The COVID-19 pandemic and its associated restrictions significantly impacted public perception of environmental risks, potentially altering sentiments expressed in water pollution complaints. Limited research exists on this effect. Additionally, climate extremes like floods and droughts can degrade water quality, indirectly impacting mental health and attitudes towards pollution. The Apalachicola-Chattahoochee-Flint (ACF) River basin in the southeastern US, including Alabama, is particularly vulnerable to such events, experiencing both water scarcity and limited resources for environmental management. The Alabama Department of Environment Management (ADEM) provides a web-based platform for water pollution complaints, offering valuable data for participatory environmental surveillance. This study uses text mining and sentiment analysis of over 10,000 ADEM complaints to examine how historical climate extremes (drought of 2016-2017 and floods of 2016 and 2019) and the COVID-19 pandemic (2020-2021) affected public sentiment (emotion and attitude) in these reports, also considering socioeconomic factors by comparing complaints from high-GDP counties and other counties. The research questions focus on how climate extremes altered public sentiment before COVID-19, how pandemic-related sentiment differs from that during climate extremes, and the extent to which socioeconomic factors influenced these sentiments.

Existing literature highlights the link between the COVID-19 pandemic and changes in risk perception and behavior (Wise et al., 2020; Qian & Li, 2020), as well as the pandemic's impact on mental health (Giuntella et al., 2021; WHO, 2022). Studies also show how climate extremes affect mental health (Berry et al., 2010; Cianconi et al., 2020; Hales et al., Stanke et al., 2013; Ebi & Bowen, 2016; Bell et al., 2018) and the challenges in predicting and managing their impact on water quality (Kam et al., 2021; Zhang et al., 2022; Antwi et al., 2021). The value of participatory surveillance systems in various fields, including environmental monitoring (Wang & Jia, 2021; Sun et al., 2021), is also established. However, there's a lack of research specifically on how pandemics and climate extremes affect the sentiment in public water pollution complaint reports.

The study utilized 10,690 water pollution complaints from the Alabama Department of Environmental Management (ADEM) website from 2012 to 2021, collected using BeautifulSoup. Data on droughts (from the US Drought Monitor), floods (from NOAA), and COVID-19 cases (from the COVID Tracking Project) were also gathered. The study defined specific periods for drought (D-2017: May 2016-April 2017), floods (F-2016: Oct 2015-April 2016; F-2019: Aug 2018-July 2019), and the COVID-19 pandemic (C-2020, C-2021: 2020, 2021). TextBlob was employed for sentiment analysis, calculating polarity (positive to negative) and subjectivity (objective to subjective) scores for each complaint. Bootstrapping (10,000 samples with replacement) was used to compare the percentage of positive/negative and subjective/objective complaints during these periods against the overall study period. This allowed for statistical significance testing. The impact of socioeconomic development was analyzed by comparing complaints from the top 15 GDP counties versus the remaining 52 counties. The analysis focused on percentage changes in complaint sentiment during each period relative to the baseline of the entire study period.

Annual water pollution complaints showed no significant changes from 2012 to 2018, with peaks in 2019 (related to specific pollution incidents) and 2020-2021. Monthly complaints exhibited seasonality, correlating with precipitation and temperature. From 2012-2021, 34% of complaints expressed negative emotion, and 32% showed objective attitudes. The top 15 GDP counties showed more objective and fewer negative complaints. During F-2016 (flash floods), there was no significant change in sentiment. However, D-2017 (drought) significantly increased negative complaints by +35% at the state level, with no significant attitude change. F-2019 (riverine floods) also resulted in a +25% increase in negative complaints and a −25% decrease in subjective complaints. In 2020 (C-2020), negative complaints decreased by −30% statewide, while subjective complaints increased by +20%. This trend persisted in 2021 (C-2021). Socioeconomic factors played a differential role in sentiment alterations before and after COVID-19. An additional analysis on a smaller dataset of complaints from 2019 and 2021 showed the impact of environmental incidents on the increase of positive sentiments.

The study demonstrates that COVID-19 significantly altered public sentiment in water pollution complaints. The decrease in negative sentiments during the pandemic might be due to reduced outdoor activity, constrained industrial activity and a shift in public attention towards public health concerns. The increase in subjective complaints could reflect the impact of pandemic-related stress and anxiety on how people perceived and reported environmental issues. The varying responses to different types of climate extremes (flash floods vs. drought vs. riverine floods) indicate the importance of understanding the specific characteristics of such events in managing public responses. The influence of socioeconomic factors on sentiment also requires further research. The findings highlight the importance of public outreach and education programs to improve the effectiveness of participatory surveillance systems and managing public expectations during environmental crises.

This study provides valuable insights into how climate extremes and the COVID-19 pandemic impacted public sentiment expressed in water pollution complaints. The results emphasize the need for effective communication strategies and resource allocation to manage public expectations and concerns during environmental crises. Future research could investigate the underlying mechanisms influencing these sentiment changes, focusing on individual-level factors, using qualitative methods like surveys and interviews to complement quantitative analysis. Further work can also explore the integration of traditional and digital technologies for more comprehensive environmental monitoring and management.

The study relies on publicly available complaint data, which may not capture the full spectrum of public sentiment. The data lacks details on the severity of the pollution incidents themselves, which might also affect the reported sentiments. The analysis focuses on Alabama, limiting the generalizability of findings to other contexts. The reliance on TextBlob for sentiment analysis may not capture all nuances of human language.