SSSAJ Grow Your Career with SSSA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.
Soil Science Society of America Journal 64:1706-1713 (2000)
© 2000 Soil Science Society of America

DIVISION S-5-PEDOLOGY

A Functional Approach to Soil Characterization in Support of Precision Agriculture

B.J. Van Alphen and J.J. Stoorvogel

Lab. of Soil Science and Geology, Wageningen Univ. and Research Center, P.O. Box 37, 6700 AA Wageningen, The Netherlands

jeroen.vanalphen{at}aio.beng.wau.nl


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 
Managing soil variability is an integral aspect of precision agriculture (PA). Existing soil databases, however, are found to match few of the requirements for PA. The nature of these requirements and their implications for soil information need to be further explored. Ongoing developments towards a decision support system (DSS) for PA in the Netherlands have shed some light on this issue. Two soil related DSS-components are presented: (i) the construction of a soil database at the farm level and (ii) the delineation of soil functional units at the field level. Developed methods were tested in a case study for two arable fields located on Dutch marine clay soils. Basic soil data were collected in a 1:5000 soil survey and supplemented with secondary data derived through pedotransfer functions. Soil characterization focused on functional properties describing soil-specific characteristics in terms of water regimes and nutrient dynamics. Four properties were considered: (i) water stress, (ii) N-stress, (iii) N-leaching and (iv) residual N-content at harvest. These were quantified for individual soil profiles using a mechanistic–deterministic simulation model. Sensitivity to water stress was evaluated for a dry year (1989), other properties were quantified for a wet year (1987). Based on functional similarity, the soil profiles were grouped into functional classes using a fuzzy c-means classifier. Standard interpolation techniques and a boundary detection algorithm subsequently identified soil functional units in each field. Analysis of variance revealed that >65% of the spatial variation could thus be accounted for. This confirmed that (i) the proposed classification procedure was efficient and (ii) soil functional units are suitable entities to be used as management units for PA.

Abbreviations: CI, confusion index • DSS, decision support system • FCM, fuzzy c-means classification • PA, precision agriculture • PTF, pedotransfer function • TDR, time domain reflectometry


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 
THE TIGHTENING of economic and environmental constraints on agriculture has resulted in a call for more efficient management systems. Besides maximizing crop production, the input of fertilizers and biocides should be reduced to a minimum. Precision agriculture responds to this challenge by developing management strategies that incorporate field variability. Soil information is crucial here, as soils are a major source of variation.

Soil databases have been assembled in many countries to provide easy access to soil information. Some examples are the State Soil Geographic (STATSGO; National Resources Conservation Service, 1995a) and Soil Survey Geographic (SSURGO; National Resources Conservation Service, 1995b) databases in the USA and the National Soil Survey database of the Netherlands (Bregt et al., 1987). While readily available, the application of soil survey databases in PA has proven cumbersome. Stermitz et al. (1999) found SSURGO data to be of little value in explaining yield variations. In more general terms, the National Research Council (1997) concluded that "current soil surveys satisfy few of the data requirements for PA. Soil data are not at the appropriate level of detail, nor are the indexes required by PA the same as those provided by soil surveys." This was not surprising since soil survey data were never intended for use in PA. Their conclusion does, however, raise three important questions:

  1. Which requirements does PA pose on soil information?
  2. How can the desired information be produced?
  3. How can soil information be translated into recommendations for precision management?

An interdisciplinary research team in the Netherlands is currently seeking answers to these questions. Although research is still in progress, the outline of a DSS for PA is taking shape. The system, which is designed for arable farming and reflects Dutch conditions, was described in detail by Bouma et al. (2000).

The DSS is founded on a detailed soil database constructed specifically for PA. Bouma et al. (2000) state that a similar database will be required for most farms switching to precision management, as it provides the only means of reaching an adequate level of detail. The soil database contains both primary (e.g., texture, organic matter content) and secondary soil data (e.g., hydraulic parameters) for a large number of soil auger observations. Secondary data are derived through continuous pedotransfer functions (Wösten et al., 1998).

