Research Papers

Estimation of agricultural soil properties with imaging and laboratory spectroscopy

[+] Author Affiliations
Tingting Zhang

Chinese Academy of Sciences, Institute of Remote Sensing Applications, State Key Laboratory of Remote Sensing Science, Beijing 100101, China

Indiana University-Purdue University, Department of Earth Sciences, 723 West Michigan Street, Indianapolis, Indiana 46202

Lin Li, Baojuan Zheng

Indiana University-Purdue University, Department of Earth Sciences, 723 West Michigan Street, Indianapolis, Indiana 46202

J. Appl. Remote Sens. 7(1), 073587 (Feb 11, 2013). doi:10.1117/1.JRS.7.073587
History: Received February 15, 2012; Revised December 23, 2012; Accepted January 18, 2013
Text Size: A A A

Open Access Open Access

* Present address: Geography Department, Virginia Tech, 115 Major Williams Hall, Blacksburg, Virginia 24061

Abstract.  Two EO-1 Hyperion images covering a Cicero Creek reservoir of central Indiana were analyzed using partial least squares (PLS) regression to estimate soil properties, including soil moisture, soil organic matter (SOM), total carbon (C), total phosphorus (P), total nitrogen (N), and clay content. PLS results for Hyperion image spectra were compared with those for laboratory measured spectra using several statistics, including the coefficient of determination (R2) and RPD (the ratio of standard deviation of sample chemical concentration to root mean square error). PLS was conducted in two phases: phase-1 used all samples for calibration to determine outliers and then models were recalibrated after outlier removal; phase-2 split the resulting samples from phase 1 into two subsets for calibration and validation, respectively. Based on R2 and RPD values, the results from the phase-1 calibration indicate that PLS can estimate all soil properties from laboratory spectra and some soil properties from Hyperion spectra, and the phase 2 results suggest that PLS can predict SOM, total C, and total N using Hyperion reflectance spectra. It was found that spectral resolution has impacts on the PLS performance in estimating the soil properties considered in this investigation.

Figures in this Article

As a common boundary where atmosphere, hydrosphere, geosphere, and biosphere interact, soil plays an important role in the transfer of materials and energy among them. Quantitative information of soil properties is often required for environmental monitoring, modeling, and precision agriculture practices. For example, soil organic matter (SOM), the largest carbon (C) stock of continental biosphere,1 is related to the size and capacity of soil microbial population2 and controls soil structural stability.3 As a major sink for carbon, agricultural soil should be managed effectively for mitigating the emission of greenhouse gases, including carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O).1 Soil moisture is related to energy exchange between soil and atmosphere, and favorable to the production of CO2, N2O, and CH4;4 like clay content, soil moisture is also important for plant growth and soil quality. Soil nutrient conditions drive the rate of fertilizer applications that is directly related to N2O emission,5 and to the soil capacity of consuming atmosphere CH4.6 Phosphorus (P) and nitrogen (N) have been identified as pollutants carried into water bodies from agriculture land where excess fertilizers are applied.7,8 With the increasing need for quantitative soil information, remote sensing has been considered as a promising tool for rapid quantifications of a single or multiple soil properties. Numerous studies show that different soil properties can be estimated using laboratory measured and simulated hyperspectral data.815 The success of these laboratory spectra-based studies naturally lead to the exploration of imaging spectroscopy for characterizing soil properties at large scales because imaging spectroscopy not only has the capability of acquiring spectral information at several hundred spectral bands as laboratory spectroscopy does but also provides a synoptic view which cannot be achieved by laboratory spectroscopy. Ben-Dor et al.16 used a DAIS-7915 hyperspectral airborne sensor to map organic matter, soil field moisture, soil saturated moisture, and soil salinity. Their research results are promising, but limited by the quality of the DAIS-7915 data. Stevens et al.17 compared the performance of Compact Airborne Spectrographic Imager (CASI) (405 to 950 nm) and CASI+ Shortwave Infrared Airborne Spectrographic Imager (SASI) (444 to 2500 nm) sensors in estimating soil organic carbon (SOC), and concluded that the CASI+ SASI sensor outperformed the CASI sensor due to its wider spectral range. Stevens et al.18 used an ASH-160 airborne spectrometer to map SOC of cropland soils, and achieved a mapping accuracy lower than that achieved with the in situ measured spectra. The degraded performance of the ASH-160 sensor was attributed to its low signal-to-noise ratio (SNR).18 Soil clay and calcium carbonate (CaCO3) were estimated with a hyperspectral mapper (HYMAP) image which provides 126 spectral bands in the spectral region of 400 to 2500 nm.19 Although the estimation of clay content with the HYMAP image was not as good as that with laboratory measured spectra, a strong correlation between clay content and the continuum removed band depth of the HYMAP image spectra demonstrated the transferability of the band depth-based calibration model from the laboratory to field and to HYMAP sensor measurement.19 Hively et al.20 used an airborne hyperspectral imaging spectrometer to map 19 soil analyses including C, N, and P. Their results were promising for predicting soil C, but they had difficulties to predict N and P.20 In contrast to a large number of airborne hyperspectral sensor applications, few attempts have been made to map soil properties with satellite hyperspectral images. Gomez et al.21 predicted SOC using Hyperion images and concluded that the accuracy of soil mapping with Hyperion images is limited by its low SNR and 30 m spatial resolution. They also demonstrated that the spectral resolution of hyperspectral reflectance data had an insignificant effect on the SOC prediction.21

Besides the quality of hyperspectral images, a feasible statistic method is required to reliably extract the soil information from hyperspectral data with hundreds to thousands of bands. A number of statistical methods have been used to relate soil properties to spectral reflectance data.2226 Among these methods, multiple linear regression analysis (MLR), principal component regression (PCR), and partial least squares regression (PLS) are commonly used to spectrally quantify soil properties.22 MLR uses a linear equation to correlate a response variable (i.e., chemical concentration) with two or more explanatory variables (i.e., spectral wavelength).23 The number of spectral wavelengths that could be used in MLR is limited because a larger number of spectral bands than the number of samples can result in rank deficiency problems. Both PCR and PLS are full-spectrum methods. PCR is simply principal component analysis (PCA) of spectra followed by a regression against chemical compositions,22 while PLS is a rotated PCA applied to both spectra and chemical compositions and then finds the best relationship between them.26 Recently, several new statistical tools have been utilized for soil mapping, such as artificial neural networks (ANN) and boosted regression trees (BRT). Brown et al.27 built BRT and PLS models with soils collected from all around the world for determining SOC, inorganic carbon, clay, cation exchange capacity (CEC), and iron (Fe). Leone et al.28 predicted sand, silt, clay, SOC, and CaCO3 by applying the ANN approach to reflectance spectra in the visible-near infrared-shortwave infrared (VIS-NIR-SWIR, 350 to 2500 nm) region and concluded that clay content and SOC could be predicted at high accuracy, sand and CaCO3 at moderate to relatively high accuracy, and silt at relatively low accuracy.

This study aims to: (1) evaluate the potential of Hyperion imagery for estimating soil moisture content, total carbon (C), total nitrogen (N), total phosphorous (P), SOM, and clay content; (2) examine the effects of spectral and spatial resolutions on the estimation of soil properties by comparing the estimates of the soil properties using Hyperion spectra to those resulting from laboratory measured spectra; (3) investigate whether the use of absorbance instead of reflectance can improve the prediction accuracy. Like neither Gomez et al.21 where only one soil property (i.e., SOC) was considered, nor Hively et al.20 where airborne hyperspectral images were used, this study used Hyperion images to simultaneously map multiple soil properties.

