Journal of Environmental Quality 31:1248-1255 (2002)
© 2002 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
TECHNICAL REPORTS
Plant and Environment Interactions
Site-Specific Phosphorus Application Based on the Kriging Fertilizer-Phosphorus Availability Index of Soils
Kai-Wei Juanga,
Day-Chyng Lioub and
Dar-Yuan Lee*,b
a Dep. of Post-Modern Agriculture, Ming Dao Univ., Peitou, Changhua, Taiwan, R.O.C
b Graduate Inst. of Agricultural Chemistry, National Taiwan Univ., Taipei, Taiwan, R.O.C
* Corresponding author (dylee{at}ccms.ntu.edu.tw)
Received for publication July 3, 2001.
 |
ABSTRACT
|
|---|
Site-specific phosphorus management is done to optimize crop production and minimize P loss from soils. The spatial variability of the available P prior to fertilizer application and the P-fixation tendency of soil both need to be taken into account for variable-rate P application. The objectives of this research were to document the spatial variability of the fertilizer-P availability index, which shows the P-fixation tendency, and to develop a strategy that takes the spatial distribution of this index into account for site-specific phosphorus application. In this study, the spatial patterns of the fertilizer-P availability index were characterized by using geostatistics. The ordinary kriging was used for spatial interpolation of the fertilizer-P availability index. Because the fertilizer-P availability index of soil is related to oxalate-extractable Fe and Al and because measuring oxalate-extractable Fe and Al is much easier than directly determining the fertilizer-P availability index, the spatial distribution of the fertilizer-P availability index can be obtained using the oxalate-extractable Fe and Al data. The spatial distribution of Olsen-extractable P, which was used to measure the available-P status prior to fertilizer-P application, was also estimated by using ordinary kriging. The required fertilizer-P amounts were then determined using the kriging estimates of the fertilizer-P availability index and Olsen-extractable P. A fertilizer-P recommendation map for the 430-ha study site in Changhua county, Taiwan was generated by using this approach for illustration. The proposed method for generating fertilizer-P recommendation maps can be used for variable-rate application to maintain an adequate P status for crop production and to potentially reduce the P loss from soils.
|
INTRODUCTION
|
|---|
CURRENTLY, the main purpose in managing phosphorus is to optimize crop production and minimize P loss from soils (Beegle et al., 2000; Frossard et al., 2000; Higgs et al., 2000; Valk et al., 2000). If the available-P status after fertilizer application is very high, then P loss from soils can be expected. An ideal application strategy is to apply sufficient P to soils for crop production without overapplication, which can negatively affect the environment. Chemical-extractable P is usually used as an availability index for fertilizer-P recommendations. Chemical extractants, such as Bray-1 (NH4F + HCl), Mehlich-1 (HCl + H2SO4), Mehlich-3 (NH4F + ethylene-diaminetetraacetic acid [EDTA]), and Olsen (NaHCO3), are regularly used in soil fertility testing (Mehlich, 1984; Wolf and Baker, 1985; Sharpley et al., 1987). However, fertilizer P added into soils will be fixed by sorption or precipitation reactions; examples are P sorbed on Fe and Al oxide surfaces and P precipitated with Fe and Al oxides (Barrow, 1983; Van Riemsdijk et al., 1984). Thus, the available-P status after fertilizer application is also dependent upon the P-fixation tendency of soil. An alternative approach to soil testing that considers the recovery rate of P application to improve fertilizer recommendations has been proposed (McLean et al., 1979). The chemical-extractable P and the recovery rate of P application were simultaneously taken into account for fertilizer-P recommendations (McLean et al., 1982). The recovery rate of P application, also called the fertilizer-P availability index (Sharpley et al., 1984; Jones et al., 1984), is a fraction of the added P recovered in the chemical extraction. Two factors determining the available-P status after fertilizer application are the level of extractable P prior to application and the fertilizer-P availability index. The fertilizer-P availability index corresponds to the P-fixation tendency of soil. There is a great value of the fertilizer-P availability index when the P-fixation tendency is weak, and vice versa. Thus, to reach a sufficient P level for crop production and to avoid a negative environmental impact, one can use the extractable P prior to application and the fertilizer-P availability index of soil to make fertilizer-P recommendations.
