of variations in climatic conditions in the greenhouse environment which allows one to optimize pepper growth control. The DSS was developed using object-

Table 3 Final results resume with four different strategies. Variable 2

−2

Leaf area index (m m ) Total dry weight (kg m−2) Fruits dry weight (kg m−2) Total fresh weight (kg m−2) Fruits fresh weight (kg m−2)

Strategy 1

Strategy 2

Strategy 3

Strategy 4

4.90 1.55 0.86 14.15 10.02

5.60 1.77 1.02 16.52 12.52

5.2 1.63 0.95 15.08 11.33

5.61 1.87 1.16 16.71 12.76

oriented methodology, which facilitates the reuse, modification or extension of each of its components. This paper presents a growth model for sweet pepper cultivation based on various studies by other authors. The model provides a good basis for further research and development in future. Furthermore, the software implementation of this model can be a useful tool for growers. This software can be used to better understand the pepper crop growth process both for teaching and research purposes as well as to optimize greenhouse production. It might be noted that we have incorporated a greenhouse climate model based on external conditions to evaluate the effect of different greenhouse structures on pepper growth. Moreover, it is possible to extend the tools use to other horticultural varieties such as cucumber and tomato. Future works A future line of work for the present study is to evaluate the model under controlled experimental conditions, including its use in salinity management. Furthermore, the model will be tried out on crops with similar agronomic characteristics such as cucumber. The Decision Support System will be implemented as a web-based tool for teaching purposes. Acknowledgments This research was funded by the Controlcrop Project, P10-TEP- 6174, project framework, supported by the Andalusian Ministry of Economy,

152

J.A. Sánchez-Molina et al. / Agricultural Systems 139 (2015) 144–152

Innovation and Science (Andalusia, Spain), and by the Spanish Ministry of Science and Innovation as well as from EUERDF funds under grant DPI2014-55932-C2-1-R. Appendix A. Variables and parameter values and units

Table A1 Model parameter outline. Symbol Units aF aR al fa b c d eq. fc Kc Ko O pLAI /tau(25) Vcmax

Description

[−]

Fruits DW average fraction [−] Roots DW average fraction [−] Leaf absorption for PAR [−] Fractions amplitude of variation [days] Sin phase [fruits, roots] [days] Fruits life cycle −1 [μmol mol ] Quantum light efficient [g PS/g CH2O] Conversion factor [μmol mol−1] CO2 Michaelis constant [μmol mol−1] O2 Michaelis constant −1 [μmol mol ] O2 partial pressure [−] Gompertz parameter μmol mol−1 [2600] [μmol mol−1 s−1] Max carboxylation vel., T = 25 °C

Value

Values reported

Variable

0.48

0.5

Wf,l,s

0.17

0.17

Wf,l,s

0.76

0.8

Pleaf

0.17

0.17

Wf,l,s

[60, 25]

[60, 22]

Wf,l,s

[79] [0.063]

[76] [0.06]

Wf,l,s Pleaf

[0.29] [300]

[0.28] [300]

Wf,l,s Pleaf

[300]

[300]

Pleaf

[21,000] [0.0018] [2600] [200]

[21,000] Pleaf [0.00161] LAI Pleaf [200] Pleaf

Table A2 Model variables resume. Symbol

Units

Description

Variable

C ff fl fr fs Γ⁎ If XLAI XLAImax XLAIini Pcrop Pleaf Rm,r Rm TS Vc Vq Wf,l,r,s,T

[μmol mol−1] [−] [−] [−] [−] [μmol mol−1] [μmol mol−1 s−1] [m2 m−2] [m2 m−2] [m2 m−2] [g CH2O m2 h−1] [μmol mol−1] [g CH2O m2 h−1] [g CH2O m2 h−1] [°C day] [μmol mol−1 s−1] [μmol mol−1 s−1] [g DW]

Intercellular [CO2] Dry weight fruits Dry weight leaves Dry weight roots Dry weight stems CO2 compensation point PAR flux incident LAI LAI initial LAI max Crop photosynthesis Pleaf Rm, T = 25 °C Maintenance respiration Thermal time CO2 assimilation Low light levels Pleaf Fruits, leaves, roots, stems, total DW

Pleaf Wf Wl Wr Ws Pleaf Pleaf LAI LAI LAI [−] Pleaf [−] [−] Rm,r Pleaf Pleaf [−]

References Aubert, B.A., Schroeder, A., Grimaudo, J., 2012. IT as enabler of sustainable farming: an empirical analysis of farmers' adoption decision of precision agriculture technology. Decis. Support. Syst. 54, 510–520. Berenguel, M., Rodríguez, F., Guzmán, J., Lacasa, D., Pérez-Parra, J., 2006. Greenhouse diurnal temperature control with natural ventilation based on empirical models. Acta Horticult. 719, 57–64.