Study Area and Soil Samples

The study area is located in Cicero Creek Watershed of central Indiana (Fig. 1) in which the row-crop agriculture is a dominant type of land uses. According to the State Soil Geographic (STATSGO) data base (available on line at http://soildatamart.nrcs.usda.gov), main soil associations in Cicero Creek Watershed are Crosby-Treaty-Miami, Miami-Crosby-Treaty, Patton-Del Rey-Crosby, and Drummer-Toronto-Wingate. Fertilizers and pesticides have been the main source of pollutants discharged into the Morse Reservoir, and cause yearly algal blooms in this drinking water reservoir.

Graphic Jump LocationF1 :

The Cicero Creek watershed in central Indiana, USA. The soil sample locations and areas covered by Hyperion images are shown.

Soil samples were collected on the same day or within a few days after satellite image acquisitions. A total of 33 surface (0 to 2 cm) agricultural soil samples were collected and each of them was collected from about a 20×20cm area. The geographic coordinates of each sampling site were recorded at one meter accuracy using a global positioning system (GPS) instrument. Soil samples were kept fresh in sealed plastic bags (17×20cm2) and stored over ice in coolers before being transported to the laboratory.

Soil samples were stored in a laboratory refrigerator (4°C) before soil property analyses. Each soil sample was analyzed for the amount of SOM, moisture, total C, total N, total P, and clay. SOM was determined with the loss on ignition (LOI) method.29 Soil moisture was measured by the gravimetric method. Total N and total C were measured by dry combustion at 1020°C using Costech 4010 elemental analyzer (Costech Analytical Technologies Inc., Valencia, California). Total P was measured by using strong acid digestion followed by the molybdate blue technique and detection with a Shimadzu scanning spectrophotometer at 880 nm.30 Clay content was determined through freeze drying, centrifuge and particle size analysis.31 Clay particle size was analyzed with a Malvern Mastersizer 2000 laser particle size analyzer after organic matter had been eliminated.

Laboratory Spectra Measurements

The field-moist soils were scanned in the laboratory using an ASD spectrometer (Analytical Spectral Devices, Boulder, Colorado) with 2151 bands in the 350 to 2500 nm spectral region and spectral resolution of 1 nm. A 50 W ASD Pro lamp (Analytical Spectral Devices, Boulder, Colorado) was used as artificial illumination with approximate 30 deg zenith angle. The fiber-optical head of the spectrometer pointed at the nadir viewing angle at a height of 30 cm above the soil surface. An 8 deg fore-optic was used, resulting in a 4.2 cm diameter field of view. Spectral measurement began with scanning a white spectralon reference panel. Three spectra were acquired for each soil sample by rotating the sample 120 deg for the second and third measurements. To reduce the effect of soil texture on measured spectra, the average of three spectral measurements of each sample was used in spectral-compositional modeling.

Hyperion Image Acquisition and Processing

The Hyperion imaging spectrometer is on board the NASA EO-1 satellite which was launched on November 21, 2000. Hyperion images are characterized by a total of 242 channels in 10 nm spectral intervals over the spectral region of 356 to 2577 nm, and acquired at 30 m spatial resolution with approximate 501 SNR. A Hyperion scene has 7.7 km cross-track width with 42 or 185 km along-track length. Out of the 242 collected channels, bands 1–7 (356 to 417 nm), bands 58 to 70 (collected by the visible near infrared (VNIR) instrument), bands 71–76 (collected by the SWIR instrument), and bands 225–242 (2406 to 2577 nm) are not calibrated. Therefore, a Hyperion image spectrum has a total of 198 bands from 427 to 2395 nm available for further analysis.

The EO-1 satellite does not acquire data continuously and its sensors are only activated to collect specific scenes upon a request. Two Hyperion scenes were acquired for the study area, respectively on April 24 and May 7, 2007. The images were acquired at around 10:00 AM local time (16:00 GMT). Figure 1 shows the area where Cicero Creek Watershed was covered by two Hyperion images with 7288×1955pixels each.

The Hyperion images were delivered from United States Geological Survey (USGS) in radiance, and they were radiometrically and geometrically corrected so that reflectance spectra could be extracted from them and related to a specified soil property. The following steps recommended by Goddard Space Flight Center14 were taken for radiometric calibration: (1) a pixel shift was applied to samples 129 to 256 in the shortwave infrared (SWIR) region to co-register this portion of the data with the visible infrared observations; (2) the VNIR bands were multiplied by a scale factor of 1.08, and the SWIR bands were multiplied by a scale factor of 1.18; and (3) the wavelength values were increased by 2 nm for all bands. These steps were completed using the Band Math and Edit ENVI Header functions of ENVI 4.5. The radiance data were then converted into surface reflectance using the Atmospheric Correction Now (ACORN) software, a commercialized software package for atmospheric calibration.32 For the geometric calibration, the Hyperion images were rectified by referencing 2006 aerial photographs of the Hamilton and Tipton counties, Indiana. High spatial resolution (2 m) aerial photographs were degraded to 30-m spatial resolution, and ground control point (GCPs) pairs were manually selected from the referencing aerial photo and the Hyperion images. A bilinear warping method was applied to project each Hyperion image into the coordination of Universal Transverse Mercator (UTM) Zone 16, NAD-1983 Datum. The registration accuracy was assessed using the ENVI dynamic overlay function.

As the two Hyperion scenes were acquired on different dates, atmospheric conditions and the sensor behaviors may not have been consistent, such that the same ground object could have been spectrally different in the two scenes. A normalization correction was applied to the Hyperion scene acquired on May 7, 2007, in which the image acquired on April 24, 2007, was used as a master image. We then mosaicked the two Hyperion images, and extracted spectra of each sampling site to build spectral-chemical compositional models. Because there were several noisy bands corrupted by atmospheric water absorption, we excluded those noisy bands and retained 150 bands for partial least squares regression. To compare the results for laboratory spectra with those for Hyperion spectra, the spectral bands and soil samples for both datasets are required to match each other. We resampled the laboratory spectra to the Hyperion spectral resolution, resulting in 150 bands from 428 to 2357 nm.

Partial Least Squares Modeling

Partial least squares (PLS) modeling was used to build relationships between soil property parameters and hyperspectral data. PLS is a standard multivariate regression method developed by Herman Wold.33,34 PLS uses a few eigenvectors of the explanatory variables so that the corresponding scores not only explain the variance of explanatory variables but also have high correlation with response variables. A simplified PLS model consists of two outer relations shown in Eqs. (1) and (2) that describe the eigenstructure decomposition of both the matrix containing the explanatory variables (i.e., spectral bands) and the matrix containing the response variables (i.e., the abundance of SOM), and an inner relation shown in Eq. (3) that links the resultant score matrices from these two eigenstructure decompositions.35Display Formula

X=TD+E.(1)
Display Formula
Y=UQ+F.(2)
Display Formula
U=BT.(3)

The first outer relation is derived by applying principal component analysis (PCA) to X, resulting in the score matrix T and the loading matrix D plus an error matrix E. In the similar way, the second outer relation is derived by decomposing Y into the score matrix U, the loading matrix Q, and the error term F. The inner relation is a multiple linear regression between the score matrices U and T in which B is a regression coefficient matrix determined via least square minimization. The prime (′) represents matrix transpose. Y is computed as Display Formula

Y=TBQ+F.(4)

The goal of PLS modeling is to minimize the norm of F while maximizing the covariance between X and Y by the inner relationship. Because the above mentioned two separate PCAs’ approaches to the derivation of PLS factors are not the best possible and could result in a weak correlation for the inner relation, a method resulting in a strong inner relation between T and U was used in this study.3537 Selecting the optimal number of latent variables is essential for building a robust PLS model. The leave-one-out cross-validation method was used to determine the optimal number of latent variables. Given a set of m samples, m-1 samples are used to develop a calibration model and the concentration of the left out sample is predicted using the calibration model. This process is repeated until each sample has been excluded once. The predicted error sum of squares (PRESS) can be calculated as Display Formula

PRESS=i=1m(y^(i)yi)2,(5)
where y^(i) and yi are the estimated and actual concentration for the left-out sample, respectively. Root mean square error of cross-validation (RMSECV) for each PLS model with a given number of latent variables is expressed as Display Formula
RMSECV=PRESSj/m,(6)
where j is the number of latent variables. In general, the number of latent variables is considered to be optimal when it yields the minimal RMSECV.

