The Earth Observation-based Anomaly Detection (EOAD) systemA simple, scalable approach to mapping in-field and farm-scale anomalies using widely available satellite imagery

Authors Organisations
Type Article
Original languageEnglish
Article number102535
Number of pages14
JournalInternational Journal of Applied Earth Observation and Geoinformation
Early online date08 Oct 2021
Publication statusPublished - 15 Dec 2021
Permanent link
Show download statistics
View graph of relations
Citation formats


To feed the world increasing population, expansion in the area under arable cultivation is expected, with the majority projected to occur in Sub-Sahara Africa and Latin American countries. However, many existing Precision Agriculture (PA) techniques are difficult to transfer to agricultural systems in these regions as they rely on prohibitively expensive crop monitoring systems. Satellite Earth Observation (EO) has the ability to provide affordable solutions, particularly to identify yield-limiting conditions within site-specific management zones (MZs). This paper presents the Earth Observation-based Anomaly Detection (EOAD) approach, a novel system for the detection of in-field anomalies through automatic thresholding of optical Vegetation Index data, based on their deviation from a normal distribution. The EOAD sets dynamic thresholds for the pixel values within a parcel by removing the atypical values in increments from the tails towards the median until the distribution is normal. The distribution normality is assessed based upon measures of skewness and kurtosis for each iteration. The anomaly detection approach demonstrated a strong agreement, 80% overall accuracy, with identified in-field anomalies when applied to rice plots in the Ibague Plateau, Colombia, using both Sentinel-2 and PlanetScope imagery. Areas identified as anomalous during the booting stage were shown to be significantly (p ⩽0.005) associated with a decrease in final yield. Additionally, the percentage of anomalies detected with the EOAD improved the detection of underperforming plots in early growth stages. Using freely available data and software, this automated approach demonstrates an exciting potential for use in improving agricultural practices in low-resource regions.


  • Agricultural fields, Anomaly detection, Precision agriculture, Sentinel-2