Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published in Agron J 100:205-212 (2008)
DOI: 10.2134/agrojnl2007.0018
© 2008 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Agricola
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Related Collections
Right arrow Forage Management
Right arrow Nutrient Management

REMOTE SENSING

A Simple Spectral Index Using Reflectance of 735 nm to Assess Nitrogen Status of Rice Canopy

Yuh-Jyuan Leea, Chwen-Ming Yanga, Kuo-Wei Changb and Yuan Shenc,*

a Div. of Crop Science, Taiwan Agricultural Research Institute, Wufeng, Taichung Hsien 413, Taiwan ROC; yjlee{at}wufeng.tari.gov.tw, cmyang{at}wufeng.tari.gov.tw
b Dep. of Leisure and Recreation Studies, Aletheia Univ., Tainan Hsien 721, Taiwan ROC; ckw550320{at}yahoo.com.tw
c Dep. of Soil & Environmental Sciences, National Chung-Hsing Univ., Taichung 40227, Taiwan ROC

* Corresponding author (yshen{at}nchu.edu.tw).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
Spatial distribution of canopy N status is the primary information needed for precision management of N fertilizer. This study demonstrated the feasibility of a simple spectral index (SI) using the first derivative of canopy reflectance spectrum at 735 nm (dR/d{lambda}|735) to assess N concentration of rice (Oryza sativa L.) plants, and then validated the applicability of a simplified imaging system based on the derived spectral model from the dR/d{lambda}|735 relationship in mapping canopy N status within field. Results showed that values of dR/d{lambda}|735 were linearly related to plant N concentrations measured at the panicle formation stage. The leaf N accumulation per unit ground area was better fitted than other ratio-based SIs, such as simple ratio vegetation index (SRVI), normalized difference vegetation index (NDVI), R810/R560, and (R1100 – R660)/(R1100 + R660), and remained valid when pooling more data from different cropping seasons in varied years and locations. A simplified imaging system was assembled and mounted on a mobile lifter and a helicopter to take spectral imageries for mapping canopy N status within fields. Results indicated that the imaging system was able to provide field maps of canopy N status with reasonable accuracy (r = 0.465–0.912, root mean standard error [RMSE] = 0.100–0.550) from both remote sensing platforms.

Abbreviations: DNs, digital numbers • LAI, leaf area index • LNA, leaf N accumulation per unit ground area • NDVI, normalized difference vegetation index • PA, precision agriculture • RMSE, root mean standard error • SI, spectral index • SRVI, simple ratio vegetation index • TARI, Taiwan Agricultural Research Institute • TIFF, tag image file format

A Simple Spectral Index Using Reflectance of 735 nm to Assess Nitrogen Status of Rice Canopy

Yuh-Jyuan Leea, Chwen-Ming Yanga, Kuo-Wei Changb and Yuan Shenc,*

a Div. of Crop Science, Taiwan Agricultural Research Institute, Wufeng, Taichung Hsien 413, Taiwan ROC; yjlee{at}wufeng.tari.gov.tw, cmyang{at}wufeng.tari.gov.tw
b Dep. of Leisure and Recreation Studies, Aletheia Univ., Tainan Hsien 721, Taiwan ROC; ckw550320{at}yahoo.com.tw
c Dep. of Soil & Environmental Sciences, National Chung-Hsing Univ., Taichung 40227, Taiwan ROC

* Corresponding author (yshen{at}nchu.edu.tw).

Received for publication January 13, 2007.
Spatial distribution of canopy N status is the primary information needed for precision management of N fertilizer. This study demonstrated the feasibility of a simple spectral index (SI) using the first derivative of canopy reflectance spectrum at 735 nm (dR/d{lambda}|735) to assess N concentration of rice (Oryza sativa L.) plants, and then validated the applicability of a simplified imaging system based on the derived spectral model from the dR/d{lambda}|735 relationship in mapping canopy N status within field. Results showed that values of dR/d{lambda}|735 were linearly related to plant N concentrations measured at the panicle formation stage. The leaf N accumulation per unit ground area was better fitted than other ratio-based SIs, such as simple ratio vegetation index (SRVI), normalized difference vegetation index (NDVI), R810/R560, and (R1100 – R660)/(R1100 + R660), and remained valid when pooling more data from different cropping seasons in varied years and locations. A simplified imaging system was assembled and mounted on a mobile lifter and a helicopter to take spectral imageries for mapping canopy N status within fields. Results indicated that the imaging system was able to provide field maps of canopy N status with reasonable accuracy (r = 0.465–0.912, root mean standard error [RMSE] = 0.100–0.550) from both remote sensing platforms.