For a specified soil property parameter, PLS modeling was carried out in two phases: the first phase to detect outliers and the second phase to conduct calibration and validation after removing outliers. In phase 1, two statistical indices, namely the leverage and the Studentized residual, were used to determine outliers. The leverage evaluates the influence that a given sample has on a PLS model and is defined as the variance of the vector of PLS factors for the sample weighted by the covariance matrix of the factor score matrix T for the calibration dataset.37 The Studentized residual is the quotient resulting from division of a residual by an estimate of its standarddeviation. The studentized residual is an important technique to detect outliers, as it indicates the lack of fit for the content of a sample, and is assumed to obey the normal (mean zero and unit variance) distribution. Samples with either Studentized residuals 3 or leverages 3+3n/m were identified as an outlier.37 Here n is the optimal number of PLS factors. The identified outliers were discarded in the second phase PLS analysis. For the second phase PLS analysis, the dataset was divided into two subsets: one for calibration and the other for validation. Separation between the calibration and validation subset was achieved by sorting the samples in the descent or ascent order, and then randomly selecting 30% of the samples for validation. In both phases, the leave-one-out cross-validation method was used to determine the optimal number of PLS factors.

Previous studies have shown that data pretreatment improved the PLS performance.36,38 While a wide range of pretreatment methods are available including mean centering, auto scale, derivatives, smoothing, multiplicative scatter correction, and orthogonal signal correction,35,38 we applied mean centering to the dataset. Mean centering subtracts the means of individual spectral bands from spectra data and similar subtractions were applied to soil property contents. Mean centering was used because it is simple and facilitates the interpretation of PLS results.

The prediction accuracy was assessed on the basis of the coefficient of determination (R2), the slope (b) of a regression line, and the root mean square error of calibration (RMSEC) and prediction (RMSEP). In addition, the following measures were used to evaluate prediction accuracy: Display Formula

MAE=m|YpredYmeas|m.(7)
Display Formula
Bias=m|YpredYmeas|m.(8)
Display Formula
NU=(1b)2×var(Ymeas).(9)
Display Formula
LC=(1R2)×var(Ypred).(10)
Display Formula
RPD=SD/RMSE.(11)

Mean absolute error (MAE) was used to estimate how predictions are close to the measurement. Bias represents overall difference between predicted and measured values. NU represents the nonunity regression line and LC is lack of correlation. The values of NU and LC are both zero when a perfect correlation is obtained.39 RPD (ratio of performance deviation) is the ratio of standard deviation (SD) of sample chemical concentration to root mean square error (RMSE) from a PLS model.40

Many researchers used RPD to evaluate the stability and effectiveness of PLS calibration and prediction.21,4143 We used RPD by following Chang et al.44 whereby results from PLS modeling are classified as good if RPD>2, as satisfactory and can be improved using different calibration strategies if 1.4RPD2, and as unreliable with RPD<1.4.

Outlier Determination

Five samples were excluded in the PLS phase 1 screening process: one sample was outside of the Hyperion image coverage; three outliers were visually identified because of poor spectral quality; and one outlier was detected due to a leverage value of 0.33 larger than (3+3*2)/29 (Fig. 2). There were 28 samples remaining for the PLS phase 1 and 2 analyses after the outlier removal. Table 1 showed the range, mean value, and standard deviation (SD) of each soil property dataset before and after outlier removal. The mean values of each soil property changed slightly and SD values decreased after outlier removal (Table 1).

Graphic Jump LocationF2 :

A leverage and Studentized residual plot for estimating SOM.

Table Grahic Jump Location
Table 1Minimum, maximum, mean, and standard deviation of six soil parameters before and after outlier removal.
PLS Modeling—Phase 1

The PLS results for laboratory measured, resampled laboratory, and Hyperion spectra are presented in Table 2. The RPD values indicate that we can estimate moisture, SOM, and total C using all the above-mentioned spectral data. PLS resulted in good estimation with laboratory data, and satisfactory with Hyperion spectra for total P, total N, and clay content. For the laboratory spectra, PLS resulted in the best estimation for moisture with R2=1.00, and the poorest estimation for clay content (R2=0.79) (Fig. 3). For the resampled laboratory spectra, PLS yielded the best estimation for SOM (R=0.84), and the lowest R2 value for clay content (R2=0.68) (Fig. 4), while for Hyperion spectra, PLS provided the best estimation for SOM (R2=0.81), and the poorest estimation for total P (R2=0.60) (Fig. 5).

Table Grahic Jump Location
Table 2Calibration results with laboratory spectra and Hyperion spectra of all usable samples.
Graphic Jump LocationF3 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with laboratory spectra.

Graphic Jump LocationF4 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with the resampled laboratory spectra.

Graphic Jump LocationF5 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with Hyperion image spectra.

The results shown in Table 2 and Figs. 345 demonstrate that (1) in terms of R2 and RPD, the PLS model resulted in better estimates for soil moisture and SOM than other soil properties when the original laboratory spectra were used; (2) a similar pattern is observed from the results for the resampled laboratory spectra, but the resultant R2 and RPD values are smaller than those with the original laboratory spectra for most of soil properties except for total P; and (3) PLS resulted in poorer estimates for most properties using Hyperion spectra than using the resampled laboratory data. These results have several implications for remote estimation of soil properties. First, these results indicate that spectral resolution had impacts on the PLS performance in estimating the soil properties considered in this study, which is in disagreement with the statement by Gomez et al.21 that the spectral resolution did not change accuracy of the PLS model for estimating SOC. Second, the image spatial resolution has effects on spectral estimation of individual soil properties. This is consistent with the observation by Gomez et al.21 that the decrease in the accuracy of PLS estimated SOC may be due to the noise and the Hyperion 30 m spatial resolution. With the presence of the inconsistent observations between this study and the previous study by Gomez et al.21 regarding the effects of the spectral and spatial resolutions on the PLS performance for estimating different soil properties, we initially attributed these to the use of R2 as the only criterion for the model evaluation. To verify this speculation, other model evaluation criteria were also examined. The results from the model evaluation with MAE, NU, and LC statistics were consistent with the value of R2 and RPD. Based on bias values, the PLS models tended to underestimate most properties.

PLS Modeling—Phase 2
Calibration and validation with laboratory and Hyperion spectra

Eight samples were randomly selected for validation from the outlier removed dataset, and 20 remaining samples were used in calibration for estimating individual soil property content. Table 3 listed the chemical range of both the calibration and the validation sets and the number of samples in each dataset. From Table 3, it can be seen that there were no significant differences between both datasets for all of the soil property parameters when mean content were compared. Standard deviations of soil samples for both calibration and validation were similar for all of the examined properties. PLS modeling was conducted with laboratory spectra and Hyperion spectra and the results were summarized in Table 4 and Figs. 678.

