Application of multivariate geostatistics in delineating management zones within a gravelly vineyard using geo-electrical sensors
Introduction
Efficient methods for accurately measuring within-field variations in soil properties are crucial for precision viticulture (Bramley, 2005). Sampling at discrete places has been the traditional means of obtaining information about the soil, but soil surveys are generally time-consuming, labour-intensive and costly, especially in the gravelly soils characterising some of the most important terroirs in the world. The large numbers of samples required in gravelly areas in order to attain a good representation of the soil properties (Buchter et al., 1994) limit the possibility of adopting an appropriate sampling intensity to determine the spatial variability within vineyards.
The potential use of ancillary data that can be intensively recorded, such as soil bulk electrical conductivity (EC) measured by electro-magnetic induction (EMI) surveys, has been well examined over the last decade. This is because data are relatively easy and inexpensive to collect (Blackmer et al., 1995, Mulla, 1997). If the sparse and more intensive data are spatially correlated, then the additional information from the ancillary data can be used to improve the estimate precision of the sparsely sampled primary variable. Several scientists have used EMI surveys to characterise soil salinity (Rhoades et al., 1999a) and nutrients (Kaffka et al., 2005), texture (Triantafilis and Lesch, 2005), bulk density related (Rhoades et al., 1999b) and many other properties (Corwin and Lesch, 2005). EMI investigations were also applied to identify morphological features such as depth to boulder clay (Brus et al., 1992) or clay pan (Sudduth et al., 1995). Although EMI is useful for looking at lateral spatial variation, it gives limited information on how conductivity varies with depth because the relationship between a specific earth domain and a particular EC reading is poorly quantified (Pellerin and Wannamaker, 2005). To improve the characterisation of the soil profile, the EMI method can be coupled with electrical resistivity measurements to increase the vertical resolution of subsurface electrical images (Rizzo et al., 2004). Few papers have used different geo-electrical techniques in an integrated way (De Benedetto et al., 2008).
Modelling the relationships between primary soil variables and EC is essential to assess and describe the spatial variability within a vineyard with sufficient precision and then identify management zones. The task is not generally easy, because EC depends on many soil properties over different spatial scales, in a very complex and non-linear way. Moreover, difficulties increase when sampling intensity is reduced by unfavourable soil conditions such those in gravelly soils. Several methods have been proposed to incorporate secondary information. A number of “hybrid” interpolation techniques, combining geostatistical technique of (co)kriging with exhaustive secondary information, have been developed and tested to improve primary variable precision (Goovaerts, 2000, McBratney et al., 2000, Frogbrook and Oliver, 2001). Kriging with external drift (Royle and Berliner, 1999, Wackernagel, 2003) is a non-stationary geostatistical technique, based on a model assumed for the conditional distribution of the primary variable and taking into account the linear relationship between primary and auxiliary data. Hierarchical spatial regression models (Triantafilis and Lesch, 2005) and regression kriging (Hengl et al., 2004) have been used as an alternative to cokriging. Another technique is an approximation of multivariate extension of kriging, known as collocated cokriging, which has proved to be well-suited to merging types of information with different resolution (Castrignanò et al., 2008).
Recent research on precision viticulture has focused on the use of management zones, which are defined as subfield regions within which the effects on the crop of seasonal differences in weather, soil, management, etc. are expected to be more or less uniform (Lark, 1998). For this purpose it is often useful to define classes from a set of multivariate spatial data stored in a geographical information system (GIS). Cluster analysis procedures have been effectively used to divide a field into potential management zones (Stafford et al., 1999). This methodology groups similar individuals into distinct classes called “clusters” in the N-dimensional character space defined by the N properties measured for each individual (Lark, 1998). Existing traditional clustering techniques produce natural groupings of the data only in the attribute space without any reference to geographical position. Spatial classification obtained by clustering does not account for the spatial correlation between observations and takes little account of gradual change, either from one class to another or within any one class. On the contrary, a multivariate geostatistical technique, called factor kriging analysis (FKA) (Castrignanò et al., 2000, Bocchi et al., 2000, Casa and Castrignanò, 2008), treats multivariate indices of spatial variation (regionalised factors) as continua in a joint attribute and geographical space.
The objective of this work was to propose a procedure, based on multivariate geostatistics, to build maps of soil attributes and classify gravelly vineyards in zones to be differently managed.