Abbreviations: DNs, digital numbers • LAI, leaf area index • LNA, leaf N accumulation per unit ground area • NDVI, normalized difference vegetation index • PA, precision agriculture • RMSE, root mean standard error • SI, spectral index • SRVI, simple ratio vegetation index • TARI, Taiwan Agricultural Research Institute • TIFF, tag image file format


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
NITROGEN IS THE MOST IMPORTANT and essential element affecting crop productivity among plant nutrients. Nitrogen fertilization plays a vital role in enhancing and stabilizing crop growth and yield production (Martens, 2001). However, excess reactive N, derived from over and/or improper application of N fertilizers, may produce detrimental effects on public health and ecosystems (Buresh et al., 1993; Mashima et al., 1999). Excess N deposition in atmosphere considered recognition as a potent contributor to global warming and stratospheric ozone depletion (Galloway and Cowling, 2002).

Precision agriculture (PA), also known as precision farming or site-specific farming, is a newly evolved management system/strategy with goals not only to increase efficiency and economical return of agricultural activities, such as fertilization and pesticides spraying, but also to achieve sustainable agriculture (Blackmer et al., 1996; Cassman et al., 2002). One of the key factors in implementing PA is the ability to provide timely information regarding spatial distribution of crop N status within a field. From this perspective, determining plant N concentration by remote sensing techniques is much more appealing than the traditional destructive chemical analyses on plant samples considering the cost and time required. Remote sensing techniques are also better than methods using chlorophyll meters (Peng et al., 1993; Adamsen et al., 1999) because reflectance measurements are adjustable to the sample area without attaching a sensor on the leaf, and thus can reduce variability of the measurements.

Rice yield is known to closely related to N status before heading stage (Cui and Lee, 2002; Ntanos and Koutroubas, 2002), and N topdressing at panicle initiation/formation stage is most crucial for rice yield and quality (Nguyen and Lee, 2006). In this regard, assessing canopy N status by remote sensing techniques at these critical periods can provide the needed information for variable rate fertilizer application at the right place and right time of growth period.

Plant leaves absorb in the visible region, mainly the blue and red portions, of incoming solar energy spectrum. Reflectance in the near-infrared region of the spectrum is primarily dependent on the leaf water content, internal leaf structure, and canopy structure/architecture (Guyot, 1990). Many researchers have tried to use visible and near infrared spectral responses from plant canopies to assess N status. Thomas and Oerther (1972) showed that leaf N content of sweet pepper (Capsicum annuum L.) could be estimated by measuring leaf reflectance at 550 nm, while Blackmer et al. (1994) indicated that reflectance at 550 nm could detect N deficiencies in corn (Zea mays L.) leaves. Shibayama and Akiyama (1986) found that reflectance values at 620 and 760 nm (regression with two variables) or 400, 620, and 880 nm (regression with three variables) correlated well with leaf N concentration, and found linear relationship between the measured and the predicted values in spite of various types and cultivars of rice. Takebe et al. (1990) stated that green color intensity values and total N content of the second leaf from the top of rice plants were highly correlated. Inoue et al. (1998) reported a close relationship (R2 = 0.96) between the normalized difference of R1100 and R660 [(R1100–R660)/(R1100+R660)] and leaf N accumulation per unit ground area (LNA) in rice when LNA was lower than 3 g m–2, and a combination of four spectral bands (R550, R830, R1650, and R2200) also gave a satisfied estimation of LNA (R2 = 0.91). In a preliminary study, Shen et al. (2000) pointed out that the first derivative of canopy reflectance spectrum at 735 nm may be a good indicator (R2 = 0.86) for assessing rice canopy N status. Xue et al. (2004) suggested that the reflectance ratio of 810 nm to 560 nm (R810/R560) was linearly related to total leaf N accumulation (R2 = 0.85), independent of N level and growth stage, and could be a useful nondestructive tool for monitoring of N content in rice plants.

