Original papers
Rectification methods for optimization of management zones

https://doi.org/10.1016/j.compag.2018.01.014Get rights and content

Highlights

  • Rectification for management zones optimization.

  • Management zones smoothed are made contiguous using a computational function.

  • Stray pixels and virtually all spots eliminated; class boundaries smoothed.

  • The rectification function facilitates the application of precision agriculture.

Abstract

The use of management zones (MZs) is an approach to precision agriculture that considers spatially contiguous subregions of the field, within which effects on the crop due to differences in soil, topography, and other abiotic factors are expected to be nearly uniform. Delimiting regions within the fields with similar yield potential and yield-limiting factors can lead for field management optimization. Regardless of the method used to delimit these zones, patches or isolated pixels generally appear. To smooth the MZs and improve their contiguity, a computational rectification function was implemented, allowing the analysis of 8 (3 × 3 mask) or 24 (5 × 5 mask) neighboring pixels using the statistical median and mode, to evaluate whether each pixel in the map should be reassigned to a different MZ. After being interpolated and normalized, sample data from three experimental fields were used to create clusters through fuzzy c-means algorithm, generating maps with two, three, four, and five classes. Then, the rectification function was applied five times on each map, which eliminated isolated pixels and virtually all patches, smoothing the boundaries between classes. The smoothness index showed higher variation in the first rectification as well as with an increase in the number of classes. The best performance was obtained with the 5 × 5 mask regardless of the statistical method used (median or mode). Our results show that these techniques are an effective way to increase the contiguity and smoothness of MZs, thereby improving their effectiveness, and are suitable for application in precision agriculture.

Introduction

The variability of nutrients in the soil directly affects crop yield; it may be related to factors such as climate, topography, organic matter, vegetation, geological processes, and soil management practices (Mallarino and Wittry, 2004). This influence occurs on different scales, complicating soil management and reducing the effectiveness of fertilizers if applied on a uniform scale (Mohammadi, 2002). As a result, different fertilization requirements may be needed for the same area. Since crop yield is influenced by soil characteristics, the study and understanding of the sources of variation is paramount to identify appropriate site-specific management (Rodrigues et al., 2016, Mzuku et al., 2005).

Spatial variability in crops is determined through monitoring and measurements, making it possible to create a plan for the correction of any deficiencies, particularly when site-specific management is intended, in order to improve soil quality, and, consequently, increase production (Davidson, 2014, Mzuku et al., 2005). Timlin et al. (1998) showed that topography and related factors such as soil depth and organic matter have a large effect on the variation of crop yields and can be used to identify management zones.

Normally, soil samples are analyzed to determine the levels of nutrients in the soil. The sampling should be dense enough to allow the determination of the variability of nutrients in the soil so that fertilizers can be used profitably and in an environmentally sustainable manner (Ferguson and Hergert, 2009, Franzen et al., 2002).

A management zone (MZ) is a subregion of a field that expresses a functionally homogeneous combination of yield-limiting factors for which a uniform rate of a specific crop input is appropriate (Bobryk et al., 2016, Doerge, 2000, Moral et al., 2010, Moshia et al., 2014). After delineation of the MZs, the number of samples needed to delineate the field soil variability can be reduced to one composite sample per zone, a technique suggested by Wollenhaupt et al. (1994) whereby subsamples are collected around georeferenced points, ensuring superior evaluation of nutrients in the area. This approach to sampling is likely to reduce laboratory costs while maintaining the level of reliability (Ferguson and Hergert, 2009, Mallarino and Wittry, 2004), and it has been shown to improve the efficiency of nutrient use, maintaining or increasing the yield and potentially reducing the overloading of nutrients into the environment (Khosla et al., 2002, Moshia et al., 2014). Many studies related to sampling density have been performed (Journel and Huijbregts, 1978, Demattê et al., 2014, Wollenhaupt et al., 1994, Franzen et al., 2002, Ferguson and Hergert, 2009, Doerge, 2000), resulting in a suggested minimum density of 1 sample ha−1 (Ferguson and Hergert, 2009) to 2.5 sample ha−1 (Journel and Huijbregts, 1978, Doerge, 2000), which should be composed of at least eight individual samples (Wollenhaupt et al. (1994)). It is reported in the literature that the use of MZs is commercially viable, but there is a need for further research to improve delineation techniques (Nawar et al., 2017).

Several kinds of data can be used to delineate MZs; however, it is advantageous to use a set of multivariate data attributes that do not vary significantly over time (topography, electrical conductivity, physical properties of the soil) and that are correlated with crop yield, thus producing more stable MZs (Buttafuoco et al., 2010, Doerge, 2000). Some researchers, such as Taylor et al., 2007, Guastaferro et al., 2010, Nawar et al., 2017, have also included crop yield as a variable in the delineation of MZs using clustering methods. According to Kitchen et al. (2005), the identification of areas of similar productivity potential called “productivity zones” may be useful when some key management decisions depend on reliable estimates of expected yield, such as the application of N fertilizer rate or seedling rate. In precision agriculture (PA), the terms “management zone” and “management class” are often used as synonyms. However, a management class is the area to which a particular treatment may be applied, whereas a management zone is a spatially contiguous area to which a particular treatment may be applied. Thus, a management class may consist of numerous zones, whereas a management zone can correspond to only one management class (Taylor et al., 2007).

