EFFECTIVENESS OF GRANULAR MATRIX SENSORS IN DIFFERENT IRRIGATION TREATMENTS AND INSTALLATION DEPTHS

: Continuous monitoring of soil moisture content plays a key role in irrigation scheduling and yield formation. This study was conducted to derive the technique and efficiency of application of granular matrix sensors (GMSs) in a sprinkler irrigation system of maize ( Zea mays L.). Two irrigation (a2 = 60%– 100% of field capacity (FC), a3 = 80%–100% of FC) treatments were imposed during two growing seasons (2010, 2012) and compared with rainfed control plots (a1). GMSs are used as an indirect method for monitoring soil water status at two depths (b1 = 15 cm and b2 = 30 cm) in order to make a decision on when to irrigate. The sensors used in this study were calibrated using a mass-based gravimetric method. In both growing seasons, irrigation treatment and installation depths have a significant influence (P<0.01) on soil water content. Sensors have shown a satisfactory response to wetting and drying periods in irrigation scheduling at 30-cm depth. Yet, due to variability of weather conditions, a slow response to wetting and drying was recorded in periods with intensive rainfall events (2010) and drought conditions with frequent irrigation intervals (2012). made for watermark readings and amount of irrigation water at each irrigation treatment. The analysis shows a weak negative correlation (r= -0.39** (2010), r= -0.29** (2012)) between soil water potential and a2


Introduction
In the context of water management for irrigation, measuring and monitoring of soil water status are an essential component of the best management practices to improve the sustainability of agriculture (Muñoz-Carpena, 2014). In addition to the direct weighing method (gravimetric method), soil water and moisture profiles can be obtained with numerous indirect methods where sensors are placed in the soil at different depths (Michot et al., 2003). An ideal soil water sensor would respond instantaneously to changes in soil water content or potential, and would be inexpensive, reliable, maintenance free, accurate within the needed range, and would produce an electrical signal suitable for electronic measurements, analysis and control (McCann et al., 1992). Depending on the soil physical properties some sensors are more reliable and effective than others . McCann et al. (1992) claim that soil moisture device such as Granular Matrix Sensor (GMS,Watermak 200SS) has all desirable qualities such as low cost and maintenance, and most important, sufficient accuracy in the 0-100 kPa range.
The GMS measures soil moisture that can be converted to soil water potential (Ψ soil ) by using a different calibration formula provided in the literature or calibrating them for a specific soil type (Intrigliolo and Castel, 2004). The sensor consists of electrodes embedded in a granular quartz material, surrounded by a synthetic membrane and protective stainless steel mesh. Inside, gypsum is used to buffer against salinity effects (Muñoz-Carpena, 2014;Perea et al., 2013). Soil moisture block is made of gypsum (gypsum block) and it consists of two electrodes enclosed in the block. Electrodes are connected to wires, which are extended to the soil surface. Wires are one meter long and they provide installation to different layers in soil. Certain disadvantages of GMS are previously described by Intrigliolo and Castel (2004) and Irmak and Haman (2001). They state that the GMS does not respond to changes at soil water potential higher than -10 kPa, and therefore it may not be a suitable tool in those cases where irrigation practices maintain a low soil tension. Also, McCann et al. (1992) have pointed out that the GMS sensor does not respond properly to rapid drying or partial rewetting of the soil, which may lead to incorrect estimation of the actual soil water status.
In spite of all limitations, the GMS sensor could be a useful tool when a relative indication of soil water content is needed, which is confirmed in previously published research of Josipović et al. (2010, 2011), Marković et al. (2012, 2014and 2015 and Marković (2013). Authors claim that yields of summer crops in average climatic years were increased from 13% to 40% in dry growing seasons, when irrigation scheduling was based on measuring of soil moisture with Watermark sensors.
The present study was designed to: derive a soil specific calibration of the GMS, present the technique of application and evaluate the efficiency in irrigation scheduling during two growing seasons (2010 and 2012).