The conventional SIs, such as SRVI or NDVI, which based on ratios of broadbands in visible and near-infrared regions of the solar spectrum (Rouse et al., 1974), have also been used for growth estimation and stress evaluation (Patel et al., 1985; Clevers, 1989; Steven et al., 1990; Bouman et al., 1992; Price and Bausch, 1995; Wooten et al., 1999). Yang and Chen (2004) monitored the change in NDVI and leaf area index (LAI) for estimating rice growth along plant development. By assessing LAI from the LAI)NDVI relationship using values of NDVI as inputs, Chen and Yang (2005) were able to predict rice yield 6 wk before harvest. Others have related NDVI to changes in plant vigor and biomass (Holben et al., 1980; Leblon et al., 1991; Gilabert et al., 1996) and for assessing the effect of moisture stress (Wooten et al., 1999). However, before applications of such biophysical attribute)SI relationship, the developed models should be validated for their feasibility and applicability.

The objectives of this research were to examine the feasibility of a derivative type SI dR/d{lambda}|735 and four ratio-based SIs [i.e., NDVI, SRVI, R810/R560, and (R1100–R660)/(R1100+R660)] in assessing N concentration of rice plants at the panicle formation stage, and then to validate the applicability of a simple imaging system unit based on the spectral model derived from the N)dR/d{lambda}|735 relationship in mapping canopy N status in a field from two remote sensing platforms.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 REFERENCES
 
Rice Cultivation and Nitrogen Treatment
Field experiments with different application rates of N fertilizer were conducted, similar to that reported by Shen et al. (2000), at the Shi-Ko experimental farm of Chiayi Branch Station (Chiayi Hsien, 23°52'24'' N, 120°21'29'' E), Taiwan Agricultural Research Institute (TARI), from 2000 to 2002 cropping seasons. Field experiments were also performed at three experimental sites in north, central, and southern part of Taiwan and data was collected from various seasons. They were the first and the second cropping seasons of 2002 at Ankung experimental farm of National Taiwan University (northern Taiwan; Taipei Hsien, 23°52'24'' N, 120°21'29'' E), the first and the second cropping seasons of 2001 at the Precision Agriculture Experimental Farm of TARI (central Taiwan; Taichung Hsien, 24°1'56'' N, 120°41'20'' E), and the first and the second cropping seasons of 2002 at the Kaohsiung District Agricultural Research and Extension Station Experimental Farm (southern Taiwan; Pingtung Hsien, 23°52'24'' N, 120°21'29'' E). Most of the collected data was used to derive the spectral model from the N)dR/d{lambda}|735 relationship, while a subset was used for model validation.

In Taiwan, rice plants normally grow under a two-season cropping system. Three-leaf-old seedlings were machine transplanted in February for the first season crop and the second season crop was transplanted in July. Transplanting density was 0.15 by 0.25 m with two to three plants per hill. Nitrogen application rates at Taipei Hsien, Taichung Hsien, Chiayi Hsien, and Pingtung Hsien experimental sites were (0, 100, 200 kg N ha–1), (0, 60, 120, 180 kg N ha–1), (0, 45, 90, 180 kg N ha–1), and (0, 45, 90, 180 kg N ha–1), as ammonium sulfate, respectively. Other macro-elements (P and K) were applied in rates of 20 kg P ha–1 and 55 kg K ha–1 in forms of calcium superphosphate and potassium chloride, respectively. It was a loamy soil with pH of 4.5 to 5.8 and organic matter from 0.011 to 0.023 kg kg–1. The experimental plots were all arranged in a randomized complete block design with at least three replications. Individual plot sizes varied according to field arrangement at each site, but were at least 10 x 10 m. The regional recommendations of field management practices, such as in-season weed and pest controls and irrigation, were followed.

Spectral Measurements and Calculation of Spectral Indices
When rice plants grown to the panicle formation stage, the most critical period that affect yield, canopy reflectance spectra were measured using a field-portable spectroradiometer (model LI-1800, LI-COR Inc., Lincoln, NE) that was connected to a remote cosine receptor (model LI-1800–02, LI-COR Inc., Lincoln, NE). For ground-truth studies, the receptor was attached to a 1.5-m extension arm and was held horizontally at a nadir viewing 1 m above the canopy surface. At this height, a target area of 1 m-radius may have occupied 80% of the view. The person holding the extension arm always wore dark clothes and stood sideways to minimize measurement interference. All the measurements were made under near cloud-free conditions within 2 h ± solar noon. Incident and reflected solar radiations were measured simultaneously, by facing the remote cosine receptor upward and downward, respectively, and the reflectance was calculated from these two data sets. The measurements were taken over the wavelength range from 400 to 1,100 nm at a scanning interval of 10 nm. Each plot was measured three times to reduce the atmospheric effects and the mean reflectance spectrum was used for data analyses. The typical canopy reflectance spectra collected in the panicle formation stage from rice plants grown under different application rates of N fertilizer in first crop of 2001 are plotted in Fig. 1 . As indicated, spectral reflectance fluctuated in the measured spectral domain and changed with N application rates. The first derivative value of canopy reflectance spectrum at 735 nm (dR/d{lambda}|735) was computed by subtracting the reflectance at 730 nm from the reflectance at 740 nm and divided by 10 nm (Shen et al., 2000).