Uniform fertilizer application is a traditional practice for maintaining optimal crop production. However, because of the spatial variation of soil fertility, excess fertilizer application usually happens in some areas within a field. In order to match fertilizer application with crop needs, variable-rate application was developed. Ndiaye and Yost (1989) emphasized that the spatial variability of nutrients in soils should be considered for variable-rate application. Mulla et al. (1992) pointed out that variable-rate application reduces the overfertilization that commonly occurs when uniform-rate application is used on a farm along with spatial variation. Cahn et al. (1994) showed the importance of spatial variation of soil fertility for site-specific crop management. Haneklaus et al. (1998) also suggested that correctly mapping soil fertility parameters is important for variable-rate application. Therefore, spatial information of nutrient status should be characterized when making fertilizer recommendations. Geostatistical methods are frequently used to map soil properties. The spatial distribution of the P status in soils has been estimated using geostatistics for variable-rate applications (Mulla et al., 1992; Cahn et al., 1994). For managing the fertilizer-P application, a spatial distribution of soil P status is thus essential for variable-rate application. Wollenhaupt et al. (1994) compared some spatial interpolation methods for mapping soil phosphorus. Schepers et al. (2000) documented the spatial patterns of phosphorous in soils for developing management strategies for variable-rate application. Thus, to develop an improved management strategy for variable-rate fertilizer-P application, the spatial variation of the P-fixation tendency also should be taken into account. One can use the spatial distribution of the fertilizer-P availability index to make the recommendation for variable-rate application of fertilizer P.
However, determining the fertilizer-P availability index of soils (McLean et al., 1982) is more cumbersome than is determining the chemical-extractable P. The fertilizer-P availability index is obtained by estimating the slope of the linear model between chemical-extractable P and added P (Sharpley et al., 1984; Jones et al., 1984). Measuring the fertilizer-P availability index is a laborious task. It is difficult to obtain enough data for investigating the spatial variation and estimating the spatial distribution of the fertilizer-P availability index using geostatistics. Thus, the spatial patterns of the fertilizer-P availability index have never been used to develop a fertilizer-P recommendation map. To our knowledge, the fertilizer-P availability index is closely related to the oxalate-extractable Fe and Al, which dominate the P-fixation tendency of soils (Sanyal and De Datta, 1991; Lookman et al., 1995; Schoumans and Groenendijk, 2000). The correlation between the fertilizer-P availability index and oxalate-extractable Fe and Al of soils is expected. The fertilizer-P availability index thus can be modeled by a function of oxalate-extractable Fe and Al. If the correlation between the fertilizer-P availability index and oxalate-extractable Fe and Al is identified, then one can use oxalate-extractable Fe and Al data to predict the values of the fertilizer-P availability index for investigating the spatial patterns of the fertilizer-P availability index.
The objectives of this research are to document the spatial variability of the fertilizer-P availability index of soils and to develop a strategy that takes the spatial distribution of this index into account for site-specific phosphorus application. In this study, the spatial variation of the fertilizer-P availability index of soils was investigated. The correlation between the fertilizer-P availability index and oxalate-extractable Fe and Al was also identified for modeling the spatial variation of the fertilizer-P availability index. Ordinary kriging was used in spatial interpolation of the fertilizer-P availability index and Olsen-extractable P. Then, the spatial distributions of both the fertilizer-P availability index and Olsen-extractable P prior to fertilizer application were simultaneously used to make a recommendation map for variable-rate P application.
|
MATERIALS AND METHODS
|
|---|
Sampling and Chemical Analyses
The study site was a field about 430 ha in area situated in Changhua county, Taiwan. Rice (Oryza sativa L.) is the main crop in this field. Chien et al. (1997) investigated the spatial variation of the soil properties of this site. This area is on the alluvial plain of the Tadu River. The soils are characterized by sandstone, shale, and slate alluvial soils. Soils (0–15 cm) were taken at 62 sampling locations. The sampling scheme is shown in Fig. 1
. The smallest sampling interval is 250 m. All soil samples were air-dried and passed through a 2-mm sieve for chemical analyses.