Sampled soil profiles are characterized in terms of their water regimes and nutrient dynamics under varying weather conditions. This is referred to as functional characterization, as opposed to traditional taxonomic characterization (e.g., Soil Taxonomy) (Soil Survey Staff, 1998). Soil functional properties are derived using a mechanistic–deterministic simulation model, which forms the core of the DSS. Based on functional similarity, the soil profiles are grouped into functional classes. This information is subsequently interpolated to identify soil functional units at the field level. These units serve as management units for PA (Van Uffelen et al., 1997).

Questions to be resolved by the DSS may include whether fertilizer or irrigation water should be applied and, if so, at which locations and at which quantities? A forward-looking approach is pursued, allowing a pro-active response to the near depletion of water and/or nutrients in (part of) a field. This requires dynamic and spatially differentiated estimates of the actual supply and demand for water and nutrients. Estimates are provided through real-time simulations for representative soil profiles located in each functional unit. The simulations quantify soil water fluxes, N-transformations and solute movement on a daily basis. In turn, these data are used to generate early warning signals for water and/or N-depletion in the root zone.

This paper describes the development of two important DSS components: (i) construction of a soil database at the farm level and (ii) delineation of soil functional units at the field level. Soil information relevant to PA is identified and methods for producing this information are presented (i.e., Questions 1 and 2 formulated above). Anticipated applications for soil information in operational decision support (i.e., Question 3) are closely considered and receive special attention in the results and discussion section.


   Materials and methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
Results and discussion
 Conclusions
 REFERENCES
 
Study Area
Research was conducted on a commercial arable farm in the central-western part of the Netherlands (51°17'N, 4°32'E). The farm covers an area of approximately 100 ha and applies a crop rotation with winter wheat, consumption potato and sugar beet as the main crops. Two fields were included in the study, covering areas of 14.7ha (Field A) and 10.5ha (Field B). Soils originate from marine deposits and are generally calcareous with textures ranging from sandy loam to clay. They are characterized as Typic Fluvaquents (Soil Survey Staff, 1998) or Mn25A-Mn45A on the Dutch 1:50000 soil map (Vos, 1984). Soil variability results from differences in texture (average clay content in 0–100 cm varies from 14–50%), soil organic matter (SOM) content (average SOM content in 0–100 cm varies from 0.5–5.8%) and subsoil composition (peat or mineral matter). With excellent drainage conditions, controlled by a dense system of pipe-drains, the area is considered prime agricultural land.

Soil Database
A detailed 1:5000 soil survey was conducted in the study area, counting approximately six soil auger observations per hectare. Results were stored in a soil database containing soil physical and soil chemical properties for individual soil layers. Texture and SOM content were estimated directly in the field and tested against a limited number of laboratory measurements to ensure accurate characterization. Based on these properties, soil layers were grouped into relatively homogeneous classes as defined by the Staringreeks (Wösten et al., 1994). This classification distinguishes between topsoil and subsoil layers, which are further differentiated by textural composition and SOM content. Sixteen classes were identified and sampled in the field. Average bulk density and saturated moisture content were determined for each class using at least four replicate samples.

Soil hydraulic characteristics were derived through a continuous pedotransfer function (PTF) developed at the DLO-Staring Center (Wösten et al., 1998). The PTF is based on soil physical measurements for 620 soil samples collected from major soil types in the Netherlands. It relates basic soil properties, such as texture, SOM content, and bulk density to a set of van Genuchten parameters (van Genuchten, 1980) that describe the moisture retention and hydraulic conductivity curves for individual soil layers. A sensitivity analysis by Vanclooster et al. (1992) identified the saturated moisture content as the most sensitive parameter to affect nitrate leaching from a Typic Hapludalf (Soil Survey Staff, 1998). Considering their results, measured saturated moisture contents were used to replace the PTF estimates.