Several approaches have been developed to delineate MZs; these are often classified as empirical or clustering according to the technique used (Córdoba et al., 2016, Guastaferro et al., 2010). The clustering approach is the most typical, and it has been used to obtain satisfactory results by many researchers (Arno et al., 2011, Moral et al., 2010, Saleh and Belal, 2014, Tagarakis et al., 2013). However, regardless of the procedure used to generate the MZs, small patches generally arise within a class. In order to resolve this issue, Lowrance (2014) created the EZZone software package, which smooths zones by merging small polygons having an area smaller than a certain threshold with larger polygons of different zones. Other researchers, such as Pramanik et al. (2013), have proposed that fields be merged by assigning weights to pixels, that is, by analyzing the asymmetry of neighboring pixels. To eliminate isolated cells or patches, Xiang et al. (2007) used post-classification majority filtering (mode). According to Lark, 1998, Ping and Dobermann, 2003, Córdoba et al., 2016, the use of spatial filters applied to the results of such classification is also recommended to improve zone contiguity. Nonetheless, none of these studies compared the mode and median statistics or used the smoothness index or Cohen’s kappa coefficient to evaluate the quality of the smoothing process.

The objective of the present study was to develop and apply computational techniques based on mode and median statistics to rectify and thus smooth MZs. The expectation was that this would eliminate the patches, making the zones more contiguous and thus more viable operationally.

Section snippets

Datasets

Data collection was based on an irregular sampling grid of three experimental fields located in rural areas of the state of Paraná, Brazil (Fig. 1) (Field A: 25°24′28′′S, 54°00′17′′ W; Field B: 25°26′49′′S, 54°04′59′′W; Field C: 25°06′32′′S, 53°49′55′′W). These fields have been cultivated under a no-tillage system for over 10 years, rotating soybean (Glycine max L.) with corn (Zea mays L.). Fields A, B, and C measured 9.9 ha, 19.8 ha, and 15.5 ha, respectively. They included 42 (4.2 sample ha−1

Variable selection

The results of the spatial correlation matrix obtained using SDUM show that the following attributes had autocorrelation at the 5% significance level: elevation, clay, silt, soil penetration resistance (SPR) 0.0–0.1 m in 2013, SPR 0.0–0.1 m in 2014, and SPR 0.2–0.3 m in 2012 in Field A; elevation, sand, clay, silt, and SPR 0.2–0.3 m in 2012 in Field B; and elevation, clay, silt, SPR 0.0–0.1 m in 2013, and SPR 0.2–0.3 m in 2014 in Field C. According to the methodology proposed by Bazzi et al.

Conclusion

The objective of this study was to develop and apply median- and mode-based computational techniques for rectifying and thus smoothing management zones (MZs), with the aim of making them more viable from an operational point of view.

The smoothness index showed higher variation in the first rectification as well as with an increase in the number of classes. Each time the function was applied, the edges were smoothed and the patches were reduced. After the fifth rectification, in most of the

Acknowledgements

The authors would like to thank the Federal University of Technology of Paraná (UTFPR), the Western Paraná State University (UNIOESTE), the Araucaria Foundation, the Coordination for the Upgrading of Higher Education Personnel (CAPES), and the National Council for Scientific and Technological Development (CNPq) for funding.

References (54)

  • J. Arno et al.

    Clustering of grape yield maps to delineate site-specific management zones

    Spanish J. Agric. Res.

    (2011)
  • C.L. Bazzi et al.

    Management zones definition using soil chemical and physical attributes in a soybean area

    Engenharia Agrícola

    (2013)
  • C.W. Bobryk et al.

    Validating a digital soil map with corn yield data for precision agriculture decision support

    Agron. J.

    (2016)
  • B. Boydell et al.

    Identifying potential within-field management zones from cotton-yield estimates

    Precis. Agric.

    (2002)
  • G. Buttafuoco et al.

    Delineation of management zones using soil properties and a multivariate geostatistical approach

    Ital. J. Agron.

    (2010)
  • J. Cohen

    A coefficient of agreement for nominals scales

    J. Educ. Meas.

    (1960)
  • R.L. Czaplewski et al.

    Expected value and variance of Moran’s bivariate spatial autocorrelation statistic for a permutation test

    USDA Forest Service

    (1993)
  • D. Davidson

    Evaluating the quality of your soil

    Crop Soils Mag.

    (2014)
  • J.A.M. Demattê et al.

    Precision agriculture for sugarcane management: a strategy applied for Brazilian conditions

    Acta Sci. Agron.

    (2014)
  • A. Dobermann et al.

    Classification of crop yield variability in irrigated production fields

    Agron. J.

    (2003)
  • T.A. Doerge

    Management Zone Concepts

    (2000)
  • R.B. Ferguson et al.

    Soil sampling for precision agriculture

    Ext. Precis. Agric.

    (2009)
  • D.W. Franzen et al.

    Evaluation of soil survey scale for zone development of site-specific nitrogen management

    Agron. J.

    (2002)
  • J.J. Fridgen et al.

    Management zone analyst (MZA): software for management zone delineation

    Agron. J.

    (2004)
  • R.C. Gonzalez et al.

    Digital Image Processing

    (2008)
  • D.M. Grzegozewski et al.

    Local influence when fitting Gaussian spatial linear models: an agriculture application

    Ciencia e Investigación Agraria

    (2013)
  • F. Guastaferro et al.

    A comparison of different algorithms for the delineation of management zones

    Precis. Agric.

    (2010)
  • Cited by (0)

    View full text