Georeferenced Multimodal Water Quality Dataset from Lago do Serpa (Amazon, Brazil)
Description
This dataset contains georeferenced physicochemical water quality measurements collected in Lago do Serpa (Itacoatiara, Amazonas, Brazil). The dataset includes spatial coordinates (latitude and longitude) and water quality parameters: pH, dissolved oxygen (DO), temperature (TEMP), apparent color (COR_A), true color (COR_V), and turbidity (TURB). Data were obtained through in-situ sampling campaigns and organized in CSV format. The dataset is suitable for spatial interpolation analyses (e.g., IDW), environmental modeling, and machine learning experiments. These data were used to evaluate multimodal spiking neural networks (SNNs) and edge inference performance across different embedded platforms, including Raspberry Pi, Android devices, and ESP32 microcontrollers, with benchmarking of latency and energy consumption. The dataset supports research in environmental monitoring, edge artificial intelligence, and spatial modeling in low-connectivity regions such as the Amazon.
Files
Steps to reproduce
1. Use `lago_serpa_raw.csv` for original spatial analysis and environmental assessment. 2. Use `lago_serpa_augmented_x10.csv` for machine learning training experiments. 3. Spatial interpolation can be performed using inverse distance weighting (IDW) or kriging methods. 4. All coordinates are provided in WGS84 / UTM Zone 21S. 5. Refer to `data_dictionary.csv` for variable definitions and units.
Institutions
- Universidade Federal do AmazonasAmazonas, Manaus