Site description
The 2-year (2010, 2012) study was carried out at the research station of Agricultural Institute in Osijek (45°32`N, 18°44`E, and elevation 88 m). The experimental site is located in eastern Republic of Croatia characterized by moderate continental climate. The soil at experimental site is hydromeliorated hypogley, with a silty clayey loamy texture. Main soil physical properties Effectiveness of granular matrix sensors in different irrigation treatments 259 (Josipović, 2007;Marković 2013) are presented in Table 1. Studied soil horizons are slightly porous (41.8%), with moderate soil water capacity (36.6% to 37.1%) and small air capacity (5.3% to 6.2%). The texture of horizons is silty clay loam with clay content from 32.5 % in arable to 31.3% in sub-arable horizons. A high content of clay at the experimental plot reduces porosity and air capacity. The irrigation water had an average EC of 0.96 dS m -1 , pH 7 and Clcontent 29 mg l -1 . The weather data were obtained from an Automatic Weather Station (AWS) of national Meteorological and Hydrological Service located 1.5 km from the plot location and included rainfall amount (mm) and soil temperature ( o C). At the experimental site, the groundwater level was measured in an observation well using a piezometer (Marković et al., 2015). Observations were carried out once a week during the growing season in both years of the study.

Irrigation scheduling
The experiment had three irrigation treatments (a1, a2 and a3) and three replicates in a randomized complete block design. Irrigation scheduling was based upon the method which recommends irrigating when the available soil moisture is depleted to an allowable depletion. The depletion is expressed as a percentage of available water or field capacity (FC): a1 = rainfed condition; a2 = 60% to 100% of field capacity (FC); a3 = 80% to 100% of FC. Irrigation was performed with a traveling sprinkler system. The amount of water added in one irrigation interval was 35 mm. Water for this system was pumped from the well 37 m deep and 5 to 7 Ls -1 flow rate by using an electric pump (5.5 kW). The total size of the irrigation plot was 235 m 2 plus side borders which prevented overlapping of water from different irrigation treatments. The soil moisture content was measured with two granular matrix sensors (GMS, Watermark model 200SS, Irrometer Company, Inc.) per irrigation treatment (18 GMS in total). Prior to installation, field sensors were soaked overnight for three days and dried during the day. Wet sensors were installed in soil at two depths (15 cm and 30 cm in the plant row). Soil probes were used to make a hole to the desired installation depth. Sensors were installed before maize planting and stayed in the soil until the harvesting time. Soil water content was measured with a digital soil moisture meter (Watermark, Irromenter Company, Inc., Riverside, California). Measuring of soil water content was performed twice a week or after irrigation interval and a significant rainfall event until the end of August. The position of a sensor and the end of each row containing the sensor were marked.

Calibration of Watermark sensors
Sensors for measuring soil water status must be calibrated to provide actual soil moisture content . In our study, the calibration of the GMS was based on soil water content measured by the gravimetric method. Ten soil samples were taken in Kopecki's cylinders with a volume of 100 cm 3 from the experimental plot. Watermark sensors were placed in cylinders and in controlled laboratory settings. In the beginning, gross weight, cylinder weight, soil sample and sensor weight were recorded. Soil samples with the GMS were wetted with distilled water and weighted on an analytical balance. During wetting and drying, soil moisture was measured with a Watermark hand-held device. Sensor readings and daily gross weight were recorded until the readings of the majority of sensors were 199 kPa (dry soil). Then the soil samples were oven-dried (105 o C) and weighted again. The calibration process was repeated in the 2012 growing season. Results of the calibration are presented as calibration curves (Figure 1) with given regression equations. As it can be seen from Figure 1, there is some variability between individual sensors. Since this could be a problem in the field, it is recommended to use average readings of a number of sensors. According to some previous results of McCann et al. (1992) and Wang (1988), the variation in soil moisture of sensors increased with increasing dryness. According to results of the calibration process, available water holding capacity (AWHC) was 37%, vol. On average, 20 readings were taken (Table 2) and they ranged from 0 to 199 kPa where 0 stands for 100% FWC while 199 stands for dry soil. Average soil water tension in the first calibration process was 81.43 kPa and 84.51 kPa in the second. The correlation coefficient between GMS readings and gravimetrical soil water content in the first calibration process was very significant, complete and negative (r = -0.91**), while in the second year, it was very significant, very strong and negative (r = -0.86**).