Figure 1
View larger version (15K):
[in this window]
[in a new window]

  		Fig. 1. The typical canopy reflectance spectra (350–1,100 nm) collected in the panicle formation stage from rice plants (Oryza sativa cv. Tainung 67) grown under different application rates of N fertilizer at the Precision Agriculture Experimental Farm of Taiwan Agricultural Research Institute (Taichung Hsien, Taiwan) in first crop of 2001.

 
The values of NDVI, defined as (RNIR – RRED)/(RNIR + RRED), and SRVI, defined as RNIR/RRED, were calculated from the simulated broadband reflectance of red (RED, 610–680 nm) and near-infrared (NIR, 780–890 nm) regions as those of the sensors of SPOT-5 satellite. The reflectance of simulated broadband sensors was acquired by using the mean value of reflectance averaging over the respective waveband regions of hyperspectral data (Yang and Chen, 2007). The reflectance ratio of 810 nm to 560 nm (R810/R560) (Xue et al., 2004) and the normalized difference of R1100 and R660 [(R1100 – R660)/(R1100 + R660)] (Inoue et al., 1998) were also calculated to correlate with plant N concentration.

Plant Nitrogen Analysis, Image Processing, and Model Validation
Whole plant samples, three-hill per sampling, taken from the targeted regions on the days of spectral measurements, were oven-dried at 60°C for 48 h before milling. The chemically analyzed plant N concentration was determined by micro-Kjedahl method (Bremner, 1996) as that modified by Lee et al. (2002).

To validate the applicability of the simplified imaging system unit based on the spectral model derived from the N–dR/d{lambda}|735 relationship, additional field experiments were conducted at the Precision Agriculture Experimental Farm of TARI in the first crops of 2001 and 2002, respectively. The lightweight design system unit, which composed of a monochrome digital camera (EDC-1000L, Electrim Corp., Princeton, NJ) with a wide-angle lens (PHF6 1.4, Canon Inc., Tokyo, Japan), a rotative circular tray with bandpass filters 730 and 740 nm (Andover Corp., Salem, NH), and an IBM PC/AT compatible computer for command input and image storage, was assembled following the method of Lee et al. (2007). The self-written software, which provided the camera with basic features, such as image acquisition and display, anti-blooming control, exposure time setting, gain and bias adjustment, and pixel gain correction, was installed to the computer for operation. The software also has procedures to do reflectance correction, geometric correction, orthorectification, and resampling. Linear relationships between digital numbers (DNs) and reflectance of control panels at corresponding wavelength were first established using reflectance control points in each acquired image. The reflectance control panels of known reflectance spectra were placed on the ground during the time of image acquisition. The reflectance of other pixels in the image was then linearly interpolated based on the established relationship.

To perform analytical rectification, numerical procedures using two-dimensional projective transformation method, as described by Wolf (1983) for tilted photography, were used. The method permitted the use of redundant amounts of ground control points; hence least squares computational techniques were used to determine the most probable values for the geometric correction parameters. Generally, eight or more control points were used for each scene. Once the geometric correction parameters have been determined, the rectified and ratioed coordinates of every pixel in the image were computed accordingly. The rectified images were resampled using the nearest-neighbor method as described by Schowengerdt (1997) and grid size specified by the user. It generally took about an hour to complete all the image processing work for sets of scene taken from a measurement trip to the end users. In this study, a sound image that covered the size of the experimental field taken from the flight during the panicle initiation/formation stage of rice crop was enough to yield the needed N map for variable-rate application of N fertilizer. The acquired imageries can be saved in tag image file format (TIFF) or PCX (PC Paintbrush) formats files for compatibility with other application software.

