Linking social media data with geospatial information to analyse changes in human sentiments in and along surface water environments

Abstract

Social media data represent a valuable source of information on human activity patterns and emotional responses in relation to natural environments. These data can provide insights into the drivers of human sentiments toward freshwater ecosystems, especially in contexts where traditional survey methods are insufficient or resource intensive. A better understanding of the relationship between human sentiments and the perceived value of freshwater environments can support the integration of public perspectives into ecosystem management and regional development. In this paper, we present a replicable method for acquiring, cleaning, and analysing geolocated Twitter data from 2011 to 2018 from Germany. The method includes multiple data cleaning and filtering steps to prepare the dataset for identifying spatial and temporal trends in sentiments and to determine the primary drivers of emotional responses to water bodies. The demonstrated workflow includes the following steps:

• Geo-located Tweets were collected via the Twitter API, then sorted, indexed, and subjected to filtering and cleaning to ensure data quality.

• Language detection and sentiment analysis using a lexicon-based method (Polyglot), suitable for limited computing power, short-text social media sentiment analysis, particularly in the context of analysing the content posted by individuals spending time in freshwater ecosystems.

• Geospatial enrichment, incorporating contextual data such as weather, population density, and other location-based variables.