Table Grahic Jump Location
Table 3Calibration and validation datasets: the number of samples, chemical range, mean, and standard deviation for soil properties considered in this study.
Table Grahic Jump Location
Table 4Statistical parameters resulting from the PLS calibration and validation with laboratory and Hyperion reflectance spectra.
Graphic Jump LocationF6 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with laboratory spectra.

Graphic Jump LocationF7 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with the resampled laboratory spectra.

Graphic Jump LocationF8 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with Hyperion image spectra.

Both R2 and RPD statistics indicated that PLS with laboratory measured spectra achieved good calibration results for SOM, moisture, total N, and clay content, respectively, with R20.91 and RPD>2; the calibrations for total P and total C were not as good as for the other properties, but satisfactory. PLS with the resampled laboratory spectra yielded accurate estimates for moisture, SOM, and clay content, but the R2 and RPD values were lower than those resulting from PLS with 2151 bands laboratory spectra. The PLS results for the Hyperion image spectra demonstrated good calibration for moisture and clay content with R2>0.79 and RPD>2, and satisfactory calibration for SOM, total C, and total N with RPD>1.5.

In general, the validation results indicated the PLS capability for predicting soil properties from hyperspectral data. All RPD values for the PLS validation were less than two except the value for total P with the resampled laboratory spectra, implying that almost all soil properties cannot be predicted with hyperspectral data at a good accuracy. PLS with the laboratory spectra (2151 bands) had satisfactory validation for total P, total C, total N, and clay content. PLS using the resampled laboratory spectra achieved a good validation for total P and a satisfactory validation for moisture, total C, and clay content, respectively. With Hyperion spectra, the PLS validation was satisfactory for SOM, total C, and total N.

Calibration and validation with laboratory and Hyperion absorbance spectra

All results reported in the previous sections were obtained through PLS modeling of the reflectance spectra of the soil samples. According to Whiting et al.25 the relationship between spectral reflectance and soil moisture is exponential, and a similar relation may also apply for other soil properties. A number of studies have shown that the logarithmic transformation of reflectance (i.e., log1/R) can improve the PLS performance for estimating soil properties.45 In this study, the PLS models using the log1/R value of the laboratory spectra and Hyperion spectra were also evaluated and the results are summarized in Table 5 and Figs. 91011. Table 5 demonstrates that PLS generated the better validation for estimating soil moisture, SOM, and total C with laboratory spectra after transformation of reflectance to absorbance based on the higher R2 and RPD values than that using reflectance for these properties. Moisture can be predicted accurately using logarithmic-transformed laboratory spectra. RPD values for total C increased with resampled laboratory and Hyperion absorbance data. PLS was able to predict most soil properties using different spectral data with satisfactory RPD values, except for total N with laboratory spectral data. PLS had the poorest prediction for moisture and clay using Hyperion absorbance data.

Table Grahic Jump Location
Table 5Statistical parameters resulting from the PLS calibration and validation with laboratory and Hyperion absorbance spectra.
Graphic Jump LocationF9 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with laboratory absorbance spectra.

Graphic Jump LocationF10 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with the resampled laboratory absorbance spectra.

Graphic Jump LocationF11 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with Hyperion absorbance spectra.

PLS was applied to laboratory measured reflectance spectra, resampled lab spectra, and Hyperion image spectra. These datasets were different in spectral (e.g., original lab spectral versus resampled) or spatial (resampled lab versus Hyperion spectra) resolution, or both (original lab versus Hyperion spectra). We compared the PLS performances for these datasets. The effects of spectral and spatial resolutions on the PLS performance for predicting soil properties were evaluated based on the coefficient of determination (R2) and RPD values resulting from the phase-2 validation (Table 4). PLS with the laboratory resampled spectra resulted in more accurate estimates for moisture (RPD=1.42) and total P (RPD=2.18) than those with the laboratory data (Table 4). PLS with Hyperion image spectra produced higher RPD for predicting SOM (RPD=1.91) and total N (RPD=1.41) than with the resampled laboratory data (Table 4). The PLS performance for predicting these properties was not sensitive to the degraded spatial resolution, i.e., 30 m Hyperion image resolution. When PLS was applied to the original laboratory spectra for the estimation of soil moisture, the spectral absorption regions around 1400 and 1900 nm due to water were retained, but the result was not reliable and might have been indicative of overfitting (Table 4). After water absorption regions were removed, the prediction was satisfactory with the resampled laboratory spectra, but unreliable with Hyperion image spectra. This is likely due to the fact that even if some bands within the two water absorption regions around 1400 and 1900 nm were removed, some other spectral bands that are adjacent to these two absorptions and affected by the shoulders of the two absorptions were still present in Hyperion image data. The PLS prediction of moisture content with the original and resampled laboratory absorbance spectra demonstrates better accuracies than those using the corresponding reflectance spectra. For example, the PLS validation for moisture on original laboratory spectra generated RPD larger than two. A low R2 value for the PLS calibration of moisture with the laboratory absorbance data was driven by a largely deviating soil sample that was not identified as an outlier using the criteria of Studentized residual and leverage (Table 5 and Figs. 9 and 10).

In both phase 1 and phase 2, PLS achieved good calibration for SOM (Tables 2 and 4). These results are expected and consistent with previous studies which show that SOM can be reliably estimated from lab or in situ measured reflectance spectra.17,46,47 However, the PLS validation in phase 2 yielded poor estimates for SOM using laboratory data with RPD values of less than 1.4 (Table 4). This result is comparable to those from several investigations in which RPD values of less than 1.4 were reported for SOC (e.g., Stevens et al.18). One reason for this poor PLS performance for estimating SOM could be the narrow range of SOM in the validation dataset. Due to the complexity of soil, a dataset with a wide range of SOM content is required to achieve reliable prediction of SOM.21 Another reason may have contributed to exponential effect of increasing moisture on the apparent strength of absorption features of SOM.48 However, this effect due to soil moisture is somehow accommodated by converting spectral reflectance into absorbance49 as suggested by the good or satisfactory validation of using absorbance data.

SOM consists of organic carbon, organic nitrogen, and organic phosphorus. In our dataset, total C and total N show high correlations to SOM with a correlation coefficient of 0.96 and 0.93, respectively. The correlation coefficient between total C and total N is 0.94, indicating that mineral portion is less important in total C and N of the analyzed soils, as expected for soil samples that were collected from cultivated fields. These correlations also explain why predictions for total C and total N had a similar trend, and why they are comparable to the findings by Wetterlind et al.,50 who reported fairly stable estimation for total N highly correlated to SOM. Compared to the results reported by Hively et al.20 for estimating C, a lower R2 (0.48) value for validation with Hyperion reflectance data may be associated with the increased radiometric noise from space borne data and coarse spatial resolution.

In contrast to Chang et al.,44 who poorly predicted total P, we predicted this parameter satisfactorily with the laboratory spectra used in this study. However, consistent with the results reported in Hively et al.20 the poor prediction for total P with Hyperion data was obtained, which may indicate a low spectral response of total P at the Hyperion spectral resolution.