For near ground canopy N mapping in 2001, the imaging system unit was raised 15 m above canopy surface by a mobile lifter. A field plot of 0.5 ha, 100 by 50 m, was divided into three equal regions and different rates of N fertilizer were applied. From east to west, the rates were 90, 0 kg ha–1 (first region); 0, 45 kg ha–1 (second region); and 45, 135 kg ha–1 (third region). There were eight spots randomly selected from each region for spectral measurements and plant samplings. Because the viewing angle of imageries was taken 15 degrees from normal, only the 20 by 20 m area of each region (closest to the lifter) was analyzed to reduce the potential interference due to bi-directional reflection. For middle altitude canopy N mapping in 2002, the system unit was mounted on a helicopter flying 600 aboveground surface. A 1-ha field plot, 200 by 50 m, was applied with varied rates of N fertilizer to produce a spatially heterogeneous rice canopy. The plot was divided into 16 regions and each region was further split up to six equal-area grids. Plant sampling and spectral measurement were made on each grid and resulted in a total of 96 spots obtained from the experimental plot.

Two sets of field imageries of 730 and 740 nm were acquired by the imaging system at panicle formation stage from both years. After the necessary reflectance correction and orthorectification, the images taken from both remote sensing platforms were resampled into 0.5 m (near-ground image) and 2 m (middle altitude image) square grids. The values of dR/d{lambda}|735 for each resampled grids were computed. Maps of canopy N status over the experimental paddy fields were generated using the spectral model derived from the N–dR/d{lambda}|735 relationship as aforementioned. To generate canopy N status maps, plant samples were harvested the next day from the targeted regions in locations where the field imageries were taken and were analyzed for N status.

The assessed plant N concentration (%) was computed from the spectral model derived from regression analysis of the N–dR/d{lambda}|735 relationship using values of dR/d{lambda}|735 as inputs by the General Linear Model of StatSoft (2001). The accuracy of the regression model was examined by a linear correlation analysis between the chemically analyzed values and the image analyzed values of plant N concentration yielded from the N maps. The RMSE between the analyzed and the assessed values was also calculated for comparing the precision of assessment. The formula used is

Formula
where Xi is the analyzed value and Xi is the assessed value.


   RESULTS AND DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
REFERENCES
 
This study was conducted primarily to examine the feasibility of a derivative type of spectral index, dR/d{lambda}|735, proposed by Shen et al. (2000), in assessing N concentration of rice plants, and to compare the differences in N assessment using the well-known ratio-based SIs NDVI and SRVI. As listed in Table 1 , similar to SRVI, dR/d{lambda}|735 provided a wider range of variability (40–113%) relative to 81 to 101% for NDVI, and an improved clear-cut to N gradients than NDVI when N application rates >90 kg ha–1. Thus, both dR/d{lambda}|735 and SRVI are considered as suitable SIs in distinguishing growth differences resulted from different rates of N fertilizer than NDVI, particularly for the medium (90–120 kg ha–1) and high (180–200 ka ha–1) rates of N applications. Collins (1978) and, more recently, Baret et al. (1992), Demetriades-Shah et al. (1990), and Mauser and Bach (1995) concluded that derivative type of SIs were very sensitive to changes in green LAI and chlorophyll concentration, while they tended to minimize the spectral noise caused by soil background and atmospheric effects. This study confirms the feasibility of SI dR/d{lambda}|735 in assessing N concentration of rice plants and provides a distinguishable assessment of N concentrations as SRVI.


View this table:
[in this window]
[in a new window]

  		Table 1. Changes of simple ratio vegetation index (SRVI), normalized difference vegetation index (NDVI), and the first derivative differential value of canopy reflectance spectra at 735 nm (dR/d{lambda}|735) for rice plants (cv. Tainung 67) grown under different application rates of N fertilizer in different locations from the first cropping season of 2000 to the second cropping season of 2001.

