Abstract: Nitrogen (N) is one of the important elements for optimum growth as well as high yield in oil palm plantation industry. Hence, estimation of N content in plant tissue is crucial for plantation management to minimize cost of production to produce high yields. Conventional method such as foliar analysis is expensive and not a real time analysis. Implementation of chlorophyll meter and spectral approaches has been used widely in estimation of N in various types of crops but a few studies have been carried out with regard to oil palm. Therefore, this study was conducted to evaluate multiple proximal sensors for N estimation in mature oil palm. The experiment was conducted in a plantation setup in Randomized Complete Block Design (RCBD) with three replications. Three levels of N treatments; 0, 1 and 2 kg N per palm per year as ammonium chloride were applied in split applications within two different planting years plots. The data was collected in February and October 2014. The combination of physiological and spectral models fit the best (R2 = 1.00) for foliar N or N rate estimation in matured oil palm. Generally, SPAD meter was not suitable to estimate N content in matured oil palm leaves but the function can be compensated by the spectroradiometer. The variation of single and combination of single physiological or spectral models in estimating N was highly influenced by age and sampling time. Palm stem diameter was important in this study though it is not sensor-measured.
INTRODUCTION
Like many other cash crops, good agronomic practices involving oil palm nutrient management is undeniably essential to optimize the yield production per hectare. Since the crop constantly removed nutrients through the harvested Fresh Fruit Bunch (FFB) or sequestered in the standing biomass, the consumed nutrients must be abundantly replenished (Wahid et al., 2005). According to the amount required for oil palm growth, N ranked after K and mostly taken up by the plant in the form of soluble nitrate ion (NO3‾).
Insufficient N will affect chloroplast development and functions, leaf area index (Goh and Hardter, 2003), palms height (Uwumarongie-Ilori et al., 2012) and consequently oil and Fresh Fruit Bunch (FFB) (Goh and Hardter, 2003). On the other hand, excessive N applications will increase the susceptibility to disease and insect pest such as leaf-eating caterpillars and bagworms (Goh and Hardter, 2003). Over-application of N may also induce B deficiency which leads to white stripe (Goh and Hardter, 2003). Additionally, the largest portion of the oil palm production cost is attributed to fertilizers where their prices are highly fluctuated (Goh et al., 2003). It is estimated that the oil palm industry will benefit RM 117.25 million per year if excessive ammonium nitrate application as much as 0.25 kg/palm/year can be avoided (Goh and Po, 2005). Thus, it is of an equal importance to be able to estimate the optimum requirement of N to prevent economic loss in plantation industry and for producing maximum oil palm yield.
Foliar analysis is conventionally being practiced to observe nutrient status in palms. However, considering Malaysia alone planted a total of 5.3 million ha (MPOC., 2014) of oil palm, this practice is not feasible since it is costly, time consuming, labour intensive and not accessible for small scale plantations or individual oil palm planters. In seeking for alternative to the conventional foliar analysis, the indirect chlorophyll estimation by the employment of Chlorophyll Content Meter (CCM) has become popular in recent years. The non-destructive, portable and real time instrument has been illustrated to dependably indicate foliar N of perennial crops such as for jatropha (R2 = 0.99) (Nyi et al., 2012), Asian pear (R2 = 0.76) (Ghasemi et al., 2011) and timber trees (R2 = 0.85-0.93) (Percival et al., 2008). The basic principle of the CCM operation is based on the level of photon absorbed by chlorophyll at red (650 nm) and near infrared (940 nm) wavelengths (Peterson et al., 1993; Blackmer et al., 1994) and hence the close connection between extractable leaf chlorophyll (CxHxOxN4Mg) and N content.
