An API-based Framework for Clustering Meteorological Time Series for Agricultural Applications

Abstract

Climate variability plays a key role in shaping agricultural outcomes, influencing crop growth, yield and resource management strategies. The ability to discern meaningful patterns in complex meteorological data sets is invaluable for optimizing agricultural practices and ensuring global food security. Climate zoning, for example, is important knowledge that contributes to the accuracy of agricultural recommendation systems. In this context, this work proposes an API-based framework to group meteorological data in time series format for agricultural applications. The main objective of the framework is to enable the identification of climate patterns in variables collected by meteorological stations, in order to provide climate zoning for agricultural decision-making, in a reliable and automated way. After an extensive and descriptive systematic review of existing literature, statistical techniques, clustering algorithms and similarity metrics were brought together, extended and implemented in API format. The set of applied methods was ordered in a logical and efficient sequence, guiding the tasks of data extraction, pre-processing, feature engineering, clustering and validation. The applicability of the framework was validated through two case studies using meteorological data, from FAWN/FL and SIMAGRO-RS, followed by a discussion of the results. The results obtained indicated the feasibility of using the framework, its contributions and limitations, highlighting its potential for improving input into agricultural decision systems. Finally, insights obtained from clustering are highlighted and ways of using the API are proposed in an interconnected way with agricultural systems.