View larger version (22K):
[in this window]
[in a new window]
|
Fig. 1. Study site and the sampling points. The coordinate system is 2° UTM (Universal Transverse Mercator for Taiwan).
|
|
To measure the fertilizer-P availability index of soil, enough K2HPO4 solution was added and thoroughly mixed to each soil sample to provide 0, 15, 30, 60, and 120 mg P/kg soil. Each soil sample was moistened to field capacity. Then, the soil sample was incubated at room temperature (25°C) for 10 d. The available P in the soil after incubation was measured using the Olsen extraction method (Olsen et al., 1954). Olsen extraction is a soil test recommended for analyzing the P availability of dryland and paddy soils in Taiwan, especially for nonacid soils (Chang, 1981). The Olsen-extractable P concentration was determined using the molybdate–ascorbic acid method (Watanabe and Olsen, 1965). The Olsen-extractable P levels at various P amounts added to the soil for each soil sample were measured. The relationship between the P amounts added and the corresponding Olsen-extractable P level for all the tested soil samples was found to be linear and similar to what has been reported by previous researchers (Sharpley et al., 1984; Jones et al., 1984). The relationship was then modeled by simple linear regression. The fertilizer-P availability index (FP) was the slope of the linear model between Olsen-extractable P (POls) and added P (Padd) for each soil sample (Jones et al., 1984):
 | [1] |
where P0 is the level of Olsen-extractable P when no P is being added and represents the level of Olsen-extractable P of soils prior to fertilizer-P application. To characterize the relationship between the fertilizer-P availability index and oxalate-extractable Fe and Al, the oxalate-extractable Fe and Al (Feox and Alox) in each soil sample were also determined (McKeague and Day, 1966). Then, the data for FP, P0, Feox, and Alox of soils at 62 sampled locations were obtained.
Correlation and Regression Model
The correlation matrix was used to investigate the linear relationship among FP, Feox, and Alox. The scatter plots of FP vs. Feox and FP vs. Alox were used to visualize the correlation. Then, a regression model was used to predict the value of FP with Feox and Alox. The model is set by:
 | [2] |
where a, b, and c are parameters. The regression data analysis tool in Microsoft Excel (Microsoft, 2000) was used to obtain the model parameters. For validating the regression model, a subset with 20 data points randomly drawn from the original data set with 62 points was used to establish the regression model. The remaining 42 data points were used for validation. The coefficient of determination (r2) was used for assessing the reliability of the regression model:
 | [3] |
where FPi and FPi* are respectively the observed and predicted values of the fertilizer-P availability index at the 42 points and 
P is the mean of FPi.
Geostatistical Analyses
Geostatistics has been used frequently for spatial interpolation of soil properties. The theory of geostatistics can be found in textbooks (Journel and Huijbregts, 1978; Isaaks and Srivastava, 1989; Goovaerts, 1997). In this study, geostatistical analyses, including spatial variation modeling (variogram) and spatial interpolation (ordinary kriging) were conducted to characterize the spatial distributions of the fertilizer-P availability index and Olsen-extractable P.
Variogram
Based on the assumption of intrinsic stationarity for spatial data, the spatial variation structure of a data set that is only dependent on distance (h) between two locations where the variable values are z(xi + h) and z(xi) can be shown by the following semivariogram:
 | [4] |
where 
(h) is the experimental semivariance and N(h) is the pair number of z(xi + h) and z(xi). For kriging estimation, a variogram model should be fitted to such an experimental semivariogram. In this study, the spherical model was used:
 | [5] |
where Co is called the nugget effect, Co + C is the sill value, and a is the range. The software package GS+ (Gamma Design, 1994) was used to construct experimental semivariograms and variogram models for estimating the spatial distributions of the fertilizer-P availability index and Olsen-extractable P.