Nonetheless, in oil palm related study, the CCM has been limitedly explored. Law et al. (2014) tested the SPAD 502 in oil palm seedlings and found that the meter provided reliable estimate of foliar N (r = 0.73). However, Jifon et al. (2005) and Pinkard et al. (2006) demonstrated that the relationship between SPAD-502 meter and foliar leaf N observed from controlled-environment crops cannot be simply up-scaled to the crops grown in the field, mainly due to the thickening of leaves through the rearrangement of leaf cell during the exposure to direct sunlight that changes the characteristics of light absorption and transmission and hence the meters readings. In comparing the strength of relationship between SPAD-502 meter and foliar N, they also found that the crops grown under controlled environment resulted in stronger relationships than for ones that were field-grown. It is also worth noting that the relationship describing the CCM measurements and foliar N is often influenced by other secondary factors such as crop N partitioning characteristic (Turnbull et al., 2007), variety (Schaper and Chacko, 1991; Shaahan et al., 1999; Jifon et al., 2005) growing condition (Simorte et al., 2001; Liu et al., 2012; Jifon et al., 2005), sampling time (Neto et al., 2011), nutrient deficiencies, water stress and pest and diseases (Peryea and Kammereck, 1997; Pestana et al., 2005).
On the other hand, Leaf Area Index (LAI) is among the physiological parameters of interest measured in oil palm studies since it indicates photosynthetic efficiency (Noor and Harun, 2004). Besides, LAI is crucial for crop monitoring and productivity as well as important components for canopy structure analysis (Pocock et al., 2010). LAI increase as plant grows as it influences the light interception (Ewert, 2004). Comparable to the conventional foliar nutrient analysis, the manual method of measuring LAI is inconvenient for efficient plantation management. Based upon this argument, researchers such as Awal and Wan Ishak (2008) and Noor et al. (2002) had evaluated the use of indirect technique of estimating LAI such as the LAI-2000 Plant Canopy Analyzer (PCA), where the former authors found that the correlation coefficient between manually measured LAI and PCA LAI is weak (r = 0.57). They also reported that the PCA give inconsistent LAI readings in oil palms. However, Behera et al. (2010) found high correlation coefficient (r = 0.99) between manual LAI measurement and PCA for jatropha at sensor angle of 90°. Other than factors of age and sensor angle position, N concentration also influences LAI reading (Pierce et al., 1994).
In mature oil palm nutrient study, plant growth indicator such as height and diameter has not been intensively studied in relation to N response. Nevertheless, since N application rates can induce differences in physiological parameters as such chlorophyll concentration, LAI and net assimilation rate (Corley and Mok, 1972; Uwumarongie-Ilori et al., 2012), these two parameters can also provide information regarding crop growth and health independent of chlorosis. For oil palm seedlings, N organic fertilizers were found to significantly influence the seedlings height and diameter reading (Uwumarongie-Ilori et al., 2012). This result is concurrent with the finding by Gul et al. (2006) and Owolabi et al. (2013) who reported that N organic fertilizers increased the plant height and diameter in oil palm seedlings, respectively.
Numerous efforts towards spectral approach in the study of plant N have been done for perennial crops (Min and Lee, 2005; Perry and Davenport, 2007; Gomez-Casero et al., 2007). Min and Lee (2005) reported that the blue, red, near-infrared and shortwave-infrared bands were worthy for N detection in oranges using spectrophotometer. Following the study by Perry and Davenport (2007), who reported that the narrow band indices such as MCARI and RVSI sensitive to N treatment of apples. Meanwhile, study of Gomez-Casero et al. (2007) indicated the NIR as the most sensitive wavelengths in distinguishing the N treatment in olive. However, the study of foliar nitrogen in oil palm using spectral approach is still lacking in Malaysia. While Nguyen et al. (1995) used SPOT image to estimate macro and micro nutrients of oil palm. In a different study utilizing Landsat-5 TM image, Nor Azleen et al. (2003) tested three vegetation indexes including Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI) and Atmospherically Resistant Vegetation Index (ARVI). The best accuracy is achieved by SAVI (91%) which also give the only positive correlation to foliar N content. However, the performance of the equation is limited by characteristics; site specific, need more effort to make it more accurate and insensitive to the small changes in foliar N contents.