Data analysis
The statistical analyses of data were conducted using the STATISTICA (StatSoft, Inc., 2011 -data analysis software system, version 10.0) and included analysis of variance (ANOVA), regression and correlation analyses. Soil moisture contents for each irrigation treatment and growing season are presented in graphical form.

Results and Discussion
Seasonal dynamics of soil water potential (kPa) in different irrigation treatments and at GMS installation depths are presented in Table 3. During the first year of the study, on average, soil water potential ranged from -49.9 kPa in rainfed plots to -29.5 kPa in a3 irrigated plots. As for soil water potential at different installation depths, the soil water potential was for 39% lower at the 30-cm depth in comparison to the 15-cm depth (15 cm = 15.9 kPa; 30 cm = 31 kPa). Both irrigation treatments as well as GMS installation depths had a significant impact (P<0.01) on soil water potential (Table 3). At 15-cm depth, both irrigation treatments (a2 and a3) significantly (P>0.01) increased soil water content, while at 30-cm GSM installation depth, only a3 irrigation treatment had a significant impact on soil water status (Table 3).  McCann et al. (1992) claim that soil water potential in an irrigated environment changes more rapidly closer to the soil surface than it does at greater depths, which is in accordance to results of our study. Yet, beside the GMS depth of installation, detection of wetting and drying periods was influenced by environmental conditions prior to ground water level as well as the intensity and amount of rainfall. The first year of the study was wetter than normal for this area (Figure 3). Annual rainfall was 1038.2 mm which is nearly 60% above the longterm average, LTA (1961-1990 = 650.6 mm). Furthermore, the amount of rainfall in the growing period was by 53% higher than normal. The second year of the study was warmer and drier than normal. Annual rainfall was 566.2 mm, which is 13% below LTA while the amount of rainfall in the growing period was by 34.2% below LTA. As a result of the intensity and duration of rainfall during the extremely rainy year of 2010, the level of groundwater was very high (Figure 3). In May, the highest point of groundwater was 20 cm, which had a significant effect on irrigation scheduling and yield of maize grain. Previously published results by Marković et al. (2015) showed the significant yield reduction by 7% in a3 irrigation plots in comparison with control plots. In the second year of the study, the level of groundwater was 294 cm and the highest point had no effect on irrigation scheduling and yields (Marković et al., 2015).
During the first year of the study, two irrigation events occurred in July and one in August in a3 irrigation plots. The decision of when to irrigate was based upon an average of soil water potential from two GMS depths. In a3 irrigation plots soil dried near to -40 kPa, which according to the calibration process is soil water status near 80% of FC. Although, if the soil water potential is analyzed as a single GSM depth, it is clear than that the irrigation event should be prolonged since the soil water potential at 30-cm depth was near -20 kPa. At a3 treatment, irrigation increased soil water content to the point that caused plants to suffocate due to a lack of oxygen and therefore yield reduction. The highest yield was obtained in rainfed plots (9.2 t ha -1 ) while a significantly (P<0.01) lower yield (8.59 t ha -1 ) was obtained in fully irrigated plots (Marković, 2013). Furthermore, it seems that some wetting process (from irrigation event or rainfall) at a3 treatment was not detected since in the April-June period soil moisture was very high due to the amount of rainfall and high ground water level. At the beginning of July GMS sensors responded to the first irrigation event which increased soil water content close to 100% of FC in a3 irrigation plots. Since the amount of rainfall in the July period was insufficient for maize needs (50% lower than LTA), there was the second irrigation event in July and the third event in the second decade of the month of August. As it seems, in our study GMSs have shown a slow response to drying and wetting cycles in the lower soil layer (30 cm) since the soil water potential was near -40 kPa although the soil water content near the sensor was close to 100% of FC. Similarly, Irman and Hasman (2001) and Taber et al. (2002) report that the sensors have a slow reaction to changes in moisture at a high level of water content. In a2 irrigation plots one irrigation event (July 22) was detected by the GMS. At that point soil water potential decreased from -79 kPa to -10 kPa ( Figure 2). During the second year of the study, on average, soil water potential ranged from -113.23 kPa in rainfed plots to -50.56 kPa in a3 irrigated plots. As