Ordinary Kriging
The ordinary kriging estimator zOK*(xo) at an unsampled location xo is defined by:
 | [6] |
where m is the number of surrounding observations z(xi) selected for estimation and 
i is the weight of z(xi). The weights should sum to unity to make zOK*(xo) unbiased:
 | [7] |
By minimizing the kriging variance 
2OK to determine the weights, the kriging estimate can be found with Eq. [6]. The values of 2OK also can be obtained by:
 | [8] |
where µ is the Lagrange multiplier. In this study, the kriging estimation of the fertilizer-P availability index (FP) and Olsen-extractable P (P0) prior to P application was performed using the software package GEO-EAS (Englund and Sparks, 1988).
Cross-Validation
To evaluate the reliability of kriging estimation, cross-validation was used, with the mean error (ME) and mean square relative error (MSRE) of kriging-estimated values being calculated (Voltz and Webster, 1990). The ME is a measure of the estimation bias, and it should be close to zero for unbiased methods. It is defined by:
 | [9] |
where n is the number of data points, z(xi) is the observed value, and zOK*(xi) is the kriging estimate. The index MSRE is a measure of consistency for kriging estimation. It is defined by:
 | [10] |
where sOK(xi) is the kriging standard deviation corresponding to zOK*(xi). A reliable estimation should be with a MSRE value close to 1.
Recommendation Mapping
Based on the measurement of the fertilizer-P availability index by Eq. [1], the value FP of the fertilizer-P availability index and the level of Olsen-extractable P prior to P application (P0) were used to determine the amount (PF) of fertilizer P required for a given sufficient level (PSL) by:
 | [11] |
The sufficient available-P status of Taiwanese soils for rice production is between 23 and 26 mg P/kg soil of Olsen-extractable P (Yang, 1991). Because rice is the main crop in the study field, the sufficient level PSL was set to be 24.5 mg P/kg soil. The kriging estimated values of FP and P0, showing the spatial distributions of the fertilizer-P availability and Olsen-extractable P prior to P application, were put into Eq. [11] to obtain the values of PF for mapping the requirement of fertilizer P. The whole procedure of this study is schematically shown in Fig. 2
.
|
RESULTS AND DISCUSSION
|
|---|
Descriptive Statistics
The descriptive statistics of the fertilizer-P availability index (FP), the level of Olsen-extractable P (P0) prior to P application, and oxalate-extractable Fe and Al (Feox and Alox) for the soil samples are shown in Table 1
. The mean value of FP is 0.49. This means that almost 50% of fertilizer P added to the soils will be fixed. The coefficient of variation (CV) of FP is not large, and the data distribution is also slightly skewed. However, the deviation between minimum and maximum is near 0.6. The spatial variation of the P-fixation tendency of the soils cannot be ignored. This should be taken into account when making fertilizer-P recommendations. The mean P0 value is about 30 mg/kg. This indicates that the average P status prior to application is higher than the given sufficient range, 23 to 26 mg P/kg soil of Olsen-extractable P. However, the P0 data distribution shows greater variation than for FP. The minimum for P0 is less than 2 mg/kg and the maximum is almost 80 mg/kg. Compared with FP, the P0 skewness is relatively high. The great variation and high skewness of P0 reveals that the P status prior to fertilizer-P application was heterogeneous. Thus, the variable-rate application is needed in this field.
View this table:
[in this window]
[in a new window]
|
Table 1. Descriptive statistics of the fertilizer-P availability index (FP), Olsen-extractable P (P0) prior to fertilizer-P application, and oxalate-extractable Fe and Al (Feox and Alox) of the investigated soils.