While relative chlorophyll content measured using spectral techniques has been a major subject of study to examine N response in matured oil palm, there are few studies that explore the effects of indirect measurement of physiological parameters such as LAI, plant height and diameter in relation to oil palm N status. Thus, the objectives of this study are:
• | To investigate potential multiple proximal sensors such as chlorophyll meter, LAI Plant Canopy Analyser, clinometer and spectroradiometer (exception of measuring tape for diameter) for estimating nitrogen nutritional content of matured oil palm |
• | To construct mathematical model for N detection in oil palm, based on results derived from the previous objective |
MATERIALS AND METHODS
Study area: The study was carried out in oil palm plantation belonging to United Malacca Berhad located in Melaka Pinda, Malacca, Malaysia. Two plots of Tenera palms with different planting years were selected, 2005 (henceforth MP05; 2.377779°N and 102.265658°E) and 2002 (henceforth MP02; 2.380374°N and 102.238012°E). At the measurement time, the palms age was 9 and 12 years, respectively. The study area included 147 and 148 standing palm per hectare (SPH) for field MP02 and MP05, respectively. In 2013, the total rainfall recorded was 1213 mm with average of 101.08 mm monthly. In early stage of this study, soil samples at depths of 0-15 cm were collected and analysed according to wet digestion method for N content in the soil.
Treatments: Ammonium Chlorite (AC) was applied as N fertilizer in four split applications which were made in November, March, Jun and September. There were three levels of N treatments, 0, 1 and 2 kg AC per palm. The application rate is based on the agronomic practices as implemented by the plantation to minimize disturbance to the palm. Other nutrients were applied as standard agronomic practices. The fertilizers were broadcast around the weeded circle as practiced by the estate.
Experimental design: The experiment was conducted in Randomized Completely Block Design (RCBD) design with 3 replications. Therefore, the experimental field comprises of a total of 18 subplots. The treatments were randomly assigned to each subplot and each subplot contained 16 uniform palms based on visual observation. The plots were carefully selected to ensure that the appearance and growth of the 16 palms were uniform.
Measurements: Two types of data was collected in this study which was physiology and spectral data. Physiology data included LAI, height, diameter, relative chlorophyll content and N tissue content. Intensive field measurements were conducted from 5th-10th February and 8th until 13th October in 2014. The average rainfall during intensive field measurement is shown in Fig. 1. Three and eight oil palm stands in each subplot were selected for physiological and spectral data measurements in February and October 2014, respectively. The LAI was measured on the selected palms using the LAI-2000 Plant Canopy Analyser. Readings were taken at 5 points which included 1 point for open sky and the rest 4 points is taken below canopy reading.
Fig. 1: | Distribution of Relative Humidity (RH), temperature and rainfall od UMB estate |
The height and diameter were measured on each 3 and 8 selected palms within the subplot using clinometer and measurement tape, respectively. The clinometer measures the angle of the target which was the crown. Then the height is calculated using the Eq. 1:
(1) |
where, ° is the angle measured by clinometer, distance (m) is from palm base to the observer position, height is observer height from ground to eye level.
(2) |
where, circumference is the measurement reading from tape.
The diameter was taken at breast height and calculated by Eq. 2. SPAD chlorophyll meter readings were measured on 6 leaflets from frond 17 for each previous selected palm by SPAD Minolta 502 and the readings were averaged. The leaflets chosen were at the end part of fronds located after the thorn position. The same leaflets were used for leaf N nutrient content analysis. The leaflets were oven dried at 70°C for 72 h then ground in the grinder. Then, the sample was analysed by conducting wet digestion using sulphuric acid and hydrogen peroxide for N determination (Miller and Miller, 1948).
Spectral data was collected using spectroradiometer GER 1500 model. It covers spectral wavelength from 350-1050 nm. The spectroradiometer was used to measure the 6 individual leaflets aforementioned that were detached from frond 17. The reflectance reading was taken from 2 h before and after solar noon during clear sky. The position of the fiber optic was ensured to be close enough to the sample so that it avoided the forecasted shadow. The spectroradiometer was calibrated using white spectralon. The data was imported to Microsoft Excel for further processing, where the average of blue, green, red and NIR reflectance were obtained by averaging reflectance of wavelengths ranging from 430-470, 530-570, 630-670 and 700-1100 nm, respectively.
