A Comparison between Standard and Functional Clustering Methodologies: Application to Agricultural Fields for Yield Pattern Assessment

Abstract

The recognition of spatial patterns within agricultural fields, presenting similar yield potential areas, stable through time, is very important for optimizing agricultural practices. This study proposes the evaluation of different clustering methodologies applied to multispectral satellite time series for retrieving temporally stable (constant) patterns in agricultural fields, related to within-field yield spatial distribution. The ability of different clustering procedures for the recognition and mapping of constant patterns in fields of cereal crops was assessed. Crop vigor patterns, considered to be related to soils characteristics, and possibly indicative of yield potential, were derived by applying the different clustering algorithms to time series of Landsat images acquired on 94 agricultural fields near Rome (Italy). Two different approaches were applied and validated using Landsat 7 and 8 archived imagery. The first approach automatically extracts and calculates for each field of interest (FOI) the Normalized Difference Vegetation Index (NDVI), then exploits the standard K-means clustering algorithm to derive constant patterns at the field level. The second approach applies novel clustering procedures directly to spectral reflectance time series, in particular: (1) standard K-means; (2) functional K-means; (3) multivariate functional principal components clustering analysis; (4) hierarchical clustering. The different approaches were validated through cluster accuracy estimates on a reference set of FOIs for which yield maps were available for some years. Results show that multivariate functional principal components clustering, with an a priori determination of the optimal number of classes for each FOI, provides a better accuracy than those of standard clustering algorithms. The proposed novel functional clustering methodologies are effective and efficient for constant pattern retrieval and can be used for a sustainable management of agricultural fields, depending on farming systems and environmental conditions in different regions.

Environmental controls on Landsat‐derived phenoregions across an East African megatransect

“…The simplicity and efficiency of the K‐means clustering algorithm have resulted in its application across many disciplines (Bradley and Bradley 1998, Hoffman et al 2008, Kumar et al 2011, Mills et al 2011, Senthilnath et al 2017, Wang et al 2017, Pascucci et al 2018). Because K‐means is independent of location, the algorithm can categorize pixels in a Landsat scene that are not spatially close but belong to the same phenoregion (Kumar et al 2011).…”

“…There are many unsupervised statistical methods that have been used to determine an ideal number of clusters, such as Bayesian statistics (Senthilnath et al 2017) and the Hierarchical method (Chen et al 2005, Corstanje et al 2016, Grafius et al 2018). The elbow and silhouette methods (Subbalakshmi et al 2015, Rial et al 2017, Scharsich et al 2017, Wang et al 2017 a , b , Pascucci et al 2018) were used in this study because of their ease of interpretation and reasonable processing time.…”

“…An overall clustering quality is given between 1 and −1 for good clustering and poor clustering solutions, respectively (Subbalakshmi et al 2015). The clustering quality is based on similarity (shorter Euclidean distances between observation and cluster centroid) and dissimilarity (larger nearest neighbor distances between clusters; Pascucci et al 2018). The fpc (Hennig 2018) package was used to perform the silhouette method.…”

Unsupervised Monitoring Vegetation after the Closure of an Ore Processing Site with Multi-Temporal Optical Remote Sensing

“…The clustering algorithms largely used in remote sensing can be classified in three main classes: centroid-based methods such as K-means algorithm, neighborhood-based methods and hierarchical clustering [ 33 , 34 , 35 , 36 ]. The K-means approach is the best known clustering algorithm widely used in the last few decades [ 29 , 30 , 31 , 37 ], even if it can be unstable for large data sets. The two other classes limit this drawback but have higher computational costs.…”

“…The two other classes limit this drawback but have higher computational costs. In order to enhance the performance and reduce processing time without losing variability, dimension reduction techniques such as Principal Component Analysis (PCA) or spectral indices are often applied before clustering [ 30 , 35 , 37 ]. Moreover, the exploitation of the vegetation spectral indices rather than spectral signatures allows the clustering algorithm to focus only on specific vegetation traits [ 33 ].…”