|
|
Relationship of Fertilizer-Phosphorus Availability Index and Oxalate-Extractable Iron and Aluminum
Table 2
shows the correlation matrix of FP, P0, Feox, and Alox. The correlation coefficients of FP vs. Feox and FP vs. Alox are -0.63 and -0.61, respectively. The fertilizer-P availability index is thus closely related to the oxalate-extractable Fe and Al in soils. The scatter plots shown in Fig. 3
also confirm the linear relationship among FP, Feox, and Alox. A linear-formed model as Eq. [2] thus can be established to predict FP with the Feox and Alox data. In order to determine the model parameters (a, b, and c), the 20 randomly selected data points of FP, Feox, and Alox were used. The parameters (a, b, and c) and the correlation coefficient (r) of the model shown in Table 3
were obtained by regression. For validating the model, the scatter plot of predicted and observed values of FP at the other 42 data points was also shown in Fig. 4
. Most of the values in the scatter plot are close to the 1:1 line. The coefficient of determination (r2) between the observed values and the predicted values of FP based on the model for the 42 data points is found to be 0.51, which is significant at P < 0.05. This indicates that the regression model is reliable and thus the values of the fertilizer-P availability index can be predicted using the data of oxalate-extractable Fe and Al based on the regression model. Therefore, we tried to investigate the spatial patterns of the fertilizer-P availability index using the 42 predicted values of FP. A composite set of FP values, including the 20 observed and 42 predicted values, was then generated for spatial interpolation.
View this table:
[in this window]
[in a new window]
|
Table 2. Correlation coefficients for the fertilizer-P availability index (FP), Olsen-extractable P (P0) prior to fertilizer-P application, and Oxalate-extractable Fe (Feox) and Al (Alox).
|
|
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 3. Scatter plots of (a) oxalate-extractable Fe (Feox) vs. the fertilizer-P availability index (FP) and (b) oxalate-extractable Al (Alox) vs. the fertilizer-P availability index (FP).
|
|
View this table:
[in this window]
[in a new window]
|
Table 3. Paramaeters and the correlation coefficient (r) for the regression model of the fertilizer-P availability index (FP) versus the Oxalate-extractable Fe (Feox) and Al (Alox).
|
|
Spatial Interpolation of Fertilizer-Phosphorus Availability Index
For spatial interpolation of the fertilizer-P availability index, two data sets of FP were used: original (with 62 observed values) and composite (with 20 observed and 42 predicted values). The semivariograms of FP obtained using the original and composite data sets are shown in Fig. 5a and 5b
, respectively. These spatial structures show moderate spatial dependence. Chien et al. (1997) documented that the ratio of nugget to sill can be regarded as an index to measure the spatial dependence. If this ratio is between 0.25 and 0.75, the variable has moderate spatial dependence. However, it should be noted that the nugget effect forming a large proportion of the total variance (sill) as found in the semivariogram of FP may cause uncertainty in kriging estimation. Moreover, the variogram models in Fig. 5 are very similar. This indicates that the predicted values of FP show the same type of spatial structure as the observed values of FP. The results of cross-validation for the kriging estimates of FP using the original data set and the composite data set are shown in Table 4
. This reveals that the kriging estimation is reliable for the two data sets. The spatial distributions of the fertilizer-P availability index estimated by using ordinary kriging with the original data set and the composite data set respectively are shown in Fig. 6a and 6b
. In both figures, the values of the fertilizer-P availability index located to the right and left of the middle of the field are low. The values of the fertilizer-P availability index located above and below the middle of the field are high. The spatial patterns of Fig. 6 are very similar, indicating that the composite data set with 42 predicted values of FP could be used to estimate a reliable spatial distribution of the fertilizer-P availability index as the original data set did. Thus, based on the correlation between the fertilizer-P availability index and the oxalate-extractable Fe and Al, one can use the data of oxalate-extractable Fe and Al to predict the values of the fertilizer-P availability index for spatial interpolation. Since measuring the oxalate-extractable Fe and Al is much easier than directly determining the fertilizer-P availability index of soils, the spatial distribution of the fertilizer-P availability index can be easily obtained by using the data of oxalate-extractable Fe and Al.
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 5. Experimental semivariograms and variogram models of the fertilizer-P availability index (FP) obtained using (a) the original data set with 62 observed values and (b) the composite data set with 20 observed and 42 predicted values.