Based on the layer information, soil profiles were classified according to the standards of the Dutch 1:50000 national soil map (Bregt et al., 1987). This classification focuses on genetic origin, topsoil texture, profile structure, and lime content. Following from its traditional design, the classification was chosen as a reference for the proposed functional classification.

Monitoring sites were installed at strategic locations in each field of the farm to provide a validation set for the simulation model. Groundwater levels and soil moisture contents were measured weekly during 1997. Moisture contents were measured at two depths (approximately 15 cm and 45 cm) using time domain reflectometry (TDR). As a result of their laborious character, soil-N measurements were concentrated on a single winter wheat field. Throughout 1998, monthly samples were collected at 10 sites and two depths (0–30 cm and 30–100 cm). Nitrate–N, ammonia–N and total-N concentrations were measured in a 50-mL KCl (1 M) soil extract using the Technicon Auto Analyzer II (Bran + Luebbe, Norderstedt, Germany).

Simulation Model
Dynamic simulations of soil–water–plant interaction were conducted with the mechanistic–deterministic simulation model WAVE [water and agrochemicals in soil and vadose environment] (Vanclooster et al., 1994). WAVE integrates four existing models that describe (i) one-dimensional soil water flow (SWATRER) (Dierckx et al., 1986), (ii) heat and solute transport (LEACHN) (Hutson and Wagenet, 1992), (iii) N cycling (SOILN) (Bergström et al., 1991), and (iv) crop growth (SUCROS) (Spitters et al., 1988). Differential equations governing water movement and solute transport are solved with a finite difference calculation scheme. For this purpose soil profiles were divided into 1-cm compartments. Water movement is described by the Richards' equation (Richard, 1931), which combines the mass balance and Darcian flow equations.

Water stress is calculated according to Feddes et al. (1978). Maximum uptake rates are defined by a sink term, which is considered constant with depth. Water uptake is reduced at high and low pressure head values, according to crop-specific thresholds. Stress resulting from N-deficiency occurs when required N-concentrations in the plant cannot be sustained by actual uptake rates. Crop production is then reduced proportionally to the ratio of actual over required uptake.

Verhagen (1997) made two conceptual changes to the model:

  1. Water uptake by plant roots was originally modeled assuming preferential uptake in the upper soil compartments, thereby excluding roots in deeper layers. After revision, water uptake is calculated as an integral over the entire root zone.
  2. Nitrogen uptake was originally controlled by the N-concentration in the leaves, which was specified as model input. After revision, N-concentrations and N-uptake are related to biomass production as described by Greenwood et al. (1990). Nitrogen concentrations are calculated as a negative exponential function of biomass accumulation.

Selecting and Quantifying Soil Functional Properties
Soil characterization focused on functional properties that describe soil-specific characteristics in terms of water regimes and nutrient dynamics. Selecting these properties is a subjective procedure. In order to increase transparency, some general guidelines were followed. First, the anticipated application of soil information was considered. In case of the DSS, soil information is used to facilitate precision management of water and nutrients. Selected functional properties should, therefore, be relevant to management decisions regarding irrigation and/or fertilization. Since K and P are applied at low frequencies and according to standard rules, attention was focused on N-fertilization. Another important consideration was the significance assigned to environmental parameters. In the Netherlands, as in the whole of Europe, new legislation is being implemented that imposes strict limits on N-fertilization of agricultural land. With levies imposed on excess N-fertilization, staying within limits has gained economic relevance. Environmental parameters were therefore included in the study. Finally, soil properties were required to (i) be quantitative rather than qualitative and (ii) provide insight into the actual supply and demand for water and N across fields.

Considering the above, four soil functional properties were selected:

  1. water stress in a dry year;
  2. N stress in a wet year;
  3. N leaching from root zone in a wet year;
  4. residual N-content at harvest in a wet year.

The first two properties describe the sensitivity of a soil to the effects of major growth-limiting factors. These are directly related to crop production and are relevant from an economic perspective. The third and fourth properties are environmental parameters that describe the pace at which nitrates are leached from the root zone.