 
The correlation between plant N concentrations (%) and values of dR/d{lambda}|735 at panicle formation stage under different rates of N applications in first crop of 2000 at Chiayi Hsien and first crop of 2001 at Taichung Hsien were then investigated. Results showed that the N–dR/d{lambda}|735 relationship was a linear function (R2 = 0.679, P < 0.001, N = 17) (Fig. 2 ), similar to that reported by Shen et al. (2000). The coefficient of determination (R2) was found to be greater than that of the N–NDVI relationship (R2 = 0.471, P < 0.010, N = 17). When data from different cropping seasons and various locations were pooled, the N–dR/d{lambda}|735 relationship remained a linear function in the measured range of dR/d{lambda}|735 (R2 = 0.514, P < 0.001, N = 58), irrespective to cropping seasons and locations (Fig. 3 ). By further analyzing correlations of plant N concentrations to other ratio-based SIs, results showed that the N–SRVI relationship obtained the highest R2 value (0.519), followed by R810/R560 (R2 = 0.453), NDVI (R2 = 0.355), and (R1100–R660)/(R1100+R660) (R2 = 0.111) (Fig. 4 ). Along with SRVI, the derivative index dR/d{lambda}|735 is again proved a better indicator in assessing plant N concentration.


Figure 2
View larger version (18K):
[in this window]
[in a new window]

  		Fig. 2. Correlations between plant N concentrations (%) and the first derivative values of canopy reflectance spectra at 735 nm (dR/d{lambda}|735) (A) and normalized difference vegetation index (NDVI) (B) measured at the panicle formation stage for rice plants (Oryza sativa cv. Tainung 67) grown under different application rates of N fertilizer at Chiayi Hsien in first crop of 2000 and at Taichung Hsien in first crop of 2001.

 

Figure 3
View larger version (24K):
[in this window]
[in a new window]

  		Fig. 3. Correlation between plant N concentrations (%) and the first derivative values of canopy reflectance spectra at 735 nm (dR/d{lambda}|735) measured at the panicle formation stage for rice plants (Oryza sativa cv. Tainung 67) grown under different application rates of N fertilizer at Taipei Hsien, Taichung Hsien, Chiayi Hsien, and Pingtung Hsien in first or second crop from 2000 to 2002.

 

Figure 4
View larger version (31K):
[in this window]
[in a new window]

  		Fig. 4. Correlation between plant N concentrations (%) and simple ratio vegetation index (SRVI), normalized difference vegetation index (NDVI), (R1100 – R660)/(R1100+ R660), and R810/R560 measured at the panicle formation stage for rice plants (Oryza sativa cv. Tainung 67) grown under different rates of N fertilizer at Taipei Hsien, Taichung Hsien, Chiayi Hsien, and Pingtung Hsien in first or second crop from 2000 to 2002.

 
Many SIs, such as NDVI and SRVI, have been frequently used as biophysical parameters for describing growth performance of live green canopy material (Rouse et al., 1974). They were developed with an attempt to reduce spectral effects caused by the atmosphere and soil background (Huete, 1988). These ratio-based SIs incorporated the functions of red and near-infrared bands of the vegetation spectrum and resulted in a nonlinear measure to growth traits (Yang and Chen, 2004). Results of the present study further demonstrated the applicability of these ratio-based SIs [i.e., SRVI, NDVI, R810/R560, and (R1100–R660)/(R1100+R660)] in assessing N concentration of rice plants, as that of the derivative index dR/d{lambda}|735.proposed by Shen et al. (2000).

Applications of the N–dR/d{lambda}|735 relationship to a wider range of habitats were validated with other datasets collected from different seasons in various locations. As indicated in Fig. 5 , relationship between the chemically analyzed and the model assessed values of plant N concentration was linearly related (r = 0.554, P < 0.010). Results suggest that SI derived from the N–dR/d{lambda}|735 relationship is valid for the experimental regions studied in assessing N concentration of rice plants, and the yielded spectral model may have a potential in mapping canopy N status within field by spectral remote sensing. That is, by using the composite regression equation, changes in plant N concentration may be roughly assessed and monitored from canopy reflectance data for rice plants grown in the selected locations of Taiwan, in spite of different cropping seasons. However, the loose distribution between the chemically analyzed and the model assessed values in turn suggests that location and climatic effects may impose a wide variability in using the spectral model. Similar responses to the applications of N fertilizer have been observed in corn (e.g., Hanway, 1962; Karlen et al., 1987) and sorghum [Sorghum bicolor (L.) Moench] (e.g., Myers, 1978; Muchow et al., 1990), and were attributed to the differences in climatic, soil and genotypic factors across seasons and locations.