for the soil water potential at different installation depths, the soil water potential was by 14.2% lower at the 30-cm depth in comparison to the 15-cm depth (15 cm = -90.4 kPa; 30 cm = -77.56 kPa). Both irrigation treatments as well as GMS installation depth had a significant impact (P<0.01) on soil water potential (Table 3). As for irrigation x GMS depth interaction, significantly (P<0.01) lowest soil water potential was recorded in a3b2 plots = -46.16 kPa. The second year of our study was extremely dry and warm. Due to a lack of rainfall during the June-August period (Figure 3), the total amount of water added with the irrigation system was 175 mm in a2 and 245 mm in a3 irrigation plots. Severe drought caused soil water potential from -89 kPa in the first decade of June to -187.5 kPa at the end of growing season (Figure 3). Frequent irrigation events at both irrigation treatments did not manage to replenish soil reservoir up to 100% of FC. As presented in Figure 3, sensors were not completely rewetted during irrigation intervals. The reason for this could be that the sensor did not manage to maintain equilibrium with soil water content. McCann et al. (1992) have reported that material may not be able to conduct sufficiently rapidly below 50 kPa of resistance transmission. In the second year of our study variation in resistance of the GMS to different irrigation treatments was more notable since the year of 2012 was extremely dry and warm so the soil water content was very low. In control plots (a1) soil water potential was very high during the entire growing period. It ranged from -60 to -199 kPa at the end of August. It was very difficult to keep soil water content at high level (80% to 100% of FC) because of the lack of rainfall in the July-August period. Soil water potential in a3 plots in most of the growing period was below 50 kPa. Short periods of wetting and drying in June (Figure 3) caused soil water potential to range from -50 to -7.25 kPa. The sensor responded to a change of soil water content but the soil water potential between irrigation intervals did not increase to the minimum value (100% of FC). This means that the sensor responded slower than expected yet it responded very well to drying periods that began close to saturation. According to McCann et al. (1992), during a wetting cycle sensors only respond rapidly and accurately when soil water potential becomes high enough to permit sufficient rewetting of the sensor. This problem was more expressed in a2 irrigation plots since, as it can be seen from Figure 3, resistance did not fall below -36 kPa during the whole growing season.
As for soil water content variation in both growing seasons, deeper GMS showed greater water content variation especially in the dry year of 2012. In both growing seasons, deeper sensors (30 cm) showed a better response to wetting and drying. This result is in accordance to the one given by Perea et al. (2013) since these authors claim that sensors at 30-cm depth have a poor response to irrigation water in comparison to 0-15-cm depth.
The analysis of correlation was made for watermark readings and amount of irrigation water at each irrigation treatment. The analysis shows a weak negative correlation (r= -0.39** (2010), r= -0.29** (2012)) between soil water potential and a2 irrigation treatment, while in fully irrigated plots (a3), the connection was moderate and negative (r= -0.54** (2010), r= -0.51** (2012)). This means that the more water is in the sensor, the more conductive is the medium between the electrodes, that is, the soil water potential decreases as water content increases (Hignett and Evett, 2008). All correlations were significant at the level of P<0.01 (**).

Conclusion
Results of our study show that granular matrix sensors (GMSs) had a good reaction to periods of drying and wetting in irrigation scheduling. The response to changes in soil moisture was slow in conditions where soil water content was at high level. For instance, in growing seasons with intensive rainfall or in drought conditions with frequent irrigation events. Due to differences in results between sensors, it is recommended to use average readings of multiple sensors and to make a calibration for a specific soil type. According to the results from our study, the soil water potential in the extremely wet growing season ranged from 49.9 kPa (a1) to 29.5 kPa (a3) while in the dry growing season it ranged from 113.23 kPa to 50.56 kPa (a3). As for sensor installation depth, water potential ranged from 50.9 kPa (15 cm) to 31 kPa (30 cm) during 2010 and from 90.4 kPa (15 cm) to 77.56 kPa (30 cm) during 2012. A shallow depth installation of the GMS (< 30 cm) is not recommended as a reliable indicator for irrigation scheduling, especially in extreme weather conditions (high ground water level, intensive rainfall events).