Statistical analysis: The N effects on the ground data and spectral measurements were established by analysis of variance using PROC ANOVA in SAS software. The means were compared using Least Significant Difference (LSD) at 0.05 level of probability. Standard regression analysis was performed using PROC REG to analyse the relationship between the actual ground data and spectral data of N tissue content or N rate. Coefficient of determination (R2) was used as metrics for measuring the relative amount of variations explained by the relationships and the efficiency of the indices. Standard correlation was conducted using PROC CORR to analyse the correlation between each parameters measured and N content.
RESULTS
Single model: Almost all single physiological and spectral models for both MP02 and MPO5 demonstrate weak regression below 0.500 to N rate and foliar N content, as presented in Table 1 and 2. In most of conditions, single physiological and spectral models give better relation with N rate compared to foliar N content regardless of age and sampling time. Only SPAD-N rate relationship in MP05 shows the highest regression of 0.661. Height shows consistency in N rate models while Height and Diameter is the most consistent model describing foliar N. For the spectral model, the highest R2 with regard to N rate is shown by the Green (R2 = 0.292) model of MP05 in October. All four spectral of MP02-foliar N models display the highest r2 during October sampling.
Correlation: All physiological and spectral models with regard to age and sampling time demonstrated non-significant correlation to N rate and foliar N (Table 1 and 2). Again, only the SPAD-N rate relationship of MP05 shows strong significant correlation of r = 0.813. For MP02, correlation values for all physiological models to estimate foliar N measured in February severely decreased in October. On the other hand, the spectral models illustrated a contradict pattern where the correlation values were initially low in February but later increased in October. For MP05, there was no distinct pattern on correlation values for both N rate and foliar N models either for physiological or spectral models. Similar observation was made for MP02 for the N rate models.
Combination of single physiological or spectral model: Irrespective to age and sampling time, the combination of all physiological or spectral parameters consistently demonstrated the best relationship to N rate and foliar N content, as displayed in (Table 3 and 4). The best physiological model is LAI+SPAD+Height+Diameter (N rate = 0.389-0.767, foliar N = 0.183-0.798) while the spectral model is expressed as Blue+Green+Red+NIR (N rate = 0.407-0.742, foliar N = 0.242-0.766). In comparing the performance of the physiological model and spectral model for N rate and foliar N estimation, the latter depicted stronger relationship in many conditions.
Table 1: | Linear regression and Pearson correlation coefficient between physiological and spectral parameters with N rate and foliar N for MP02 and MP05 in February |
NS: Non significant, **Significant at p = 0.05 |
Table 2: | Linear regression and Pearson correlation coefficient between physiological and spectral parameters with N rate and foliar N for MP02 and MP05 in October |
NS: Non significant, **Significant at p = 0.05 |
Table 3: | Linear regression coefficient for combination of physiological and spectral type for MP02 and MP05 in February |
Table 4: | Linear regression coefficient for combination of physiological and spectral type for MP02 and MP05 in October |
It is found that N foliar estimation using both physiological and spectral models for both fields were affected by the sampling time. The R2 values explaining the relationship to foliar N at first were low and moderate in February but increased to moderate and high values in October. However, an exception could be made to physiological model of MP02. For the N rate, it is observed that variation in sampling time did not affect the strength of the relationship. Although the LAI+SPAD+Height+Diameter and Blue+Green+Red+NIR models demonstrate the highest R2, these models are applicable only to certain conditions. For e.g., the spectral model of MP02 is applicable to both N rate and foliar N but the application is restricted in February for N rate and in October for foliar N. The inconsistency in the application of different models is also observed for MP05, where the estimation of N rate is best done using the physiological model in February but in October, the spectral model is best implemented. Meanwhile, the physiological model is worth for foliar N estimation in October.