Figure 5
View larger version (22K):
[in this window]
[in a new window]

  		Fig. 5. Comparisons between the chemically analyzed and the model predicted N (%) from the spectral model derived from the N-dR/d{lambda}|735 relationship for rice plants (Oryza sativa cv. Tainung 67) grown under different rates of N fertilizer at Taipei Hsien, Chiayi Hsien, and Pingtung Hsien in both first and second crops of 2002.

 
A simplified imaging system based on the aforementioned spectral model was then assembled, as that reported by Lee et al. (2007), to examine its potential in mapping canopy N status within fields from two different remote sensing platforms. Two sets of field images, compiled from reflectance data measured with 730 and 740 nm sensors of the system, were taken 15 m above canopy surface by a mobile lifter for rice plants grown at the panicle formation stage of first crop in 2001 (Fig. 6 ). Field plots were applied with different rates of N fertilizer (0, 45, 90, and 135 kg N ha–1), and the spatial distribution in N status within and between rice populations can be compared in the gray scale (Fig. 6A). As the chemically analyzed and the image analyzed values from these three plots were closed to 1:1 ratio (r = 0.810, P < 0.010; r = 0.912, P < 0.010; r = 0.465), results demonstrated the feasibility and applicability of this system unit in providing rough assessments of canopy N status near ground (Fig. 6B). The plot with r = 0.465 was seemingly underestimated from N map due to the improper functioning of the device at the time of measurement.


Figure 6
View larger version (55K):
[in this window]
[in a new window]

  		Fig. 6. (A) Maps of canopy N status of rice paddy obtained from three different regions of a rice field applied with different rates of N fertilizer, taken with a simplified imaging system unit mounted on a mobile lifter. (B) Validation of the relationships between the chemically analyzed and the image analyzed N (%) values from three different regions of a rice field applied with different rates of N fertilizer.

 
The same system unit was also mounted on a helicopter flying 600 m aboveground surface to take image and reflectance data in first crop of 2002. As shown in Fig. 7 (top), spatial distribution of canopy N status across rice populations was observed by using the unit. However, the image analyzed values were slightly lower than those of the chemically analyzed N, and were loosely distributed along the 1:1 ratio line (r = 0.620, P < 0.001), with a slope of 1.122 and RMSE of 0.149 (Fig. 7 bottom). The observed loose relationship may be caused by growth variation among plants and atmospheric interference. Therefore, for practical application in the future, an atmospheric calibration in adjusting the bias of interference and a more refined geometric adjustment need to be incorporated for improving the accuracy. Nevertheless, since the system was assembled here in our lab with minimum cost compared to the commercial ones, we believe that the unit is affordable to general users and be freely mounted on platforms of various heights as desired. Recently Moran et al. (1997) and Daughtry et al. (2000) pointed out that, image-based remote sensing system may be the key leading to the success of sustainable and efficient agricultural practices. An inexpensive and easy-to-use field-portable imaging spectrometry like this one could have a promising potential in the near future.


Figure 7
View larger version (39K):
[in this window]
[in a new window]

  		Fig. 7. Top: Maps of canopy N status of rice paddy obtained from three different regions of a rice field applied with different rates of N fertilizer, taken with a simplified imaging system unit mounted on a helicopter. Bottom: Validation of the relationships between the chemically analyzed and the image analyzed N (%) values from three different regions of a rice field applied with different rates of N fertilizer.

 
In conclusion, this study demonstrated the feasibility of the derivative index dR/d{lambda}|735 along with other ratio-based SIs in assessing N concentration of rice plants at the panicle formation stage, and validated the applicability of the spectral model derived from the N–dR/d{lambda}|735 relationship used in a wide range of cropping seasons and locations. In conjunction with a self-developed simplified imaging system, the spectral model can also be used for mapping canopy N status within field from varied remote sensing platforms. The unit is capable of acquiring spatial information of N of the target regions and provide images with acceptable N assessment and variability. However, a more precise geometric and radiometric adjustments and atmospheric calibration are suggested for future field applications. Such procedure is critical to variable rate technology implemented in the site-specific management commercially (Moran et al., 1997).


   ACKNOWLEDGMENTS
 
This work is financially supported by Council of Agriculture (91AS-5.1.3-CI-C2 and 92AS-1.1.6-CI-C1) and Ministry of Education (under the ATU plan), Executive Yuan, Taiwan ROC.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Agricola
Right arrow Articles by Lee, Y.-J.
Right arrow Articles by Shen, Y.
Related Collections
Right arrow Forage Management
Right arrow Nutrient Management

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome