Table of Contents
Introduction
Satellite-based remote sensing data gained importance for agriculture-related studies throughout the last decade. The distinct advantages of Spaceborne Synthetic Aperture Radar (SAR), describing one set of possible sensors to be used, are twofold: First, SAR operates independent of cloud coverage and, second, data acquisition is possible even at night-times. Compared to optical sensor data whose usage is often limited by cloud cover, this implies an increased effective temporal resolution. The variety of use cases of radar data in the field of agricultural studies is currently outlined by several reviews identifying cropland monitoring, crop parameter monitoring, and crop classifications as significant threads of the research field. Cropland monitoring mainly comprises studies on soil moisture estimations as well as surface roughness-based analyses of field cultivations. Crop parameter monitoring often aims at correlating temporal courses of SAR-based features with crop heights, biomasses or phenological stages. Finally, crop classifications differentiating types of fruits achieved very promising accuracies in recent years. Similar to crop parameter monitoring, crop classifications are of importance in estimating crop yields on a regional level but also to enforce practices adhering to agricultural policies. Belonging to the broader field of land use classifications, they also form the basis for soil erosional, climate-related and landscape change research.
The aim of the current post is to provide a concise overview of the existing research on SAR-based crop classifications. Therefore bibliometric analyses on hundred of studies will be carried out. The guiding questions of the analyses are:
1. How did the number of publications develop in time and space?
2. What are landmark contributions forming the intellectual basis of the research field?
Methods
To capture the current knowledge base, publications with their abstracts and relevant metadata were harvested from Scopus (Elsevier) and Web of Science (Clarivate). Both databases were searched for all publications fulfilling the criteria that their title and/or abstract contained all the following keywords: “radar” or “SAR” or “Sentinel-1” or “Radarsat” or “ERS” or “TerraSAR” and “crop” and “classification”. Search results downloaded in bibtex format were processed according to the workflow shown below.
Merging, filtering and subsequent creation of temporal and citation plots was done using R libraries including bibliometrix. The geocoding of study areas aiming at identifying the spatial foci of the studies was implemented via a jupyter notebook. As most publications do not have machine-readable metadata on study sites, an automated text-based toponym recognition and resolution was applied here. Both R and python scripts are stored in an open accessible GitHub repository.
Results
The main results are visualised alongside with a brief description below. For extended explanations, please take a look at the full version of the paper provided at the end of this post.