The accuracy of the PLS model for clay was low, especially with Hyperion data (Table 4). One reason could be the narrow range of clay content as shown in Table 1, i.e., the variation of clay content within such small range could not be well characterized spectrally.51 The most important reason was the masking effects of soil water content on the clay absorption at 2200 nm. This subdued clay absorption at 2200 nm was caused by increased soil moisture, and exhibited lowered SWIR reflectance and reduced spectral contrast. Therefore, the absorption of soil moisture in the SWIR region may have degraded the correlation of the clay absorption to clay content. SOM might have a similar effect but should not be as significant as soil moisture as the former has a weaker effect on the SWIR reflectance than the latter. Lagacherie et al.19 and Gomez et al.52 reported a degraded model performance when laboratory spectra were replaced with HYMAP image spectra for clay estimation, indicating the effect of the spatial resolution of spectral data on the clay prediction accuracy. Nonetheless, the SNR could be more important factor that determines the PLS performance. A low SNR of the Hyperion image in the SWIR region could make the spectral response of clay at 2200 nm more vulnerable to the spectral masking effect due to soil moisture. Along-track striping of the VNIR and SWIR data from Hyperion images may also be a source of errors for clay prediction.53

One may question the appropriateness of matching one soil sample to a 30×30m Hyperion pixel. We suggest that for the area investigated in this study each sample from a visually 30×30m homogenous field area should be representative because (1) each soil sample was collected within a 30×30m surface area located at least 30 m away from the field edges to avoid the spectral contamination from the paved road and perennial vegetation; (2) the study area is flat and dominated by large corn and soybean fields with relatively homogeneous soil properties. It should be pointed out that observations for calibration and validation could be different when PLS is applied to soil samples with larger property variability or those collected in other study regions with soil types different from those investigated here. Additional soil samples must be collected and added into the current dataset for further analysis in order to assess the PLS performance thoroughly.

This study has used PLS modeling of laboratory data with high and low spectral resolution and Hyperion data for estimating soil constituents: soil moisture, total P, SOM, total C, total N, and clay content. The PLS model was built first with all soil samples to determine outliers, and then the soil samples excluding the outliers were used to evaluate the capability of PLS to predict soil properties using remotely sensed spectral data. Overall, the PLS model can predict half of the studied soil properties satisfactory with Hyperion image spectra except for moisture, total P, and clay content. The PLS prediction for soil moisture and clay content was unreliable using Hyperion image reflectance spectra, possibly due to the low SNR of Hyperion spectra. Compared to reflectance spectra, absorbance spectra can be used to accommodate the exponential effect of soil moisture on soil spectra and improve the PLS performances for estimating soil moisture and SOM with laboratory data, and total C with the resampled laboratory and Hyperion data. PLS regression on Hyperion data failed to effectively predict clay content, which primarily originated from the interfering effect of soil moisture on the clay absorption feature and the low-SNR of Hyperion data in the SWIR region. Using satellite hyperspectral data with a SNR equivalent to those for HyMap and AVIRIS is expected to improve the spectral estimation of clay content.

The first author received an 18-month visiting scholar fellowship from China Scholarship Council. This research was funded by five programs: (1) CAS Knowledge Innovation Program (Grant No. KZCX2-EW-320); (2) Research on spectral response mechanism of the lake sediments in Lop Nor area and reconstruction of indicative information for environmental evolution (Grant No. OFSLRSS201106) provided by State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing Applications, Chinese Academy of Sciences; (3) Key technology researches on Soil and water environment parameters in Qinghai Lake watershed (Grant No. 2012BAH31B02-03) provided by Department of science and technology of Qinghai Province; (4) The fund of State Key Laboratory of Remote Sensing Science (Y1Y00201KZ); (5) Research Support Funds Grant (RSFG) program provided by Indiana University-Purdue University at Indianapolis.

