GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
Wireless Sensor Networks to Precisely Monitor Substrate Moisture and Electrical Conductivity Dynamics in a Cut-Flower Greenhouse Operation John D. Lea-Cox 1, Félix Arguedas Rodriguez 2 1 University of Maryland, Department of Plant Science and Landscape Architecture, 2120 Plant Sciences Building, MD 20742-4452. USA.
[email protected] 2
[email protected] Andrew G. Ristvey 3 University of Maryland, Wye Research and Education Centre, 124 Wye Narrows Drive, Queenstown, MD 21658. USA.
[email protected] David S. Ross 4 University of Maryland, Department of Environmental Science and Technology, 1431 Animal Sciences/Agric. Eng. Building, MD 20742-2315. USA.
[email protected] George Kantor 5 Carnegie Mellon University Robotics Institute, 5000 Forbes Ave., Pittsburgh, PA 15213.
[email protected] Keywords: irrigation, scheduling, matric potential, Ech20 sensors Abstract The precise monitoring and control of irrigation and electrical conductivity in realtime has been a goal of many greenhouse producers for many years, but most technologies are either extremely expensive and/or do not have the precision for use in porous soilless substrates. We have deployed a low-cost wireless sensor network, developed by Carnegie Mellon University, in a commercial cut-flower operation in Maryland. This sensor network is monitoring the water content and electrical conductivity of the solution in a porous perlite substrate, using capacitance (Echo20 EC-5, and 5TE) sensors from Decagon Devices (Pullman, WA). The challenge is to accurately measure spatial and temporal dynamics in real-time, to ensure precision irrigation and nutrient applications and improve the uniformity and quality of cut-flower Antirrhinum cultivars produced by this operation. We are working towards a full integration of these control capabilities with the grower, so that he can fertigate based upon substrate matric potential or electrical conductivity setpoints with this re-circulating hydroponic system. INTRODUCTION We have previously outlined the general requirements for deploying sensor networks in intensive nursery and greenhouse operations (Lea-Cox, Ristvey et al., 2008; Lea-Cox et al., 2009), recognizing the nature and the complexity of these operations. Greenhouse environments are especially challenging, given the rapid environmental changes that can occur on a daily basis within the greenhouse microclimate, despite good control of temperature, ambient light and relative humidity (RH). We are testing moisture and electrical conductivity sensors in a commercial greenhouse facility near Baltimore, Maryland, to aid better irrigation and nutrient management decisions. This cut-flower operation specializes in snapdragon – Antirrhinum majus (L.), growing plants hydroponically using a perlite substrate in bags. Horticultural soilless substrates generally have low water-holding capacities and smaller ranges of easily-available water (EAW) for optimum plant growth (deBoodt and Verdonck, 1972). Easily-available water can be sensed and measured, based on the matric potential (Ψm) of the substrate. Soilless substrates generally hold EAW in a Ψm range from 0 to -10 kPa, with the majority of free water available from 0 to -5 kPa (deBoodt and Verdonck, 1972). This Ψm is
1
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
between 10 and 100 times lower than similar plant-available water tensions in soils, which means that that standard method of using water tension to measure soil moisture content does not apply well to soilless substrates. We have shown (Arguedas-Rodriguez et al, 2007a; Arguedas-Rodriguez et al, 2007b) that easily-available water in soilless substrates is limited, with this source of perlite having one of the lowest water contents at Ψm between -1 and -10 kPa (Table 1). Scoggins and van Iersel (2006) compared the performance of four commercially-available EC sensors in several soilless substrates, but did not include the Decagon 5TE sensor. None of the tested sensors had the ability to provide real-time data (i.e., they were stand-alone sensors). EC measured with these probes was highly correlated with the three standard substrate extraction (pour-through, 1:2 dilution and saturated media extract) methods over a range of fertilizer concentrations. This indicates that in situ EC sensors can provide an accurate measure of nutrient availability. In addition, sensing EC in the root zone in real-time would provide specific data for individual production beds, which can be used to precisely monitor plant nutrient requirements, rather than the imprecise single EC (return tank) data that is currently used. GREENHOUSE OPERATION USED IN THIS STUDY This operation is in continuous production (Fig. 1), growing 3 to 4 full cycles of various cultivars during each year, based on their photoperiodic responses. Supplementary light was not used to extend the daylength, for energy and cost considerations. Temperature and relative humidity control was achieved through a combination of passive venting of the greenhouse using a Gower’s Choice control system (HortiMaX, Rancho Santa Margarita, CA, USA), combined with the use of fogging and fan venting, when necessary. Horizontal air-flow fans were used for uniform mixing of air within the greenhouse at all times. Fogging was automatically initiated for 10 seconds when RH dropped below 65%. Heating was controlled by use of thermostats. The snapdragons were grown in horticultural grade perlite (Pennsylvania Perlite Corporation, Bethlehem, PA) in 3.6 m long x 0.3 m wide bags, which are planted at a density of 98 plants / m2 (Fig. 1). Average plant height at maturity was approximately 1.2 m, including a prime quality flower spike (#1), which has to be greater than 30 cm in length with at least 12 florets. This operation utilizes a closed (re-circulating) hydroponic system to fertigate the plants at least 8 times per day. Conventionally, fertigation was scheduled based on the accumulation of total radiation (J•m-2), to ensure fertigation events occur at least once per hour from 6 am onwards. This fertigation schedule did not operate past 3 pm, in order for the substrate and plant microclimate to dry down each afternoon, for good disease control. OBJECTIVES The above mentioned commercial operation performs optimally during most of the year, but floral yield and quality (percentage #1 cuts) typically declines in the warmest summer months between June and August, perhaps due to incipient (unmeasured) water stress in the mature crop at this time. Given the limited rooting volumes in the bag, the plant density, a 15hour daylength during summer and the absence of nutrient deficiencies in the crop, it is likely that diurnal water relations are implicated in this yield decline. The primary objectives for this ongoing study are therefore to: 1. provide real-time data on substrate moisture and electrical conductivity (EC), for better management decisions on a daily basis; 2. automatically schedule irrigations based on the measurement of plant-available moisture (matric potential, Ψm) in perlite substrate;
2
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
3. satisfy diurnal crop water requirements on a day-to-day basis, given the increasing demand for water by the developing crop, over time, and 4. provide in situ substrate EC data, to ensure that nutritional requirements are being met by the hydroponic system. RESULTS AND DISCUSSION Sensor Network Development. We have previously reported on the deployment and performance of two wireless sensor networks that we tested in three different production environments (Lea-Cox, Black et al., 2008; Lea-Cox, Ristvey et al., 2008). One network was commercially available from Decagon Devices (not shown); the other (Fig. 2) was a non-commercial research network developed by the Carnegie Mellon Robotics Institute (Zhang et al., 2004; Lea-Cox, Ristvey et al., 2008). Each sensor network consisted of a system of radio-powered “nodes” that were deployed in a plant production area, to which a number of environmental sensors were connected. Any combination of soil moisture and electrical conductivity sensors, soil and air temperature, relative humidity, tipping rain gauge and light (photosynthetically-active radiation) sensors can be connected to the radio nodes, according to the specific sensing requirements of the grower. The nodes monitored data on a per minute basis, and logged the average data every 5 minutes, to conserve battery life and memory. The accumulated data was then transmitted at 900 MHz or 2.4 GHz using a battery operated radio card to a ‘base radio station’, whenever required. The base station was connected to a computer, which used custom software to plot and display the information from each of the nodes (Fig. 2; inset). In this way, a grower can develop a network of sensors that allows for the monitoring of environmental data in the greenhouse, in real time. The advantages of these networks are obvious – they provide information at the “micro-scale” which can be expanded to any resolution for a specific operation, for specific needs. Measuring Real-Time Plant-Available Water. Substrate matric potential is the critical measurement for plant water uptake (Jones, 2008). In order for us to understand physiological responses to water deficits, the volumetric water content has to be related to matric potential, with different calibrations being required for each substrate (Jones, 2008). For this reason, Arguedas-Rodriguez (2009) related the volumetric water content to the output of Ech20 EC-5 and 5TE sensors and simultaneously generated desorption curves, using a tension table. Consequently, we have been able to show that the range of plant-available water (0 to - 40 kPa) was much lower in soilless substrates than in soils (data not shown) and more especially, this range of available water was very limited in perlite, coir and 100% pine bark, compared to peat-based substrates (Fig. 3). Although it is likely that mature plant root systems can extract substrate moisture at lower Ψm than this, Kiehl et al. (1992) and others have found significant growth reductions at substrate Ψm as low as -16 kPa. This has significant implications for the measurement and automatic control of irrigation in these substrates, as this represents a very narrow range for scheduling irrigations in these substrates, especially when plant root : shoot ratios are as small as in this particular cut-flower operation. Measuring Real-Time Electrical Conductivity. Many sensors are now available that are apparently capable of reading instantaneous EC in soilless substrates, but the measurement of real time EC is not trivial (Arguedas-Rodriguez, 2009). Most sensors measure bulk EC (ECa), which is the total electrical conductivity associated with the surface soil / substrate ionic charge plus the ions in solution. Bulk EC measurements
3
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
therefore overestimate the available ions in solution, measured as pore water electrical conductivity (p). Obviously, the amount of water contained in the substrate at any one time has a profound effect on the concentration of salts in solution, and vice versa (Inoue et al, 2008). Temperature also has an effect on ionic concentration, requiring that sensors have a temperature compensation ability for precise measurement (Scoggins and van Iersel, 2006). It is therefore evident that a sensor needs the capability to simultaneously measure three variables – water content, temperature and ECa – to provide precise measurement of pore water EC. ArguedasRodriguez et al., 2009 recently showed that as a commercial greenhouse substrate (Sunshine LC1; 80 Peat: 20 perlite) loses moisture, the precision of both Ech20-TE and 5-TE EC sensors decreases. We think that this is primarily related to increasing proportions of air in the substrate, which increases the error in measuring pore water EC. This has an important practical implication as noted by Scoggins and van Iersel (2006) in that it is likely that these sensors will not be very accurate below a substrate moisture water content of about 35%. However, we should note that these moisture levels in this substrate equates to a matric potential of -15 to -20 kPa (Arguedas-Rodriguez et al., 2009), which is less than the accepted soil moisture ‘set point’ of -10 kPa for the optimal growth of plants in soilless substrates. It therefore seems likely that with placement of these sensors in the more saturated zone at the base of containers that reliable real-time EC data will be possible, at least in soilless substrates with higher proportions of peat. Database and Graphic User Interface Development, for Precision Irrigation and Nutrient Management. Visualizing large data sets that vary both temporally and spatially is a challenge. A sixnode, 60-sensor Ech20 5TE network can theoretically generate 180 data points per minute (648,000 data points per day). Given this type of information overload, it is imperative that we develop software programs that include more robust database capabilities, combined with userfriendly graphical interfaces. The database must provide the statistical backbone that can generate real-time averages for both monitoring and control purposes (Fig. 2). These systems also need to be web-enabled, so that employees can access sensor data with hand-held devices in the greenhouse, using the same wireless networks that transmit the data to the office computer (server). More importantly, we need to connect our management capability for precision water applications with a more precise knowledge of actual plant water use. We think that modeling plant water use for indicator species (Kim and van Iersel, 2009) will be essential to provide feedforward prediction capabilities, to enable advanced irrigation scheduling. CONCLUSIONS We have implemented a 6-node wireless sensor network for the monitoring substrate moisture and EC in the root zone. In future experiments, we will monitor irrigation frequency, nutrient use, crop growth and floral quality and yield, to compare to previously obtained data. We plan an economic cost-benefit analysis, to provide information about the savings and utility that this sensor network can provide small greenhouse growers. Ultimately, we plan to expand this to a 28-node sensor network, so that we can automatically control irrigation and electrical conductivity for each production zone. Irrigation events can already be independently monitored and precisely controlled by the Carnegie Mellon nodes, though a micro-pulse subroutine (LeaCox et al., 2009). The control of EC will require the future integration with the existing EC control system; at this point, the current EC control is probably adequate for crop development, given that the operation is relatively small and the recirculating hydroponic system allows for continuous fertigation events throughout the day.