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. The mean error (ME) and mean square relative error (MSRE) obtained from cross-validation for the kriging estimation using the original data set [FP(o)] and the composite data set [FP(c)] of fertilizer-P availability index and using the data of Olsen-extractable P (P0) prior to fertilizer-P application.
|
|
View larger version (26K):
[in this window]
[in a new window]
|
Fig. 6. Spatial distributions of the fertilizer-P availability index estimated by using ordinary kriging with (a) the original data set with 62 observed values and (b) the composite data set with 20 observed and 42 predicted values.
|
|
Spatial Interpolation of Olsen-Extractable Phosphorus prior to Fertilizer-Phosphorus Application
The semivariogram of P0 was also developed for the spatial interpolation of Olsen-extractable P prior to fertilizer-P application (Fig. 7) . The experimental semivariogram shows a great nugget effect and a short range. Compared with the variogram models of FP (Fig. 5), the inference range of the variogram model of P0 is much smaller. Olsen-extractable P, which is used to measure the available-P status, thus can be regarded as an extrinsic soil property. Nkedi-Kizza et al. (1994) showed that the available-P status of soils usually depends on tillage and fertilizer application. On the other hand, the fertilizer-P availability index is an intrinsic soil property because it is the result of soil genesis. Chien et al. (1997) pointed out that the inference range of the variogram model for an extrinsic soil property is usually shorter than that of the variogram model for an intrinsic soil property. Using a variogram model with great nugget variance (Fig. 7) for kriging estimation may cause uncertainty. In Table 4, the ME and the MSRE values indicate that the kriging estimation of P0 is reliable. The estimated values of P0 were used to map the spatial distribution of Olsen-extractable P prior to fertilizer-P application, shown in Fig. 8
. The greatest estimated value of P0 is higher than 60 mg/kg and appears in the lower right- and left-hand corners of the field. On the other hand, the smallest estimated value is lower than 10 mg/kg and appears in the upper right-hand corner of the field. The spatial distribution of available-P status prior to fertilizer-P application in this field clearly shows large variability. Thus, variable-rate P application is needed in this field to maintain a sufficiently high level of available P for crop production and to minimize the loss of P due to overapplication.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 7. Experimental semivariogram and variogram model of the Olsen-extractable P (P0) prior to fertilizer-P application.
|
|
Fertilizer-Phosphorus Recommendations
We made fertilizer-P recommendations based on Eq. [11]. The kriging estimates of FP and P0, respectively shown in Fig. 6a and Fig. 8, were simultaneously used to determine the required fertilizer P amount. The mass of soil with a bulk density of 1500 kg/m3 in 1 ha to a depth of normal plowing (15 cm) weighed about 2.25 million kg. The required amounts of fertilizer P (P2O5) per hectare of soil were calculated. The P recommendation map is shown in Fig. 9
. The upper left- and right-hand corners of the field require a very high level of fertilizer-P application where both the levels of FP and P0 are relatively low (Fig. 6a, 8). The results suggested that when the Olsen-extractable P level prior to fertilizer-P application is low and the P-fixation tendency of soil is strong, a high level of fertilizer-P application will be recommended. It seems that in this study field, the variability of P recommendation rates as shown in Fig. 9 is dominated by the spatial distribution of P0 because of the greater spatial variability of P0 than that of FP. However, it is expected that in the study site, the effect of the spatial variability of the fertilizer-P availability index on the fertilizer-P recommendation map would become more evident when P0 becomes less variable after several years of site-specific fertilizer application. Therefore, the proposed method for P recommendation mapping is potentially useful for maintaining a sufficient P status for crops and reducing the loss of P from soils.
View larger version (37K):
[in this window]
[in a new window]
|
Fig. 9. Phosphorus recommendation maps obtained using the spatial distributions of the fertilizer-P availability index and Olsen-extractable P prior to fertilizer-P application.
|
|
|
CONCLUSIONS
|
|---|
In this study, we found that the fertilizer-P availability index of soil is closely related to the oxalate-extractable Fe and Al in soils. The measurement of oxalate-extractable Fe and Al is much easier than directly determining the fertilizer-P availability index. Thus, one can use the data of oxalate-extractable Fe and Al to predict the values of the fertilizer-P availability index for spatial interpolation. The spatial distribution of the fertilizer-P availability index can be estimated by using kriging. The spatial distributions of the fertilizer-P availability index and Olsen-extractable P prior to fertilizer-P application are both taken into account for fertilizer-P recommendation mapping. The proposed method for generating a fertilizer-P recommendation map is essential for variable-rate P application. It is a useful approach for maintaining a sufficient P status for crops and potentially reducing P loss from soils.