Combination of physiological and spectral model: Regardless to all conditions, the LAI+SPAD+Height+ Diameter+Blue+Green+Red+NIR is the perfect model to estimate both N rate and foliar N (Table 5-12). This model with all eight physiological and spectral parameters exhibited the highest R2 = 1.000. Among the top five foliar N models, the models excluding SPAD display strong relationships to foliar N with R2 values range from 0.930-1.000 as listed in (Table 5, 7, 9 and 11). For the N rate models, only MP02-February and MP05-October show strong relationships without the SPAD combination but for MP02-October and MP05-February, SPAD readings were required to achieve strong relationship (R2 = 0.949-1.000).
Table 5: | Linear regression coefficient for combination of physiological and spectral indices by Foliar N for MP02 in February |
Table 6: | Linear regression coefficient for combination of physiological and spectral indices by N rate for MP02 in February |
Table 7: | Linear regression coefficient for combination of physiological and spectral indices by Foliar N for MP05 in February |
For estimating N rate or foliar N content, stem diameter is a must parameter in any top five models regardless of age and sampling times.
Table 8: | Linear regression coefficient for combination of physiological and spectral indices by N rate for MP05 in February |
Table 9: | Linear regression coefficient for combination of physiological and spectral indices by Foliar N for MP02 in October |
Table 10: | Linear regression coefficient for combination of physiological and spectral indices by N rate for MP02 in October |
Table 11: | Linear regression coefficient for combination of physiological and spectral indices by foliar N for MP05 in October |
Table 12: | Linear regression coefficient for combination of physiological and spectral indices by N rate for MP05 in October |
For foliar N models, diameter, green and red are required in the combination of high R2 models. However, for effective N rate models, LAI and height are essential additional besides diameter.
DISCUSSION
Single model: From the result, all single models neither physiological nor spectral are efficient for N rate or foliar N estimation in matured oil palm with the exception of SPAD-N rate model of MP05. However, the SPAD-N rate model is not applicable to all conditions as the model is restricted to age and time. SPAD has been utilized in N detection for oil palm seedling (Law et al., 2014); while the result from this study shows that the sensor is incompetent for matured oil palm with an exception to MP05. This may be due to variation in SPAD reading related to leaf age as supported by Percival et al. (2008). Besides, there was poor responses between N applied and foliar N which may influence the SPAD reading. Furthermore, the N status of plant affect the SPAD reading (Netto et al., 2005). Secondary factors such as leaf thickness (Campbell et al., 1990), varieties (Law et al., 2014), chlorophyll or foliar N distribution and sampling season (Chang and Robison, 2003) and complex sink-source of N partitioning (Loh et al., 2002) were also influencing SPAD performance. Although there is limitation in using SPAD but the application is near to real time and better than the expensive foliar analysis in monitoring N.
The Green model shows consistency to N rate model even though the value is low. The Green N rate-model can be applied if a single parameter have to be used in estimation of N as many authors found the used of green wavelengths gave positive results in N estimation. Gitelson et al. (2003) reported the reciprocal reflectance of green band (520-550 nm) was sensitive to chlorophyll content of beech, wild vine, maple and chestnut. Datt (1998) also reported that model based on green and NIR was sensitive to different ranges of chlorophyll content among Eucalyptus species.
Combination of single physiological or spectral model: This model displays better R2 values compared to single model. LAI+SPAD+Height+Diameter and Blue+Green+Red+NIR models can be used for N rate or foliar N estimation in matured oil palm. In most of conditions, both physiological and spectral model shows moderate to strong relationship to foliar N or N rate in younger oil palm (MP05) compared to older oil palm (MP02). Thus, the age factor contributed in differences of N estimation in oil palm. This finding is concurrent with Hartley (1988) and Foster (2003) that reported foliar N concentration in oil palm decreased with age. Similar observation was also made by Kamau et al. (2008) in tea plantation. According to Sari et al. (2006), the accurate result from spectral properties is the reflection of the internal and external structure of plants which are age and growth stages dependent.