Selected properties were quantified for individual soil profiles using the WAVE model (i.e., point simulations). Required soil parameters (e.g., van Genuchten parameters, SOM-contents, groundwater levels) were extracted directly from the soil database. Winter wheat was selected for crop growth simulation, considering the availability of soil-N measurements for model validation. Management parameters, defining the strategy for split N-fertilization, were defined according to general practice (Table 1) . Turnover rates for SOM-pools were taken from Droogers and Bouma (1997), who conducted incubation experiments for similar soil types in the region. Initial soil-N concentrations were chosen to correspond to the average concentration measured in February 1997 (60 kg N ha-1 within 0–100 cm).


View this table:
[in this window]
[in a new window]
  		Table 1 Selected N-fertilization strategy for winter wheat

 
Two simulations were conducted for each profile, which describe soil behavior under the extreme conditions of a dry year and a wet year (Table 2) . Sensitivity to water stress was expressed as the ratio of actual over potential evapotranspiration (ETact/ETpot). ETpot was calculated by multiplying the Makkink reference evapotranspiration (Makkink, 1957) with a series of crop-specific factors as tabulated by Feddes (1987). Similarly, N-stress was expressed as the ratio of actual over potential N-uptake (Nact/Npot). Note that both ratios were chosen in accordance with common reduction factors that are applied in crop modeling.


View this table:
[in this window]
[in a new window]
  		Table 2 Summary of precipitation (Pre) and potential evapotranspiration (ETpot) for a wet year (1987) and a dry year (1989)

 
Fuzzy Classification
Fuzzy c-means classification (FCM) was applied to identify classes of functionally similar soil profiles. Several authors, e.g., Burrough (1996) and McBratney and Odeh (1997), have described FCM classification. As opposed to traditional discrete classifiers, FCM expresses class-membership on a continuous scale of zero to one. Observations are assigned a specific membership vector that contains partial membership values for each designated class. The sum of these membership values is by definition equal to one (i.e., constant sum constraint). The concept of partial class-membership provides additional interpretative information that would not be available if a discrete classifier were used. Fuzzy c-means classification also enables the derivation of validity measures to assist in the selection of an appropriate number of classes, and finally, membership values can be interpolated with standard techniques to provide for continuously varying soil maps.

Fuzzy c-means classification was performed on the four functional properties derived in the simulation runs (i.e., water stress, N-stress, N-leaching, and residual N-content). Scaling and mutual correlation were accounted for by calculating the Mahalanobis distance between observations. The FUZNLM program (Vriend and Van Gaans, 1984), based on an FCM algorithm by Bezdek et al. (1984), derived multiple classifications for the two fields included in this study. In both fields the number of functional classes varied between 2 and 7. Two validity measures were derived for each classification: the fuzziness performance index (F') and the normalized classification entropy (H'') (Roubens, 1982). These measures quantify a degree of non-fuzziness that is maximized when F' and H'' reach their minimum value. McBratney and Moore (1985) indicate that the corresponding number of classes reflects a balance between structure and continuity that is generally pursued. The appropriate number of classes is derived from the data set, thereby eliminating a source of subjectivity from the classification.

Interpolation and Boundary Detection
Class-membership values for all soil profiles were interpolated using ordinary kriging (Journel and Huybregts, 1978). This required separate interpolations for each functional class, resulting in multiple grid maps. These grids were combined to derive a single confusion index map (CI-map) for both fields. The CI is defined as (Burrough et al., 1997)

in which µ1,i is the highest membership value in grid cell i and µ2,i is the second largest membership value in the same grid cell. In cases where CI-values approach 0, one functional class clearly dominates, leaving little confusion about class membership. Where CI-values approach 1, two classes have similar membership values, which results in confusion about class membership. On the CI-map these areas indicate transition zones between the functional classes. A boundary detection algorithm (ESRI, 1998) placed boundaries along these zones to delineate soil functional units. Each unit was assigned a representative soil profile that best matched the functional characteristics of its class (i.e., a selection of the highest membership value).

