This dataset encompasses data produced and analyzed in the context of the scientific article "Severe droughts reduce river navigability and isolate communities in the Brazilian Amazon" published at Communications Earth & Environment in 2024 (doi: 10.1038/s43247-024-01530-4). Three main analyses were performed for the article: (1) spatial analysis of accessibility of human settlements; (2) content analysis of media articles reporting impacts of severe droughts from 2000-2020, followed by temporal and spatial analysis; (3) hydrological analysis of water levels. The dataset includes: geospatial data, tabular data, and R scripts.

QGIS, 3.28.6

R, 4.+

Excel, Microsoft Office 365

METHODOLOGICAL INFORMATION

1. Description of methods used for collection-generation of data:

Distance analysis: This dataset presents the shortest distance between non-indigenous localities and indigenous villages in the Brazilian portion of the Amazon and (1) major water bodies during high water seasons; (2) major water bodies during low water seasons; (3) all roads, independent of road conditions. We defined major water bodies as floodable areas, mapped at 1-km spatial resolution, as estimated by the intercomparison raster map of minimum and maximum inundation in the Amazon (Fleishmann et al. 2022). For roads, (paved or unpaved) we used a combination of road datasets from the “Sistema Nacional de Aviação”, SNV, for official roads as of 2020, “Centro de Sensoriamento Remoto” (CSR/UFMG) as of 2016, and IMAZON as of 2012, the latter ones to account for unofficial roads (see Supplementary Material assoc. to the article for complete links to roads and inundation extent). One dataset was used to map non-indigenous localities: "Localidades" including remote rural settlements, villages, and cities derived from “Bases Cartográficas Contínuas” from the “Instituto Brasileiro de Geografia e Estatística” (IBGE) as of 2021 (Santos de Lima et al., 2024 Comms. Earth & Env. DOI: 10.1038/s43247-024-01530-4). The original dataset for non-indigenous localities, part of the collecion "Bases Cartográficas Contínuas 1:250.000", can be accessed through this link: https://www.ibge.gov.br/geociencias/cartas-e-mapas/bases-cartograficas-continuas/15759-brasil.html?=&t=downloads. To download the exact dataset, follow this: Downloads > bc250 > versao2021 > geopackage. After downloading the file "bc250_2021_11_18.zip", unzip the file and search for the geopackage layer "bc2050_2021_11_18 - lml_posic_geo_localidade_p". Technical information regarding the original dataset can be found here: https://geoftp.ibge.gov.br/cartas_e_mapas/bases_cartograficas_continuas/bc250/versao2021/informacoes_tecnicas/bc250_documentacao_tecnica.pdf (only in Portuguese). The dataset used to locate Indigenous villages is from the “Fundação Nacional dos Povos Indígenas” (FUNAI) updated in 2020. The original dataset can be accessed through this link: https://www.gov.br/funai/pt-br/atuacao/terras-indigenas/geoprocessamento-e-mapas. To access the dataset choose: "Download de dados geográficos das Aldeias" > "Aldeias". The boundaries of the Amazon River basin were obtained via "Agêcia Nacional de Águas" (ANA) (Coleção Ottobacias codificadas 2017, Nível 1, Bacia do Rio Amazonas) (https://metadados.snirh.gov.br/geonetwork/srv/api/records/0c698205-6b59-48dc-8b5e-a58a5dfcc989). From this website, choose: "Recursos Online" > "Bacias Hidrográficas Otto Nível 1 (gpkg) ". The boundaries of the Brazilian federal states that are part of the Amazon basin were extracted from the original dataset from IBGE (2018) (https://www.ibge.gov.br/geociencias/organizacao-do-territorio/malhas-territoriais.html).

Digital Media articles: Data collection of articles from digital media outlets was performed using Google Search engine. To collect the results, we employed the software platform Apify (https://apify.com/), which collects the URLs returned by specific queries into the Google Search engine. We set the scraper to return the first 100 search results for the query “amazônia seca navegabilidade” (amazon, drought, navigability) and “amazônia seca isoladas” (amazon, drought, isolated). Probably due to pagination issues, the scraper ended up collecting a little more than 100 URLs for each query: 113 for the first one (performed on 30 July 2020) and 117 for the second one (performed on 2 September 2020). A third data collection was performed on 27 January 2021 to consider only news from 2020. In total, the search returned 80 additional URLs for the query “amazônia seca navegabilidade” and 84 for the query “amazônia seca isoladas”. (Source: Santos de Lima et al., 2024 Comms. Earth & Env. DOI: 10.1038/s43247-024-01530-4)