|
ACKNOWLEDGMENTS
|
|---|
This research was sponsored by the National Science Council, Republic of China, under Project no. NSC-88-2313-B-002-019 and NSC-89-2313-B-002-143.
|
REFERENCES
|
|---|
- Barrow, N.J. 1983. On the reversibility of phosphate sorption by soils. J. Soil Sci. 34:751–758.
- Beegle, D.B., O.T. Carton, and J.S. Bailey. 2000. Nutrient management planning: Justification, theory, practice. J. Environ. Qual. 29:72–79.[Abstract/Free Full Text]
- Cahn, M.D., J.W. Hummel, and B.H. Brouer. 1994. Spatial analysis of soil fertility for site-specific crop management. Soil Sci. Soc. Am. J. 58:1240–1248.[Abstract/Free Full Text]
- Chang, I.H. 1981. Methods of soil tests currently used in Taiwan. p. 9–26. In Diagnostic techniques for crop fertilizer requirements. Taiwan Agric. Res. Inst., Taichung.
- Chien, Y.J., D.Y. Lee, H.Y. Guo, and K.H. Houng. 1997. Geostatistical analysis of soil properties of mid-west Taiwan soils. Soil Sci. 162:291–298.
- Englund, E., and A. Sparks. 1988. GEO-EAS, geostatistical environmental assessment software, user guide. USEPA Rep. 600/4-88/033. USEPA, Las Vegas, NV.
- Frossard, E., L.M. Condron, A. Oberson, S. Sinaj, and J.C. Fardeau. 2000. Processes governing phosphorus availability in temperate soils. J. Environ. Qual. 29:15–23.
- Gamma Design. 1994. Geostatistics for the environmental sciences. Version 2.3. Gamma Design, Plainwell, MI.
- Goovaerts, P. 1997. Geostatistics for natural resources evaluation. Oxford Univ. Press, New York.
- Haneklaus, S., H.M. Paulsen, D. Schrer, U. Leopold, and E. Schnug. 1998. Self-surveying: A strategy for efficient mapping of the spatial variability of time constant soil parameters. Commun. Soil Sci. Plant Anal. 29:1593–1601.
- Higgs, B., A.E. Johnston, J.L. Salter, and C.J. Dawson. 2000. Some aspects of achieving sustainable phosphorus use in agriculture. J. Environ. Qual. 29:80–87.[Abstract/Free Full Text]
- Isaaks, E.H., and R.M. Srivastava. 1989. An introduction to applied geostatistics. Oxford Univ. Press, New York.
- Jones, C.A., C.V. Cole, A.N. Sharpley, and J.R. Williams. 1984. A simplified soil and plant phosphorus model: III. Testing. Soil Sci. Soc. Am. J. 48:810–813.[Abstract/Free Full Text]
- Journel, A.G., and C.J. Huijbregts. 1978. Mining geostatistics. Academic Press, New York.
- Lookman, R., N. Vandeweert, R. Merckx, and K. Vlassak. 1995. Geostatistical assessment of the regional distribution of phosphate sorption capacity parameters (Feox and Alox) in northern Belgium. Geoderma 66:285–296.
- McKeague, J.A., and J.H. Day. 1966. Dithionite and oxalate extractable Fe and Al as aids in differentiating various classes of soils. Can. J. Soil Sci. 46:13–22.
- McLean, E.O., T.O. Oloya, and J.L. Adams. 1979. Soil tests to inventory the initially available levels and to assess the fates of added P and K as bases for improved fertilizer recommendations. Commun. Soil Sci. Plant Anal. 10:623–630.