Analysis of Variance
Coefficients of determination were calculated to verify the efficiency of the fuzzy classification procedure. The coefficient of determination (R2 in percent) is defined as (Devore and Peck, 1986)

in which R2 is calculated as a function of SSW, the sum of squares within strata, and SST, the total sum of squares. Available soil functional and soil physical properties were stratified according to the class-membership of the corresponding soil profiles. R2 values were calculated for each property, indicating the percentage of the total spatial variation explained by the functional classes. This was interpreted as an efficiency indicator. A second series of R2 values was derived using traditional soil classes for stratification. This series served as a reference for the functional classification.


   Results and discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
Conclusions
 REFERENCES
 
Model Validation
Model performance was tested against measurements available in the soil database. Figure 1 presents simulated and measured (1997) soil moisture contents for three sites located in separate fields. The overall coefficient of determination is 63%. Figure 2 presents simulated and measured (1998) soil N concentrations. Each measurement represents an average concentration based on 10 samples taken from a single field. Steep increases of N concentrations were induced by split fertilizer applications on Days 52, 113 and 140. The overall coefficient of determination equals 71%.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 1 Measured ({diamondsuit}) and simulated (—) soil moisture contents. Measurements were collected from three winter wheat fields (Sites 1–3) during the 1997 growing season

 


View larger version (16K):
[in this window]
[in a new window]
  		Fig. 2 Measured ({diamondsuit}) and simulated (—) soil N contents. Measurements were collected from a winter wheat field during the 1998 growing season

 
Soil Functional Classification
Figure 3 presents validity measures derived for the different FCM classifications in Fields A and B. In both cases F' and H'' clearly identified an appropriate number of classes. Their values decreased to a minimum before continuing to increase slightly with the number of classes. Minimizing both measures resulted in four soil functional classes in Field A and three soil functional classes in Field B.



View larger version (12K):
[in this window]
[in a new window]
  		Fig. 3 Validity measures calculated for the fuzzy c-means classifications in Fields A and B

 
Tables 3 and 4 present soil functional and soil physical properties for the designated classes. Nitrogen stress never occurred in either field, which caused actual and potential N-uptake to coincide (Nact/Npot = 1). This was not surprising since management parameters reflected common practice with abundant N-fertilization. Water stress did occur and is represented by low values of ETact/ETpot. In both fields the sensitivity to water stress was highest for Class 2. The corresponding soil profiles appeared to contain a heavy non-calcareous subsoil layer hindering the upward flux of groundwater during summer. The presence of this layer is reflected in the relatively high clay content within 0–100 cm.


View this table:
[in this window]
[in a new window]
  		Table 3 Soil functional and soil physical properties (± SD) for the functional classes in Field A. Count indicates the number of soil profiles grouped in each class; R2 values indicate the percentage of spatial variation explained by the functional (Func.) and traditional taxonomic classes (Tax.)

 

View this table:
[in this window]
[in a new window]
  		Table 4 Soil functional and soil physical properties (± SD) for the functional classes in Field B. Count indicates the number of soil profiles grouped in each class; R2 values indicate the percentage of spatial variation explained by the functional (Func.) and the traditional taxonomic classes (Tax.)

 
A greater variation of N-leaching and residual-N contents explained the larger number of classes in Field A (4 vs. 3 in Field B). Soil organic matter content appeared to play a dominant role through the effects of N mineralization. Due to the abundant fertilization, mineralized N was not entirely used for crop production and, as a consequence, contributed to N-leaching and/or residual-N contents. Water regimes strongly influence this effect, as can be illustrated for Field A. Nitrogen leaching is practically identical for Classes 2 and 3, even though Class 2 contains 50% more SOM. This implies that soils in Class 3 leach their N residues at a greater pace. Most likely, this is a consequence of their lighter texture (less clay), which reduces water retention capacity and increases hydraulic conductivity. Compared with Class 3, Class 1 combines a higher residual-N content with 25% less SOM. In this case a heavier texture increases water retention and reduces hydraulic conductivity; the residence time for residual-N therefore increases.