The sampling time affected either the foliar N or N rate estimation models regardless to palm age. Most of N foliar models display higher R2 in October may cause by contribution of high Relative Humidity (RH) and rainfall. The RH above 75% throughout the year are favorable for optimum oil palm growth (Verheye, 2010; Carr, 2011). Meanwhile in February, low RH and less than 65 mm rainfall were recorded in Fig. 1. Besides, there was water scarcity during the sampling time as there was little rainfall in January, February and March consecutively. Thus, this low RH increased transpiration in palm and also inadequate water for root uptake during the transpiration process that leads to partial or full stomatal closure (Smith, 1989). The closure of stomata will block the entry of CO2, eventually disturbed the photosynthesis and others physiological process in the palm. Hence, sampling time must be taken for consideration during data collection.
Generally, the spectral model demonstrated strong relation to N rate or foliar N in most conditions. Numerous studies have been conducted using spectral responses and have been used widely in many crops for N prediction (Davenport et al., 2005; Alchanatis et al., 2005; Suarez and Berni, 2012). In addition, integration of multiple spectral bands from the visible and NIR bands in a model do increase the probability of model succession especially for wide range species (Blackburn, 2007). However, implementation of these models also constrained to sampling time and palm age.
Mostly, N rate models demonstrated better result rather than foliar N models in estimation of N in matured oil palm regardless of sampling time and age. The differences in foliar N estimation are related to N uptake and the complex sink-source relationship of N allocation within perennial crops. The N uptake in oil palm corresponded with age (Von Uexkull and Fairhurst, 1991) , varieties (Law et al., 2012), length and distribution of functional root system (Corley and Tinker, 2003). Besides, factors such as age and plant genotype (Osaki et al., 1993), N supply (Millard and Neilsen, 1989), light and temperature (Muller et al., 2005) influenced the allocation of N in crops. Therefore, these factors may cause the difference in foliar N models.
Combination of physiological and spectral model: Combination of both physiological and spectral information has yielded the perfect model since the N allocation in palm is distributed all over the palm organs such as stem, roots and rachis (Ng et al., 1968). Thus, the model that accommodates all of N sinks is appropriate to estimate N content of matured oil palm. Spectral model alone, for instance, is not efficient in estimating N as the spectral model is characterized by the variations of N that is only available at the foliar level.
Combination of single and physiological model with exclusion of SPAD frequently displayed strong relationship with foliar N. Spectroradiometer have the ability to compensate the function of SPAD in N estimation. Regarding to information availability, only single index value is obtained from SPAD whereas the spectroradiometer provided a lot of information from the reflectance reading. SPAD only covers in red (640 nm) and NIR (940 nm) region (Munoz-Huerta et al., 2013) whereas, spectroradiometer spans visible and NIR region. A blue and red wavelength was well-known to be sensitive to chlorophyll absorption while the green wavelength reflects the health status of plant (Reddy and Matcha, 2010). Meanwhile, the NIR wavelength associates with the leaf structure such as spongy mesophyll cell which give the information about the plant stress or senescence (Knipling, 1970).
Diameter is important parameters in all top five of foliar N or N rate models as it contribute to the perfectness of N estimation models although it was not measured by proximal sensor. Diameter was found to relate better with N estimation than the SPAD, LAI and height in mature oil palm. Due to close relation between diameter and height, the height can be estimated through diameter by using allometric relationships. There was strong relationship between height and diameter as reported by Avsar (2004) in pines tree and later supported with finding by Avsar and Ayyildiz (2005) in cedar trees.
CONCLUSION
In conclusion, results of this study displayed the combination of physiological and spectral models fit the best either for N rate or foliar N estimation of matured oil palm irrespectively to sampling time and age. The limitation of SPAD in this study makes the application optional. Without SPAD, the models are still feasible to estimate N in matured oil palm. On other hand, the single and combination of single physiological or spectral models are age and sampling time dependant. Nevertheless, Diameter or Green model is applicable if only single parameter is available for N estimation but the expected result of R2 will be relatively low. In this study, the N rate models display good result compared to foliar N model.
ACKNOWLEDGMENTS
The authors would like to thank United Malacca Berhad for the financial support and providing the research areas.