Mondini  C., Sequi  P., “Implication of soil C sequestration on sustainable agriculture and environment,” Waste Manage.. 28, (4 ), 678 –684 (2008), CrossRef. 0956-053X 
van Veen  J. A., Kuikman  P. J., “Soil structural aspects of decomposition of organic matter by micro-organisms,” Biogeochemistry. 11, (3 ), 213 –233 (1990), CrossRef. 0168-2563 
Doran  J. W., Sarrantino  M., Liebig  M. A., “Soil health and sustainability,” Adv. Agron.. 56, , 1 –54 (1996), CrossRef. 0065-2113 
Morishita  T. et al., “CO2, CH4 and N2O fluxes from a larch forest soil in central Siberia,” in Symptom of Environmental Change in Siberian Permfrost Region. , Hatano  R. , Guggenberger  G., Eds., pp. 1 –9,  Hokkaido University Press ,  Sapporo  (2006).
Hickman  J. E. et al., “Current and future nitrous oxide emission from African agriculture,” Curr. Opin. Environ. Sustainabil.. 3, (5 ), 370 –378 (2011), CrossRef. 1877-3435 
Smith  K. A. et al., “Exchange of greenhouse gases between soil and atmosphere: interactions of soil physical factors and biological processes,” Eur. J. Soil Sci.. 54, (4 ), 779 –791 (2003), CrossRef. 1351-0754 
Qafoku  O. S. et al., “Rapid methods to determine potentially mineralizable nitrogen in broiler litter,” J. Environ. Qual.. 30, (1 ), 217 –221 (2001), CrossRef. 0047-2425 
Bogrekci  I., Lee  W. S., “Spectral phosphorus mapping using diffuse reflectance of soils and grass,” Biosyst. Eng.. 91, (3 ), 305 –312 (2005), CrossRef. 1537-5110 
Ben-Dor  E., Banin  A., “Near-infrared reflectance analysis of carbonate concentration in soils,” Appl. Spectrosc.. 44, (6 ), 1064 –1069 (1990), CrossRef. 0003-7028 
Ben-Dor  E., Banin  A., “Visible and near-infrared (0.4–1.1 μm) analysis of arid and semiarid soils,” Remote Sens. Environ.. 48, (3 ), 261 –274 (1994), CrossRef. 0034-4257 
Palacios-Orueta  A., Ustin  S. L., “Remote sensing of soil properties in the Santa Monica mountains I. spectral analysis,” Remote Sens. Environ.. 65, (2 ), 170 –183 (1998), CrossRef. 0034-4257 
Ingleby  H. R., Crowe  T. G., “Reflectance models for predicting organic carbon in Saskatchewan soils,” Can. Agric. Eng.. 42, (2 ), 57 –63 (2000). 0045-432X 
Thomasson  J. A. et al., “Soil reflectance sensing for determining soil properties in precision agriculture,” Trans. ASAE.. 44, (6 ), 1445 –1453 (2001). 0001-2351 
Asner  G. P., Heidebrecht  K. B., “Imaging spectroscopy for desertification studies: comparing AVIRIS and EO-1 hyperion in Argentina drylands,” IEEE Trans. Geosci. Remote Sens.. 41, (6 ), 1283 –1296 (2003), CrossRef. 0196-2892 
Schwanghart  W., Jarmer  T., “Linking spatial patterns of soil organic carbon to topography—a case study from south eastern Spain,” Geomorphology. 126, (1–2 ), 252 –263 (2011), CrossRef. 0169-555X 
Ben-Dor  E. et al., “Mapping of several soil properties using DAIS-7915 hyperspectral scanner data—a case study over clayey soils in Israel,” Int. J. Remote Sens.. 23, (6 ), 1043 –1062 (2002), CrossRef. 0143-1161 
Stevens  A. et al., “Detection of carbon stock change in agricultural soils using spectroscopic techniques,” Soil Sci. Soc. Am. J.. 70, (3 ), 844 –850 (2006), CrossRef. 1435-0661 
Stevens  A. et al., “Laboratory, field and airborne spectroscopy for monitoring organic carbon content in agricultural soils,” Geoderma. 144, (1–2 ), 395 –404 (2008), CrossRef. 0016-7061 
Lagacherie  P. et al., “Estimation of soil clay and calcium carbonate using laboratory, field and airborne hyperspectral measurements,” Remote Sens. Environ.. 112, (3 ), 825 –835 (2008), CrossRef. 0034-4257 
Hively  W. D. et al., “Use of airborne hyperspectral imagery to map soil properties in tilled agricultural fields,” Appl. Environ. Soil Sci.. 2011, , 1 –13 (2011), CrossRef. 1687-7675 
Gomez  C., Viscarra Rossel  R. A., McBratney  A. B., “Soil organic carbon prediction by hyperspectral remote sensing and field vis-NIR spectroscopy: an Australian case study,” Geoderma. 146, (3–4 ), 403 –411 (2008), CrossRef. 0016-7061 
Sudduth  K. A., Hummel  J. W., “Portable near-infrared spectrophotometer for rapid soil analysis,” Trans. ASAE. 36, (1 ), 185 –193 (1993). 0001-2351 
Hummel  J. W., Gaultney  L. D., Sudduth  K. A., “Soil property sensing for site-specific crop management,” Comput. Electron. Agr.. 14, (2–3 ), 121 –136 (1996), CrossRef. 0168-1699 
Scharf  P. C. et al., “Remote sensing for nitrogen management,” J. Soil Water Conserv.. 57, (6 ), 518 –524 (2002). 0022-4561 
Whiting  M. L., Li  L., Ustin  S. L., “Predicting water content using Gaussian model on soil spectra,” Remote Sens. Environ.. 89, (4 ), 535 –552 (2004), CrossRef. 0034-4257 
Viscarra Rossel  R. A. et al., “Visible, near infrared, mid infrared or combined diffuse reflectance spectroscopy for simultaneous assessment of various soil properties,” Geoderma. 131, (1–2 ), 59 –75 (2006), CrossRef. 0016-7061 
Brown  D. J. et al., “Global soil characterization with VNIR diffuse reflectance spectroscopy,” Geoderma. 132, (3–4 ), 273 –290 (2006), CrossRef. 0016-7061 
Leone  A. P. et al., “Prediction of soil properties with VIS-NIR-SWIR reflectance spectroscopy and artificial neural networks: a case study on three pedoenvironments of the Campania region, Italy,” in  5th International Congress of the European Society for Soil Conservation Soils changing in a changing World: The Soils of Tomorrow , Dazzi  C. , Costantini  E., Eds., pp. 685 –698,  Catena Verlag ,  Palermo, Italy  (2008).
Heiri1  O., Lotter  A. F., Lemcke  G., “Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results,” J. Paleolimnol.. 25, (1 ), 101 –110 (2001), CrossRef.
Pierzynski  G. M. , Ed., Methods of Phosphorus Analysis for Soils Sediments, Residuals and Waters. , pp. 39 –49,  North Carolina State University ,  Raleigh, North Carolina  (2004).
Gee  G. W., Bauder  J. W., “Particle size analysis,” Methods of soil analysis, Part 1: Physical and mineralogical methods. , 2nd ed., Klute  A. , Ed., pp. 404 –408,  American Society of Agronomy ,  Madison, Wisconsin  (1986).
“Analytical Imaging and Geophysics LLC,” ACORN User’s Guide. , Stand Alone Version (2001).
Wold  H., “Nonlinear estimation by eterative least squares procedure,” Research Papers in Statistics. , David  F. N. , Neyman  J., Eds., pp. 441 –444,  Wiley ,  New York  (1966a).
Wold  H., “Estimation of principal components and related models by iterative lease squares,” in Multivariate Analysis. , Krishnaiah  P. R., Ed., pp. 391 –420,  Elsevier ,  New York  (1966b).
Geladi  P., Kowalski  B. R., “Partial least-squares regression: a tutorial,” Anal. Chim. Acta. 185, (1 ), 1 –17 (1986), CrossRef. 0003-2670 
Haaland  D. M., Thomas  E. V., “Partial least-squares methods for spectral analyses.1. Relation to other quantitative calibration methods and the extraction of qualitative information,” Anal. Chem.. 60, (11 ), 1193 –1202 (1988), CrossRef. 0003-2700 
Martens  H., Næs  T., Multivariate Calibration. , pp. 73 –296 John Wiley and Sons ,  New Jersey  (2002).
Azzouza  T. et al., “Comparison between different data pre-treatment methods in the analysis of forage samples using near-infrared diffuse reflectance spectroscopy and partial least-squares multivariate calibration method,” Anal. Chim. Acta. 484, (1 ), 121 –134 (2003), CrossRef. 0003-2670 
Gauch  H. G., Hwang  J. T. G., Fick  G. W., “Model evaluation by comparison of model-based predictions and measured values,” Agron. J.. 95, (6 ), 1442 –1446 (2003), CrossRef. 0002-1962 
Nduwamungu  C. et al., “Opportunities for, and limitations of, near infrared reflectance spectroscopy applications in soil analysis: a review,” Can. J. Soil Sci.. 89, (5 ), 531 –541 (2009), CrossRef. 0008-4271 
Willams  P. C., Sobering  D. C., “Comparison of commercial near infrared transmittance and reflectance instuments for analysis of whole grains and seeds,” J. Near Infrared Spectrosc.. 1, (1 ), 25 –32 (1993), CrossRef. 0967-0335 
Brown  D. J., Bricklemyer  R. S., Miller  P. R., “Validation requirements for diffuse reflectance soil characterization models with a case study of VNIR soil C prediction in Montana,” Geoderma. 129, (3–4 ), 251 –267 (2005), CrossRef. 0016-7061 
Du  C. et al., “Determination of soil properties using Fourier transform mid-infrared photoacoustic spectroscopy,” Vibrat. Spectrosc.. 49, (1 ), 32 –37 (2009), CrossRef. 0924-2031 
Chang  C. W. et al., “Near infrared reflectance spectroscopy-principal components regression analyses of soil properties,” Soil Sci. Soc. Am. J.. 65, (2 ), 480 –490 (2001), CrossRef. 1435-0661 
Dalal  R. C., Henry  R. J., “Simultaneous determination of moisture, organic carbon and total nitrogen by near infrared reflectance spectrophotometry,” Soil Sci. Soc. Am. J.. 50, (1 ), 120 –123 (1986), CrossRef. 1435-0661 
Hummel  J. W., Sudduth  K. A., Hollinger  S. E., “Soil moisture and organic matter prediction of surface and subsurface soils using an NIR soil sensor,” Comput. Electron. Agr.. 32, (2 ), 149 –165 (2001), CrossRef. 0168-1699 
Ben-Dor  E. et al., “Using imaging spectroscopy to study soil properties,” Remote Sens. Environ.. 113, (Suppl. 1 ), S38 –S55 (2009), CrossRef. 0034-4257 
Liu  W., Baret  F. et al., “Relating soil surface moisture to reflectance,” Remote Sens. Environ.. 81, (2–3 ), 238 –246 (2002), CrossRef. 0034-4257 
Bogrekci  I., Lee  W. S., “Effects of soil moisture content on absorbance spectra of sandy soils in sensing phosphorus concentrations using UV-VIS NIR spectroscopy,” Trans. ASABE. 49, (4 ), 1175 –1180 (2006).
Wetterlind  J., Stenberg  B., Söderström  M., “Increased sample point density in farm soil mapping by local calibration of visible and near infrared prediction models,” Geoderma. 156, (3–4 ), 152 –160 (2010), CrossRef. 0016-7061 
Selige  T., Böhner  J., Schmidhalter  U., “High resolution topsoil mapping using hyperspectral image and field data in multivariate regression modeling procedures,” Geoderma. 136, (1–2 ), 235 –244 (2006), CrossRef. 0016-7061 
Gomez  C., Lagacherie  P., Coulouma  G., “Continuum removal versus PLSR method for clay and calcium carbonate content estimation from laboratory and airborne hyperspectral measurements,” Geoderma. 148, (2 ), 141 –148 (2008), CrossRef. 0016-7061 
Staenz  K. et al., “Retrieval of surface reflectance from Hyperion radiance data,” in  Proc. IEEE Int. Geosci. and Remote Sens. Symposium , Vol. 3, pp. 1419 –1421,  IEEE ,  Ontario, Canada  (2002).

