"""Documentation page for Streamlit app.""" import streamlit as st from PIL import Image from src.config_parameters import config from src.utils_sidebar import add_about, add_logo # Page configuration st.set_page_config(layout="wide") # Create sidebar add_logo("app/img/MA-logo.png") add_about() # Set fontisize text st.markdown( """ """ % config["docs_fontsize"], unsafe_allow_html=True, ) # Page title st.markdown("# Documentation") # First section st.markdown("## Methodology") st.markdown( """ The methodology is based on the workflow depicted in Figure 1. In addition to Sentinel-1 synthetic-aperture radar SAR data, two other datasets are used through Google Earth Engine:
The WWF HydroSHEDS Void-Filled DEM, 3 Arc-Seconds dataset is based on elevation data obtained in 2000 by NASA's Shuttle Radar Topography Mission (SRTM), and it is used to mask out areas with more than 5 percent slope (see following section on limitations).
The JRC Global Surface Water Mapping Layers, v1.4 dataset contains maps of the location and temporal distribution of surface water from 1984 to 2021, and it is used to mask areas with perennial water bodies, such as rivers or lakes.
Figure 1. Workflow of the flood mapping methodology (source).
""" % ( config["docs_caption_fontsize"], config["url_unspider_tutorial_detail"], ), unsafe_allow_html=True, ) # Second section st.markdown("## Radar imagery for flood detection") st.markdown( """ While there are multiple change detections techniques for radar imagery, the one used by Sentinel-1 is one of the simplest. Active radar satellites produce active radiation directed at the land, and images are formed as a function of the time it takes for that radiation to reach back to the satellite. Because of this, radar systems are side-looking (otherwise radiation from multiple areas would reach back at the same time). To be detected and imaged, radiation needs to be scattered back, but not all surfaces are equally able to scatter back, and that ability is also influenced by the radiation's wavelength (shorter wavelengths are better at detecting smaller objects, while longer wavelengths allow penetration, which is good for forest canopies for example, and biomass studies). Sentinel-1 satellites are C-band (~ 6 cm).Figure 2. Overview of the Sentinel-1 observation plan (source).
""" % (config["docs_caption_fontsize"], config["url_sentinel_img_location"]), unsafe_allow_html=True, ) # Third section st.markdown("## Key limitations") st.markdown( """ Radar imagery is great for detecting floods, as it is good at picking up water and it is not affected by the time of the day or clouds (at this wavelength). But it has its limits, and performs actually quite bad if having to detect water in mountainous regions, especially if with narrow valleys, and in urban areas (urban canyons). The reasons are mainly around the viewing angles, which can cause image distortions. This method may also result in false positives for other land cover changes with smooth surfaces, such as roads and sand. Rough surface texture caused by wind or rainfall may also make it challenging for the radar imagery to identify water bodies. """, unsafe_allow_html=True, ) # Last section st.markdown("## Useful links") st.markdown( """ UN-SPIDER recommended practice