Water level data: We obtained time series of water level data from river gauging stations managed by the Agência Nacional de Águas (National Water Agency). Original dataset (before processing and analysis) is publicly available at: https://www.snirh.gov.br/hidroweb/serieshistoricas. The data is part of the National Hydrometeorological Monitoring Network from Brazil (Rede Hidrometeorológica Nacional, RHN). More information about RHN can be found on https://metadados.snirh.gov.br/geonetwork/srv/api/records/f85dbf06-a869-414c-afc5-bb01869e9156 (only in Portuguese). The guidelines and manuals used to operate RHN stations can be found at https://www.gov.br/ana/pt-br/assuntos/monitoramento-e-eventos-criticos/monitoramento-hidrologico/orientacoes-manuais. Some documents are available in Portuguese and others in English.

1. Methods for processing the data:

Distance analysis: After selecting only point locations occurring inside the Amazon basin, we converted the inundation (raster) maps into point data layers where each point represents the centre of a given pixel, with data values indicating the number of models that agree with the mapped wetland extent. We then used the “v.distance” function from GRASS in QGIS 3.28.6 to calculate the “minimum distance to the nearest feature”. From this data, we calculated basic statistics for straight-line distances, considering agreement of at least 2 models for the maximum inundation map and 1 model for the minimum inundation map. As the inundation map did not include some portions of the low Amazon basin, leaving 101 settlements out, for these points, the procedure to calculate nearest distance to water bodies was different. We merged two datasets: hydrography from HydroRIVERS84, and water surface from the “Instituto Nacional de Pesquisa Espacial”. After that, we used the same “v.distance” tool to calculate the nearest distance to water bodies. More details can be found in the associated article (Santos de Lima et al., 2024 Comms. Earth & Env. DOI: 10.1038/s43247-024-01530-4). Descriptive numbers per federal state were performed ovelaying the map of Brazilian federal states.

Digital Media articles: News articles collected from digital media outlets were listed in a spreadsheet and manually processed following these steps: (i) verification of each web link; (ii) manual screening & reading; (iii) application of exclusion criteria; (iv) identification of basic metadata – including year, month, and location of the event; (v) extraction of statements mentioning impacts to communities; (vi) categorization of reported impacts; (vii) identification of the affected river and associated basin; (viii) identification of administrative boundaries; and (ix) identification the closest river gauging stations via “Hidroweb”86. When a single digital media article included more than one statement of impacts, we treated them separately. We adopted the following exclusion criteria: media articles related to books, dissertations, theses, and scientific papers; news about ongoing research; news that did not specify year and/or location; news without information on drought impacts; corrupted links; reports without data on water levels and/or navigation issues related to droughts; repeated entries. If the news piece mentioned the municipality or the river segment/sub-basin, we considered the location sufficient for inclusion. Pieces with a larger scope and no specific location information were discarded. We adopted a categorization scheme based on the economic activities/sectors affected by the drought, including the following 15 categories: passenger transportation, load transportation, fuel trade, water supply, food trade/retail, medicine supply, electric power, crops, medical care, access to school, hunting, fisheries, postal services, immunization & pest control, leisure. (Santos de Lima et al., 2024 Comms. Earth & Env. DOI: 10.1038/s43247-024-01530-4)

Water level data: Data were obtained online (https://www.snirh.gov.br/hidroweb/serieshistoricas) and processed using R scripts. After excluding 10 gauging stations under the direct influence of large hydropower dams (e.g., Santo Antônio, Jirau, Belo Monte, Balbina) (see related publication), we extracted the complete time series of water level for each remaining station. We then filtered out gauging stations with time series shorter than 15 hydrological years and eliminated years with data gaps greater than 10% per month, leaving a final dataset of 90 stations whose records were analysed. The time series for each station varied in length – that is, although our analysis started in hydrological year of 1978, not all stations began collecting data that year and several were deactivated after some time. As a reference for low water levels, we identified the lowest water level that was surpassed 80% of the time (P80) over the complete historical series of each gauging station. We then counted the number of days in each hydrological year (from 2000-2021) that water levels reached values below this long-term P80 (Santos de Lima et al. 2024, Comms. Earth & Env. DOI: 10.1038/s43247-024-01530-4).

1. Instrument- or software- specific information needed to interpret the data:

Distance analysis: QGIS (Open source from OSGeo), or any other GIS software capable of reading geopackage files. Digital Media articles: Online app Apify; Microsoft Office Excel 365 Water level data: R (version 4.+). We suggest using RStudio IDE.