Biographies and photographs of the authors are not available.

© The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Citation

Tingting Zhang ; Lin Li and Baojuan Zheng
"Estimation of agricultural soil properties with imaging and laboratory spectroscopy", J. Appl. Remote Sens. 7(1), 073587 (Feb 11, 2013). ; http://dx.doi.org/10.1117/1.JRS.7.073587


Figures

Graphic Jump LocationF1 :

The Cicero Creek watershed in central Indiana, USA. The soil sample locations and areas covered by Hyperion images are shown.

Graphic Jump LocationF2 :

A leverage and Studentized residual plot for estimating SOM.

Graphic Jump LocationF3 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with laboratory spectra.

Graphic Jump LocationF4 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with the resampled laboratory spectra.

Graphic Jump LocationF5 :

Correlation between measured and PLS estimated soil properties for phase-1 calibration with Hyperion image spectra.

Graphic Jump LocationF6 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with laboratory spectra.

Graphic Jump LocationF7 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with the resampled laboratory spectra.

Graphic Jump LocationF8 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with Hyperion image spectra.

Graphic Jump LocationF9 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with laboratory absorbance spectra.

Graphic Jump LocationF10 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with the resampled laboratory absorbance spectra.

Graphic Jump LocationF11 :

Correlation between measured and PLS estimated soil properties for phase-2 calibration and validation with Hyperion absorbance spectra.

Tables

Table Grahic Jump Location
Table 1Minimum, maximum, mean, and standard deviation of six soil parameters before and after outlier removal.
Table Grahic Jump Location
Table 2Calibration results with laboratory spectra and Hyperion spectra of all usable samples.
Table Grahic Jump Location
Table 3Calibration and validation datasets: the number of samples, chemical range, mean, and standard deviation for soil properties considered in this study.
Table Grahic Jump Location
Table 4Statistical parameters resulting from the PLS calibration and validation with laboratory and Hyperion reflectance spectra.
Table Grahic Jump Location
Table 5Statistical parameters resulting from the PLS calibration and validation with laboratory and Hyperion absorbance spectra.