4
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
ACKNOWLEDGEMENTS We gratefully acknowledge funding from the Chesapeake Bay Trust, the American Nursery and Landscape Association - Horticultural Research Institute and the University of Maryland Agricultural Experiment Station, who supported this research. Literature Cited Arguedas-Rodriguez, F.R. 2009. Calibrating Capacitance Sensors to Measure Water Content, Matric Potential and Electrical Conductivity in Soilless Substrates. MS Thesis, University of Maryland, College Park. 84p. Arguedas-Rodriguez, F.R., Lea-Cox, J.D. and Ristvey, A.G. 2009. Real-Time Measurement of Electrical Conductivity in Soilless Substrates. Proc. Southern Nursery Assoc. Res. Conf. 54:216-220. Arguedas-Rodriguez, F.R, Lea-Cox, J.D., and Ristvey, A.G. 2007a. Revisiting the Measurement of Plant Available Water in Soilless Substrates. Proc. Southern Nursery Assoc. Res. Conf. 52:111-115. Arguedas-Rodriguez, F.R., Lea-Cox, J.D. and Ristvey, A.G., 2007b. Characterizing Air and Water Content of Soilless Substrates to Optimize Root Growth. Comb. Proc. Int. Pl. Prop. Soc. 57:103-110. deBoodt, M. and Verdonck, O. 1972. The Physical Properties of Substrates in Horticulture. Acta Hort. 26:37-44. Inoue, M., Ould Ahmed, B.A., Saito, T., and Irshad, M. 2008. Comparison of Twelve Dielectric Moisture Probes for Soil Water Measurement under Saline Conditions. Am. J. Environ. Sci. 4: 367-372. Jones, H.G. 2008. Irrigation Scheduling – Comparison of Soil, Plant and Atmosphere Monitoring Approaches. Acta Hort. 792: 391-403. Kiehl, P. A., Liel, J.H., and Buerger, D. W. 1992. Growth response of Chrysanthemum to various container medium moisture tensions levels. J. Amer. Soc. Hort. Sci. 117(2):224-229. Kim, J. and van Iersel, M.W. 2009. Daily water use of abutilon and lantana at various substrate water contents. Proc. Southern Nursery Assoc. Res. Conf., 54: (In Press) Lea-Cox, J. D., Ristvey, A. G., Arguedas Rodriguez, F. R., Ross, D. S., Anhalt, J. and Kantor, G.F. 2008. A Low-cost Multihop Wireless Sensor Network, Enabling Real-Time Management of Environmental Data for the Greenhouse and Nursery Industry. Acta Hort. 801: 523-529 Lea-Cox, J. D., Black, S., Ristvey, A.G. and Ross, D. S. 2008. Towards Precision Scheduling of Water and Nutrient Applications, Utilizing a Wireless Sensor Network on an Ornamental Tree Farm. Proc. Southern Nursery Assoc. Res. Conf. 53:32-37. Lea-Cox, J. D., Kantor, G. F. and Ristvey, A. G. 2009. Wireless Water Management –Using wireless sensor networks for improved, cost-effective irrigation management in nurseries and greenhouses. Amer. Nurseryman. Jan, 2009. pp. 44-47. http://www.amerinursery.com Leith, J.H. and Burger, D.W. 1989. Growth of Chrysanthemum using an irrigation system controlled by soil moisture tension. J. Amer. Soc. Hort. Sci. 114: 387-397. Scoggins, H.L. and van Iersel, M.W. 2006. In Situ Probes for Measurement of Electrical Conductivity of Soilless Substrates: Effects of Temperature and Substrate Moisture Content. HortScience 41:210-214. Zhang, W., Kantor, G. F. and Singh, S. 2004. Integrated wireless sensor/actuator networks in an agricultural application. In: Proc. of ACM SenSys, Baltimore, Maryland, pp. 314–317.
5
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
Fig 1. Greenhouse production of Antirrhinum, showing leaf area and floral spike, maintained by continuous fertigation of the perlite bag (total rooting volume = 0.177 m3; 114 plants per bag). Table 1. Mean (n=10) bulk density (BD), container capacity (CC), total porosity (TP), air space (AS) and mean (n=30) easily-available water (EAW), water buffering capacity (WBC) and progressively unavailable water (PUW) of five commercial soilless substrates in 5-cm high columns. 100% Perlite
80 Pine Bark : 20 Peat
100% Coir
100% Pine Bark
80 Peat: 20 Perlite
0.114
0.204
0.086
0.218
0.106
(mL)
349
517
341
419
534
TP
(%)
62.2
77.6
88.5
71.5
80.0
AS
(%)
11.7
2.1
38.6
10.2
2.6
BD
(g • cm-3)
CC †
Pressure (kPa)
Distribution of Water (%)
EAW
(1 to 5)
36.0
40.0
32.6
34.6
43.7
WBC
(5 to 10)
1.2
7.0
2.1
2.2
13.1
PUW
( >10 )
62.8
53.0
65.3
63.2
43.2
†
Total volume of the 5-cm column = 684 mL. Note that CC = TP - AS. Use CC values to interconvert data.
6
GreenSys 2009: ISHS Spec. Conf. High Technology for Greenhouse Systems Management. June, 2009. Quebec, Canada. Acta Hort (Accepted)
Fig. 2. The wireless sensor network, showing a Carnegie Mellon node (at left), with embedded Ech20-TE EC sensor (insert), and Sensorweb graphic user interface in the office.
Fig. 3. Volumetric water content (θ) vs. output (mV) for Ech20-5cm sensors, with fitted regressions for five commercial soilless substrates. Matric potential data for the 100% perlite substrate illustrates the small range of PAW, in comparison to other soilless substrates containing peat. All regressions P