Section snippets
Study site
The study site is a 5-ha vineyard at San Pietro in Cariano, Valpolicella (north-eastern Italy; 45°31′ N 10°53′ E, 145 m a.s.l.), located in a DOC (controlled denomination of origin) area producing Valpolicella and Amarone wines. The climate is sub-humid, with mean annual rainfall of about 850 mm distributed fairly uniformly throughout the year. From December to February the temperature rarely falls below zero, while maximum temperatures in summer vary from 25 to 30 °C during the day and 18 to 20 °C
Results and discussion
Gravel in the top layer ranged from 192 g kg−1 to 751 g kg−1, with an average of 474 g kg−1 (Table 1). Gravel of 20–100 mm diameter was the most representative fraction (67% of the total gravel), while stones >100 mm in diameter were not found in the majority of samples. Gravel content sharply increased with depth, with values higher than 600 g kg−1 in layers deeper than 40 cm (Table 2). Gravel fraction >100 mm also increased with depth, reaching a content of 117 g kg−1 in the 80–100 cm layer. This fraction
Conclusions
In this work, multivariate geostatistical analysis has allowed the relationship between EC and some soil physical properties to be described and, coupled with fuzzy c-means classification, to delineate potential management zones. The utility of EMI mapping comes from the relationships that often exist between EC and a variety of soil properties. Spatial variation of soil properties could therefore be advantageously inferred, using ancillary data, which are less expensive to obtain. In gravelly
Acknowledgements
The authors are very grateful to Dr. G. Morelli of Geostudi Astier, Livorno, for his valuable work in the realisation of the geo-electric measures. The authors also wish to thank the Pule Vinery for hosting the experiment and Marco Marconi for his support in the field operations.
References (55)
- et al.
Application of factorial kriging for mapping soil variation at field scale
Eur. J. Agron.
(2000) - et al.
Sample size determination of coarse fragment content in a stony soil
Geoderma
(1994) - et al.
Analysis of spatial relationships between soil and crop variables in a durum wheat field using a multivariate geostatistical approach
Eur. J. Agron.
(2008) - et al.
Study of spatial relationships among some soil physico-chemical properties of a field in central Italy using multivariate geostatistics
Geoderma
(2000) - et al.
Apparent soil electrical conductivity measurements in agriculture
Comp. Electron. Agric.
(2005) Geostatistical approaches for incorporating elevation into the spatial interpolation of rainfall
J. Hydrol.
(2000)- et al.
A generic framework for spatial prediction of soil variables based on regression-kriging
Geoderma
(2004) - et al.
Site-specific management in salt-affected sugar beet fields using electromagnetic induction
Comp. Electron. Agric.
(2005) - et al.
Delineation of site-specific management zones using fuzzy clustering analysis in a coastal saline land
Comp. Electron. Agric.
(2007) - et al.
An overview of pedometric techniques for use in soil survey
Geoderma
(2000)
Multi-dimensional electromagnetic modeling and inversion with application to near-surface earth investigations
Comp. Electron. Agric.
High-resolution images of the fault-controlled High Agri Valley basin (Southern Italy) with deep and shallow electrical resistivity tomographies
Phys. Chem. Earth
Mapping clay content variation using electromagnetic induction techniques
Comp. Electron. Agric.
Key soil and topographic properties to delineate potential management classes for precision agriculture in the European loess area
Geoderma
Geophysical Techniques for Sensing Buried Wastes and Waste Migration
Remote sensing to detect nitrogen deficiency in corn
Variation in grape yield and soil depth in two contrasting Australian vineyards
Understanding variability in winegrape production systems. 2. Within vineyard variation in quality over several vintages
Aust. J. Grape Wine Res.
The use of electromagnetic measurements of apparent soil electrical conductivity to predict the boulder clay depth
Geoderma
Analisi delle relazioni spaziali tra alcune proprieta’ del suolo e le misure di conducibilita’ elettrica: possibili applicazioni alla viticoltura di precisione [Assessment of spatial relationships between some soil properties and electromagnetic induction scans: potential applications to the precision viticulture]
Variogrammes et krigeages pour la gravimétrie et le magnétisme [Variograms and kriging for gravimetry and magnetism]
Sciences de la Terre, Série Informatique
Soil clay mapping by measurements of electromagnetic conductivity
Non-intrusive mapping of subsoil properties in agricultural fields with GPR and EMI
Porosity
Management zone analyst (MZA): software for subfield management zone delineation
Agron. J.
Cited by (118)
Summarizing soil chemical variables into homogeneous management zones – case study in a specialty coffee crop
2024, Smart Agricultural TechnologyAn algorithm to identify the appropriate size of a neighbourhood when analysing spatially distributed field experiments by local cokriging
2021, Computers and Electronics in Agriculture