Analysis of variance revealed the efficiency of the functional classification: coefficients of determination exceeded 65% for all functional properties (Tables 3 and 4). Traditional taxonomic classification rendered substantially lower values. The same result was found for SOM content: functional classes explained >70% of the spatial variation compared to only 17% (Field A) and 11% (Field B) for the taxonomic classification. This was not surprising since the taxonomic classes did not discriminate within the range of SOM contents encountered in the study area (0.5–5.8%). Clay content showed a somewhat different picture; while functional classes rendered a slightly higher R2 value in Field A, taxonomic classes could better describe the textural differences in Field B. This illustrates the flexibility of functional characterization; textural differences are only described as far as they have an effect on functional characteristics. Another striking difference between both classifications concerns the number of classes. Taxonomic classification rendered more classes in both fields: 7 vs. 4 in field A and 8 vs. 3 in Field B. This confirmed the efficiency of the functional classification.

Delineating Soil Functional Units
Figures 4 and 5 present grid maps describing the spatial distribution of class-membership values across Fields A and B. Semivariogram parameters used for interpolation are included in Table 5 . Transition zones between functional classes, appearing as light colored areas on the CI-map, formed clear patterns in both fields. Based on this confirmation of spatial grouping, soil functional units were delineated using CI > 0.9 as a threshold level for boundary detection. The delineated units are included in Fig. 4 and 5, along with their representative soil profiles.



View larger version (51K):
[in this window]
[in a new window]
  		Fig. 4 Soil sampling sites in Field A; spatial distribution of class-membership values; CI-map reflecting spatial uncertainty of class-membership; soil functional units with representative soil profiles

 


View larger version (46K):
[in this window]
[in a new window]
  		Fig. 5 Soil sampling sites in Field B; spatial distribution of class-membership values; CI-map reflecting spatial uncertainty of class-membership; soil functional units with representative soil profiles

 

View this table:
[in this window]
[in a new window]
  		Table 5 Spherical variogram models fitted to the membership values of the soil functional classes in Fields A and B. Goodness of fit was determined by the ratio SSD/SST [0,1]; perfect fit is attained when this ratio equals zero

 
Application in Precision Farming
Within the DSS, soil functional units serve as tools to reduce the theoretically infinite variability of soils to a limited, functional set that can be analyzed with simulation models. This is important because real-time simulations provide estimates of the actual supply and demand for water and N. A fertilization experiment conducted in 1998 confirmed the relevance of these data (Van Alphen and Stoorvogel, 2000). Figure 6 presents simulated soil-N concentrations and weekly N-uptake rates for a soil functional unit in the experimental winter wheat field. After a standard base-fertilization on Day 52, this unit received two additional N-fertilizations in reaction to early warning signals from the DSS (Days 113 and 140). Early warning signals were generated once simulated soil-N concentrations dropped below a critical threshold level, which was defined in accordance with the actual N-uptake rate. This strategy for precise fertilization was compared to traditional management in the same field. Precision management proved efficient in reducing fertilizer inputs (-23%), while slightly improving grain yields (+3%) and hectoliter weights (+4%). A similar experiment is being conducted in the study area to verify these results under different conditions.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 6 Simulated soil-N concentrations and weekly N-uptake rates used to optimize N-fertilization in a soil functional unit during 1998

 
While the implicit aim of PA is to treat each site according to its individual requirements, the enormous expense of data collection for informed decisions at this scale currently preclude the adoption of such intensive management programs (Whelan, 1999). Instead, various forms of spatial generalization have been proposed to derive management units for PA (e.g., Blackmore and Larscheid, 1997; Lark and Stafford, 1997; Van Uffelen et al., 1997; Boydell and McBratney, 1999). Based on the results of this study, it may be contested whether near-continuous variation of management operations should be pursued in all cases. Soil functional units derived in the study area could well describe the spatial variation of selected soil functional properties. Coefficients of determination exceeded 65% in all cases, meaning that <35% of the total variation was lost through generalization. If similar results can be shown for other regions, it will remain doubtful whether continuous adaptive management will become cost efficient and, if so, for which farming systems and management operations will this be the case?

