Colloquium

Key soil property identification and delineation of management zones in precision agriculture

Organisator Laboratory of Geo-information Science and Remote Sensing
Datum

do 10 april 2014 09:00 tot 09:30

Locatie Gaia, building number 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 317 48 16 00

By Tewelde Gebremichael Hagos (Ethiopia)

Abstract

Two adjacent precision farming fields located in south of the Netherlands were examined to delineate homogenous management zones for site-specific management. Spatially sensing like electrical conductivity (ECa), pH value, colour aerial photograph and digital elevation model (DEM) data sources were being obtained for these fields due to the advancements of proximal and remote sensing technologies. However, little attention was given to the over-information due to such a large volume of data from the technological progresses. The objective of the study was to develop and test the method of identifying key soil parameters to delineate management zones needed for precision farm management. Three scenario were considered for this study: i) combined analysis of both fields, ii) separate analysis of each field and iii) combined analysis of both fields but using national available geo-data sources (DEM and aerial photograph). Principal component analysis was used to identify soil variables which explain most of the soil variation for each scenario. Three principal components were retained for both scenario I and II while two principal components for scenario III. The spatial coherence and spatial distribution maps of the identified ECa’s, optical soil indices and elevation soil parameters were analysed using geo-statistical techniques. Unsupervised k-mean clustering algorithm was then performed to delineate potential management zones using the identified soil parameter for each scenario. Three optimal management zones per scenario were found most convenient based on the separable and overlapping nature of the classes. To assess the goodness of the defined management zones for each scenario, geo-referenced potato yield were examined and compared using ANOVA. In addition geo-located organic matter samples on field one were used for validation purpose. Significant mean differences of potato yield among and between the management zones for scenario I and II were found. Significant mean differences of organic matter were also found. In general the method can be operable for precision agriculture at field level using apparent electrical conductivity, colour aerial photographs and elevation sensing data sources.

Keywords: Precision farming; k-mean clustering; management zones; principal component analysis.