References

Mondini  C., Sequi  P., “Implication of soil C sequestration on sustainable agriculture and environment,” Waste Manage.. 28, (4 ), 678 –684 (2008), CrossRef. 0956-053X 
van Veen  J. A., Kuikman  P. J., “Soil structural aspects of decomposition of organic matter by micro-organisms,” Biogeochemistry. 11, (3 ), 213 –233 (1990), CrossRef. 0168-2563 
Doran  J. W., Sarrantino  M., Liebig  M. A., “Soil health and sustainability,” Adv. Agron.. 56, , 1 –54 (1996), CrossRef. 0065-2113 
Morishita  T. et al., “CO2, CH4 and N2O fluxes from a larch forest soil in central Siberia,” in Symptom of Environmental Change in Siberian Permfrost Region. , Hatano  R. , Guggenberger  G., Eds., pp. 1 –9,  Hokkaido University Press ,  Sapporo  (2006).
Hickman  J. E. et al., “Current and future nitrous oxide emission from African agriculture,” Curr. Opin. Environ. Sustainabil.. 3, (5 ), 370 –378 (2011), CrossRef. 1877-3435 
Smith  K. A. et al., “Exchange of greenhouse gases between soil and atmosphere: interactions of soil physical factors and biological processes,” Eur. J. Soil Sci.. 54, (4 ), 779 –791 (2003), CrossRef. 1351-0754 
Qafoku  O. S. et al., “Rapid methods to determine potentially mineralizable nitrogen in broiler litter,” J. Environ. Qual.. 30, (1 ), 217 –221 (2001), CrossRef. 0047-2425 
Bogrekci  I., Lee  W. S., “Spectral phosphorus mapping using diffuse reflectance of soils and grass,” Biosyst. Eng.. 91, (3 ), 305 –312 (2005), CrossRef. 1537-5110 
Ben-Dor  E., Banin  A., “Near-infrared reflectance analysis of carbonate concentration in soils,” Appl. Spectrosc.. 44, (6 ), 1064 –1069 (1990), CrossRef. 0003-7028 
Ben-Dor  E., Banin  A., “Visible and near-infrared (0.4–1.1 μm) analysis of arid and semiarid soils,” Remote Sens. Environ.. 48, (3 ), 261 –274 (1994), CrossRef. 0034-4257 
Palacios-Orueta  A., Ustin  S. L., “Remote sensing of soil properties in the Santa Monica mountains I. spectral analysis,” Remote Sens. Environ.. 65, (2 ), 170 –183 (1998), CrossRef. 0034-4257 
Ingleby  H. R., Crowe  T. G., “Reflectance models for predicting organic carbon in Saskatchewan soils,” Can. Agric. Eng.. 42, (2 ), 57 –63 (2000). 0045-432X 
Thomasson  J. A. et al., “Soil reflectance sensing for determining soil properties in precision agriculture,” Trans. ASAE.. 44, (6 ), 1445 –1453 (2001). 0001-2351 
Asner  G. P., Heidebrecht  K. B., “Imaging spectroscopy for desertification studies: comparing AVIRIS and EO-1 hyperion in Argentina drylands,” IEEE Trans. Geosci. Remote Sens.. 41, (6 ), 1283 –1296 (2003), CrossRef. 0196-2892 
Schwanghart  W., Jarmer  T., “Linking spatial patterns of soil organic carbon to topography—a case study from south eastern Spain,” Geomorphology. 126, (1–2 ), 252 –263 (2011), CrossRef. 0169-555X 
Ben-Dor  E. et al., “Mapping of several soil properties using DAIS-7915 hyperspectral scanner data—a case study over clayey soils in Israel,” Int. J. Remote Sens.. 23, (6 ), 1043 –1062 (2002), CrossRef. 0143-1161 
Stevens  A. et al., “Detection of carbon stock change in agricultural soils using spectroscopic techniques,” Soil Sci. Soc. Am. J.. 70, (3 ), 844 –850 (2006), CrossRef. 1435-0661 
Stevens  A. et al., “Laboratory, field and airborne spectroscopy for monitoring organic carbon content in agricultural soils,” Geoderma. 144, (1–2 ), 395 –404 (2008), CrossRef. 0016-7061 
Lagacherie  P. et al., “Estimation of soil clay and calcium carbonate using laboratory, field and airborne hyperspectral measurements,” Remote Sens. Environ.. 112, (3 ), 825 –835 (2008), CrossRef. 0034-4257 
Hively  W. D. et al., “Use of airborne hyperspectral imagery to map soil properties in tilled agricultural fields,” Appl. Environ. Soil Sci.. 2011, , 1 –13 (2011), CrossRef. 1687-7675 
Gomez  C., Viscarra Rossel  R. A., McBratney  A. B., “Soil organic carbon prediction by hyperspectral remote sensing and field vis-NIR spectroscopy: an Australian case study,” Geoderma. 146, (3–4 ), 403 –411 (2008), CrossRef. 0016-7061 
Sudduth  K. A., Hummel  J. W., “Portable near-infrared spectrophotometer for rapid soil analysis,” Trans. ASAE. 36, (1 ), 185 –193 (1993). 0001-2351 
Hummel  J. W., Gaultney  L. D., Sudduth  K. A., “Soil property sensing for site-specific crop management,” Comput. Electron. Agr.. 14, (2–3 ), 121 –136 (1996), CrossRef. 0168-1699 
Scharf  P. C. et al., “Remote sensing for nitrogen management,” J. Soil Water Conserv.. 57, (6 ), 518 –524 (2002). 0022-4561 
Whiting  M. L., Li  L., Ustin  S. L., “Predicting water content using Gaussian model on soil spectra,” Remote Sens. Environ.. 89, (4 ), 535 –552 (2004), CrossRef. 0034-4257 
Viscarra Rossel  R. A. et al., “Visible, near infrared, mid infrared or combined diffuse reflectance spectroscopy for simultaneous assessment of various soil properties,” Geoderma. 131, (1–2 ), 59 –75 (2006), CrossRef. 0016-7061 
Brown  D. J. et al., “Global soil characterization with VNIR diffuse reflectance spectroscopy,” Geoderma. 132, (3–4 ), 273 –290 (2006), CrossRef. 0016-7061 
Leone  A. P. et al., “Prediction of soil properties with VIS-NIR-SWIR reflectance spectroscopy and artificial neural networks: a case study on three pedoenvironments of the Campania region, Italy,” in  5th International Congress of the European Society for Soil Conservation Soils changing in a changing World: The Soils of Tomorrow , Dazzi  C. , Costantini  E., Eds., pp. 685 –698,  Catena Verlag ,  Palermo, Italy  (2008).
Heiri1  O., Lotter  A. F., Lemcke  G., “Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results,” J. Paleolimnol.. 25, (1 ), 101 –110 (2001), CrossRef.
Pierzynski  G. M. , Ed., Methods of Phosphorus Analysis for Soils Sediments, Residuals and Waters. , pp. 39 –49,  North Carolina State University ,  Raleigh, North Carolina  (2004).
Gee  G. W., Bauder  J. W., “Particle size analysis,” Methods of soil analysis, Part 1: Physical and mineralogical methods. , 2nd ed., Klute  A. , Ed., pp. 404 –408,  American Society of Agronomy ,  Madison, Wisconsin  (1986).
“Analytical Imaging and Geophysics LLC,” ACORN User’s Guide. , Stand Alone Version (2001).
Wold  H., “Nonlinear estimation by eterative least squares procedure,” Research Papers in Statistics. , David  F. N. , Neyman  J., Eds., pp. 441 –444,  Wiley ,  New York  (1966a).
Wold  H., “Estimation of principal components and related models by iterative lease squares,” in Multivariate Analysis. , Krishnaiah  P. R., Ed., pp. 391 –420,  Elsevier ,  New York  (1966b).
Geladi  P., Kowalski  B. R., “Partial least-squares regression: a tutorial,” Anal. Chim. Acta. 185, (1 ), 1 –17 (1986), CrossRef. 0003-2670 
Haaland  D. M., Thomas  E. V., “Partial least-squares methods for spectral analyses.1. Relation to other quantitative calibration methods and the extraction of qualitative information,” Anal. Chem.. 60, (11 ), 1193 –1202 (1988), CrossRef. 0003-2700 
Martens  H., Næs  T., Multivariate Calibration. , pp. 73 –296 John Wiley and Sons ,  New Jersey  (2002).
Azzouza  T. et al., “Comparison between different data pre-treatment methods in the analysis of forage samples using near-infrared diffuse reflectance spectroscopy and partial least-squares multivariate calibration method,” Anal. Chim. Acta. 484, (1 ), 121 –134 (2003), CrossRef. 0003-2670 
Gauch  H. G., Hwang  J. T. G., Fick  G. W., “Model evaluation by comparison of model-based predictions and measured values,” Agron. J.. 95, (6 ), 1442 –1446 (2003), CrossRef. 0002-1962 
Nduwamungu  C. et al., “Opportunities for, and limitations of, near infrared reflectance spectroscopy applications in soil analysis: a review,” Can. J. Soil Sci.. 89, (5 ), 531 –541 (2009), CrossRef. 0008-4271 
Willams  P. C., Sobering  D. C., “Comparison of commercial near infrared transmittance and reflectance instuments for analysis of whole grains and seeds,” J. Near Infrared Spectrosc.. 1, (1 ), 25 –32 (1993), CrossRef. 0967-0335 
Brown  D. J., Bricklemyer  R. S., Miller  P. R., “Validation requirements for diffuse reflectance soil characterization models with a case study of VNIR soil C prediction in Montana,” Geoderma. 129, (3–4 ), 251 –267 (2005), CrossRef. 0016-7061 
Du  C. et al., “Determination of soil properties using Fourier transform mid-infrared photoacoustic spectroscopy,” Vibrat. Spectrosc.. 49, (1 ), 32 –37 (2009), CrossRef. 0924-2031 
Chang  C. W. et al., “Near infrared reflectance spectroscopy-principal components regression analyses of soil properties,” Soil Sci. Soc. Am. J.. 65, (2 ), 480 –490 (2001), CrossRef. 1435-0661 
Dalal  R. C., Henry  R. J., “Simultaneous determination of moisture, organic carbon and total nitrogen by near infrared reflectance spectrophotometry,” Soil Sci. Soc. Am. J.. 50, (1 ), 120 –123 (1986), CrossRef. 1435-0661 
Hummel  J. W., Sudduth  K. A., Hollinger  S. E., “Soil moisture and organic matter prediction of surface and subsurface soils using an NIR soil sensor,” Comput. Electron. Agr.. 32, (2 ), 149 –165 (2001), CrossRef. 0168-1699 
Ben-Dor  E. et al., “Using imaging spectroscopy to study soil properties,” Remote Sens. Environ.. 113, (Suppl. 1 ), S38 –S55 (2009), CrossRef. 0034-4257 
Liu  W., Baret  F. et al., “Relating soil surface moisture to reflectance,” Remote Sens. Environ.. 81, (2–3 ), 238 –246 (2002), CrossRef. 0034-4257 
Bogrekci  I., Lee  W. S., “Effects of soil moisture content on absorbance spectra of sandy soils in sensing phosphorus concentrations using UV-VIS NIR spectroscopy,” Trans. ASABE. 49, (4 ), 1175 –1180 (2006).
Wetterlind  J., Stenberg  B., Söderström  M., “Increased sample point density in farm soil mapping by local calibration of visible and near infrared prediction models,” Geoderma. 156, (3–4 ), 152 –160 (2010), CrossRef. 0016-7061 
Selige  T., Böhner  J., Schmidhalter  U., “High resolution topsoil mapping using hyperspectral image and field data in multivariate regression modeling procedures,” Geoderma. 136, (1–2 ), 235 –244 (2006), CrossRef. 0016-7061 
Gomez  C., Lagacherie  P., Coulouma  G., “Continuum removal versus PLSR method for clay and calcium carbonate content estimation from laboratory and airborne hyperspectral measurements,” Geoderma. 148, (2 ), 141 –148 (2008), CrossRef. 0016-7061 
Staenz  K. et al., “Retrieval of surface reflectance from Hyperion radiance data,” in  Proc. IEEE Int. Geosci. and Remote Sens. Symposium , Vol. 3, pp. 1419 –1421,  IEEE ,  Ontario, Canada  (2002).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

Related Book Chapters

Topic Collections

Advertisement
  • Don't have an account?
  • Subscribe to the SPIE Digital Library
  • Create a FREE account to sign up for Digital Library content alerts and gain access to institutional subscriptions remotely.
Access This Article
Sign in or Create a personal account to Buy this article ($20 for members, $25 for non-members).
Access This Proceeding
Sign in or Create a personal account to Buy this article ($15 for members, $18 for non-members).
Access This Chapter

Access to SPIE eBooks is limited to subscribing institutions and is not available as part of a personal subscription. Print or electronic versions of individual SPIE books may be purchased via SPIE.org.