- McLean, E.O., T.O. Oloya, and S. Mostaghimi. 1982. Improved corrective fertilizer recommendations based on a two-step alternative usage of soil tests: I. Recovery of soil-equilibrated phosphorus. Soil Sci. Soc. Am. J. 46:1193–1197.[Abstract/Free Full Text]
- Mehlich, A. 1984. Mehlich 3 soil test extractant: A modification of Mehlich 2 extractant. Commun. Soil Sci. Plant Anal. 15:1409–1416.
- Microsoft. 2000. Microsoft Office 2000, SR-1. Microsoft, Redmond, WA.
- Mulla, D.J., A.U. Bhatti, M.W. Hammond, and J.A. Benson. 1992. A comparison of winter wheat yield and quality under uniform versus spatially variable fertilizer management. Agric. Ecosyst. Environ. 38:301–311.
- Ndiaye, J.P., and R.S. Yost. 1989. Influence of fertilizer application nonuniformity on crop response. Soil Sci. Soc. Am. J. 53:1872–1878.[Abstract/Free Full Text]
- Nkedi-Kizza, P., L.A. Gaston, and H.M. Selim. 1994. Extrinsic spatial variability of selected macronutrients in a sandy soil. Geoderma 63:95–106.
- Olsen, S.R., C.V. Cole, F.S. Watanabe, and L.A. Dean. 1954. Estimation of available phosphorus in soils by extraction with sodium bicarbonate. Circ. 939. USDA, Washington, DC.
- Sanyal, S.K., and S.K. De Datta. 1991. Chemistry of phosphorus transformations in soil. Adv. Soil Sci. 16:1–120.
- Schepers, J.S., M.R. Schlemmer, and R.B. Ferguson. 2000. Site-specific considerations for managing phosphorus. J. Environ. Qual. 29:125–130.
- Schoumans, O.F., and P. Groenendijk. 2000. Modeling soil phosphorus levels and phosphorus leaching from agricultural land in the Netherlands. Soil Sci. Soc. Am. J. 29:111–116.
- Sharpley, A.N., C.A. Jones, C. Gray, and C.V. Cole. 1984. A simplified soil and plant phosphorus model. II. Prediction of labile, organic and sorbed phosphorus. Soil Sci. Soc. Am. J. 48:805–809.[Abstract/Free Full Text]
- Sharpley, A.N., H. Tiessen, and C.V. Cole. 1987. Soil phosphorous forms extracted by soil tests as a function of pedogenesis. Soil Sci. Soc. Am. J. 51:362–365.[Abstract/Free Full Text]
- Valk, H., J.A. Metcalf, and P.J.A. Withers. 2000. Prospects for minimizing phosphorus excretion in ruminants by dietary manipulation. J. Environ. Qual. 29:28–36.[Abstract/Free Full Text]
- Van Riemsdijk, W.H., L.J.M. Boumans, and F.A.M. de Hann. 1984. Phosphate sorption by soils: I. A diffusion precipitation model for the reaction of phosphate with metal oxides in soil. Soil Sci. Soc. Am. J. 48:537–541.[Abstract/Free Full Text]
- Voltz, M., and R. Webster. 1990. A comparison of kriging, cubic spines and classification for predicting soil properties from sample information. J. Soil Sci. 41:473–490.
- Watanabe, F.S., and S.R. Olsen. 1965. Test of an ascorbic acid method for determining phosphorus in water and NaHCO3 extracts from soil. Soil Sci. Soc. Am. Proc. 29:677–678.
- Wolf, A.M., and D.E. Baker. 1985. Comparison of soil test phosphorus by Olsen, Bray P1, Mehlich 1, and Mehlich 3 methods. Commun. Soil Sci. Plant Anal. 16:467–484.
- Wollenhaupt, N.C., R.P. Wolkowski, and M.K. Clayton. 1994. Mapping soil test phosphorus and potassium for variable-rate fertilizer application. J. Prod. Agric. 7:441–448.
- Yang, K.S. 1991. Study on building a fertilizer recommendation system. Ph.D. thesis. Natl. Taiwan Univ., Taipei.
TOP
ABSTRACT
INTRODUCTION