Castelán-Ortega, O.A., Fawcett, R.H., Arriaga-Jordán, C., Herrero, M., 2003. A decision support system for smallholder campesino maize–cattle production systems of the Toluca Valley in central Mexico. Part I—integrating biological and socio-economic models into a holistic system. Agric. Syst. 75, 1–21. Chevalier, R.F., Hoogenboom, G., McClendon, R.W., Paz, J., 2012. A web-based fuzzy expert system for frost warnings in horticultural crops. Environ. Model Softw. 35, 84–91. Collatz, C.G., Ball, J., Berry, J., 1991. Physiological and environmental regulation of stomatal conductance, photosynthesis and transpiration: a model that includes a laminar boundary layer. Agric. For. Meteorol. 54, 107–136. Ewert, F., Rodriguez, D., Jamieson, P., Semenov, M., Mitchell, R., Goudriaan, J., Porter, J., Kimball Jr., B., P.P., Manderscheid, R., Weigel, H., Fangmeier, A., Fereres, E., Villalobos, F., 2002. Effects of elevated {CO2} and drought on wheat: testing crop simulation models for different experimental and climatic conditions. Agric. Ecosyst. Environ. 93, 249–266. Farquhar, G., Caemmerer, S., Berry, J., 1980. A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta 1, 78–90. Gijzen, H., 1994. Development of a Simulation Model for Transpiration and Water Uptake and an Integral Growth Model (Master's thesis) AB-DLO, Wageningen. Gonzalez-Real, M., Baille, A., Liu, H., 2008. Influence of fruit load on dry matter and ndistribution in sweet pepper plants. Sci. Hortic. 117, 307–315. Gonzalez-Real, M., Liu, H.Q., Baille, A., 2009. Influence of fruit sink strength on the distribution of leaf photosynthetic traits in fruit-bearing shoots of pepper plants (Capsicum annuum L.). Environ. Exp. Bot. 66, 195–202. Goudriaan, J., Laar, H.V., 1994. Modelling Potential Crop Growth Processes. Kluwer Academic Publishers, Dordrecht, The Netherlands. Heuvelink, E., 1999. Evaluation of a dynamic simulation model for tomato crop growth and development. Ann. Bot. 83, 413–422. Hoch, T., Agabriel, J., 2004. A mechanistic dynamic model to estimate beef cattle growth and body composition: 1. Model description. Agric. Syst. 81, 1–15. Jame, Y.W., Cutforth, H.W., 1996. Crop growth models for decision support systems. Can. J. Plant Sci. 76, 9–19. Marcelis, L., 1994. A simulation model for dry matter partitioning in cucumber. Ann. Bot. 74, 43–52. Marcelis, L., Heuvelink, E., Goudriaan, J., 1998. Modelling biomass production and yield of horticultural crops: a review. Sci. Hortic. 74, 83–111. Marcelis, L., Elings, A., Bakker, M., Brajeul, E., Dieleman, J., de Visser, P., Heuvelink, E., 2006. Modelling dry matter production and partitioning in sweet pepper. Acta Horticult. 718, 121–128. Nederhoff, E., Rijsdijk, A., de Graaf, R., 1992. Leaf conductance and rate of crop transpiration of greenhouse grown sweet pepper (Capsicum annuum L.) as affected by carbon dioxide. Sci. Hortic. 52, 283–301. Rabbinge, R., Ward, S., Laar, H.V., 1991. Simulation and systems management in crop protection. Plant Growth Regul. 10, 169–170. Ramirez-Arias, A., Rodriguez, F., Guzman, J., guel, M., 2012. Multiobjective hierarchical control architecture for greenhouse crop growth. Automatica 48, 490–498. Rodriguez, F., Guzman, J.L., Berenguel, M., Arahal, M.R., 2008. Adaptive hierarchical control of greenhouse crop production. Int. J. Adapt Control 22, 180–197. Rodríguez, F., Berenguel, M., Guzmán, J., Ramírez-Arias, A., 2015. Modeling and Control of Greenhouse Crop Growth. Springer International Publishing, Switzerland. Sanchez, J.A., Rodriguez, F., Guzman, J.L., Arahal, M.R., 2012. Virtual sensors for designing irrigation controllers in greenhouses. Sens. Basel 12, 15244–15266. Sánchez-Molina, J., Reinoso, J., Acién, F., Rodríguez, F., López, J., 2014a. Development of a biomass-based system for nocturnal temperature and diurnal CO2 concentration control in greenhouses. Biomass Bioenerg. 67, 60–71. Sánchez-Molina, J., Rodríguez, F., Guzmán, J., Ramírez-Arias, J., 2014b. Water content virtual sensor for tomatoes in coconut coir substrate for irrigation control design. Agric. Water Manag. 151, 114–125. Sbert, M., Perez, F., Pueyo, X., 1995. Global Monte Carlo. A Progressive Solution, in: Hanrahan, P., Purgathofer, W. (Eds.), Rendering Techniques '95. Springer Vienna. Eurographics, pp. 231–239. Scaife, M., Jones, D., 1976. The relationship between crop yield (or mean plant weight) of lettuce and plant density, length of growing period, and initial plant weight. J. Agric. Sci. (Camb.) 86, 83–91. Spitters, C., Keulen, H.V., Kraalingen, D.V., 1989. A simple and universal crop growth model simulator: Sucros87. Simul. Monogr. 32, 147–181. Teh, C., 2006. Introduction to Mathematical Modeling of Crop Growth: How the Equations are Derived and Assembled Into a Computer Program. BrownWalker Press. Whitley, D., Starkweather, T., Bogart, C., 1990. Genetic algorithms and neural networks: optimizing connections and connectivity. Parallel Comput. 14, 347–361. Wolf, J., van Oijen, M., Kempenaar, C., 2002. Analysis of the experimental variability in wheat responses to elevated {CO2} and temperature. Agric. Ecosyst. Environ. 93, 227–247.