Proposed methods for delineating management units often combine spatial and temporal analyses of yield data to identify areas showing a characteristic behavior for multiple years. Yield data are either measured on a combine harvester (Lark and Stafford, 1997), calculated with simulation models (Van Uffelen et al., 1997), or estimated from remote sensing images (Boydell and McBratney, 1999). Once management units have been established, important soil properties can be sampled in the field to identify the cause(s) of yield variations and enable site-specific treatment. Management units are thus used to stratify fields in order to increase sampling efficiency (e.g., variable N-fertilization based on N-sampling in early spring). Methods applied in the DSS add some advantages to this general concept. Yield patterns originate from the integrated effects of physical, chemical, and biological factors on crop production: therefore, they are difficult to interpret without additional soil information. Soil functional units, on the other hand, are well defined in terms of their soil functional characteristics. These may not reflect all sources of variation (e.g., the influence of pests and diseases is excluded), but they clearly describe the varying availability of major growth-determining factors (i.e., water and N).

In addition, real time simulations for representative soil profiles offer the opportunity to implement a proactive management strategy. Early warning signals generated by the DSS enable a farmer to respond before the crop is affected by stress. In this way the timing of management operations can be optimized, which has been shown a crucial factor in high-input farming systems in the Netherlands (Van Alphen and Stoorvogel, 2000). These systems are characterized by multiple treatments throughout the growing season (e.g., up to four split N-fertilizations in winter wheat) and are subjected to strict environmental regulations. The DSS, with soil functional units as its spatial component, provides a soil-based tool to maximize crop production under the constraints of environmental legislation.


   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
REFERENCES
 

  1. Soil characterization in terms of functional properties should be considered as an alternative to traditional taxonomic characterization when producing soil information for PA. Soil functionality, in this respect, should be expressed in relation to growth limiting factors that can be influenced by precision management (e.g., water and nutrients).
  2. The proposed methodology for soil characterization proved efficient in describing the spatial variability of selected functional properties. Coefficients of determination always exceeded 65%, meaning that <35% of the spatial variation was lost through generalization. Functional units are therefore considered suitable entities to be used as management units for PA.
  3. Selecting soil functional properties is a subjective procedure that should reflect a farmer's objectives. Parameters related to crop production (e.g., sensitivity to conditions of stress) will always be important, but environmental parameters (e.g., nitrate leaching) may be relevant as well. Both were considered in this study.


   ACKNOWLEDGMENTS
 
This study is embedded within a research project funded by the Dutch Board for Remote Sensing (BCRS). The research of J.J. Stoorvogel was funded by a fellowship of the Royal Netherlands Academy of Arts and Sciences. The van Bergeijk family is kindly acknowledged for supporting this research on their farm.

Received for publication November 11, 1999.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 J. Environ. Qual.Home page
M. J. Cohen, J. P. Prenger, and W. F. DeBusk
Visible-Near Infrared Reflectance Spectroscopy for Rapid, Nondestructive Assessment of Wetland Soil Quality
J. Environ. Qual., July 5, 2005; 34(4): 1422 - 1434.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
J. J. Stoorvogel, J. Bouma, and R. A. Orlich
Participatory Research for Systems Analysis: Prototyping for a Costa Rican Banana Plantation
Agron. J., March 1, 2004; 96(2): 323 - 336.
[Abstract] [Full Text] [PDF]

Home page
 J. Environ. Qual.Home page
B. J. van Alphen and J. J. Stoorvogel
Effects of Soil Variability and Weather Conditions on Pesticide Leaching-- A Farm-Level Evaluation
J. Environ. Qual., May 1, 2002; 31(3): 797 - 805.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Van Alphen, B.J.
Right arrow Articles by Stoorvogel, J.J.

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