If you look at the satellite data, it seemed completely impossible to me that a spacecraft orbiting the earth several hundred kilometers away could see a flooded road in my city. Floods are very messy, messy, and often unpredictable. However, Radar satellites have become more sensitive in the last few years, and the algorithms are more intelligent, so it is possible to monitor the water flowing in houses, fields and rivers. I wrote this article to explain how the trick works. They are not complete “AI + satellite =” satellites, but real, from the point of view of someone who has spent many nights looking at SAR (Synthetic Apertere Radar) full of noise, trying to find out what they really mean.
My basic message: To be able to detect floods in real time and be able to rely on such maps, one must go beyond visual images and understand the backscatter geometry of the backscatter bar. The satellite system of India (Radar Moging) is an excellent example of how physics-based pipelines can be robust and independent of the weather needed for the timely delivery of floods that can be used in large-scale disaster situations, such as the monsoon season.
Amazing beauty and physics of bar data
Most people see satellites as photography devices, but Sar is very different from a camera. It does not record light; In fact, it creates its own light. In the case of a satellite like risat, it is a practical operation in which the satellite sends a combined beam of microwaves to the earth and records a small part of the energy reflected back to it, called backscatter.
Why does water appear black (effect of will))
The image brightness produced is not a measure of visible light, but an account of how the radar power changes with contact with the surface. Such communication depends on how hard it is and what the material of the surface is in relation to the wavelength of the Radar.
- Dry, aggressive areas (vegetables, urban areas): Radar waves scatter in many different directions when they strike an abrasive surface, like light striking a piece of foil with water. Most of this scattered energy returns to the satellite → bright pixels.
- Smooth surface of water: A calm surface of water is like a very smooth mirror. When the radar waves hit, they reflect almost all of the energy away from the satellite, just as a mirror reflects light in one direction. The smallest amount of energy that is reflected back to the sensor → black pixels (indicating too much reflection).
Such ability to penetrate dust, rain, dust and smoke is what makes SAR indispensable for disaster response in cloudy, high humidity areas.
Core Maple map pipeline: from Echo to MAP
SAR satellite imagery is not available directly from the download. RISAT's core flood detection process is a well-established, physics-based Science Pipeline. Any mistake made at the beginning affects all subsequent results, which is why careful processing is very important.
1. Preparation of radar data
Actually the first step is to transform the raw satellite data in such a way that it produces visible measurements of backscatter. This step produces numerical values on images that show a true representation of the earth's surface that can be reliably compared to other images.
2. To reduce image noise
Speckle is granular noise, salt-and-pepper-like in SAR images by nature. This noise must be reduced in a way that does not obscure the contours of the earth, in particular, the sharp boundaries between the earth and the water.
Challenge: Improper application of the noise reduction method can remove small flood details or change water boundaries. The robust method leaves a lot of noise that can cause errors in the identification of flooded areas.
Solution: It is a clear result of the image, suitable for analysis, because special filters are introduced to smooth the noisy parts while preserving the important edges.
3. Detecting Change: Algorithmic Centerpeap
In fact, floods are a great help in renovating the earth to become radar capable – from a bright land to a place in the dark. Therefore, the comparison of a radar image taken before the Flood with one taken after allows us to find the exact locations of the fires.
One of the most effective methods is to determine the change in brightness between photos taken before and after the Flood. Those areas that have changed from land to water will have a large difference, thus presenting an area that is almost completely flooded
4. Separation and refinement of floodplains
The final tasks are about finding pixels that correspond to flooded areas and making sure the map is correct:
- PREBOLDING: An automatic method that finds those pixels change large enough to be considered 'flooding'. Thus, the first map of the flooded areas is obtained.
- Use of additional data: to deepen the accuracy, we turn to various types of spatial data. For example, we exclude areas that are always under water (such as endless lakes or rivers) and do not consider small slopes (which can easily be interpreted as dark areas on radar images due to shadows). This provides a means to eliminate false positives and ensure that the final flood map is accurate.
The nuance of radar settings and human intervention
One of the smaller decisions that has more impact than the algorithm is the choice of the right radar settings, especially the way the radar waves are sent and received (known as polarization).
Different configurations of smoke can produce different characteristics of the area. When it comes to observing floods, the arrangement of a specific polarization (often called VV polarization) is often chosen as it results in a greater uniformity between the water and the light signal from the surrounding land.
Why human judgment is still a pure ai
In current flood mapping operations, traditional methods have been found to produce more reliable results than sophisticated artificial intelligence types. This is because traditional methods are flexible and flexible.
- AI Challenge: General-Injection AI models have a hard time dealing with the natural noise in radar data. Additionally, these models fail when moved to a new location. For example, an AI model trained for flooding in a flat, urban city may not work in a mountainous, agricultural DELTA.
- The human edge: even though the same satellite data is used, two professional analysts can come up with different flood maps. This is not a myth; rather, it is a matter of nuance. The analyst applies their knowledge to:
- Adjust the flooded areas according to the terrain adjustment (noting that a flooded rice field would look different from a flooded street).
- Consider the need to find all flooded areas against the possibility of finding non-flooded areas as flooded (false alarms).
And AI is slowly gaining, especially at the right level. Advanced methods use the reliable physical principles of radar and AI to not only limit the flooding but also raise the level of detail. By doing so, the understanding of radar physics remains a primary consideration while AI is used to improve the final product.
Lasting
The RISAT program is one such initiative that achieves this by providing consistent and reliable data that is evolving to transform flood intelligence and geopatial intelligence. Currently, flood mapping is actually the meeting point of recent developments in natural models, data processing, and the use of geo-spatial expertise by human representatives.
Understanding and interpreting background patterns is a key step in moving from a casual view of the crisis to a deeper understanding of the scale and flow of the crisis, thereby allowing for timely intervention. Besides, risat and similar efforts should not be considered as technical devices installed in a certain space, but rather as a key tool that supports the harmonious operation of the analyst and the favorable environment. That is, our maps are higher and more accurate, helping teams are able to assemble and implement their tasks in a shorter time – being a perfect example of how data science can be a direct asset to humanity.
Thanks for visiting and reading.
Progress
- ISro, “Risat-1A Mission Overview,” (2022), ISRO website.
- Esa, “Sentinel-1 SAR tutorials,” (2021), ESA documents.
- JAin, kumar, singh..
- NRSC, “India's Atras Atlas floods,” (2019), National Remote Sensing Center report.
- Schumann & Moller, “Remote sensing of floods,” (2015), journal of hydrology.
