Before reading this article, please click to follow, convenient for future discussion and sharing, thank you for your support~
Text|Melon juice orange long Dr
Editor|Melon juice orange long Dr
introduction
Floods are occurring more frequently in the context of climate change, and flood monitoring capabilities are not well established, using a collaborative mapping framework to characterize the occurrence of floods in the summer of 2020.
and the impact on farmland in the middle and lower reaches of the Yangtze River Plain from the perspective of flood extent and intensity, from July to August, the total flood range was 4,936 square kilometers, and in terms of flood intensity.
Areas of 1,658 square kilometres, 1,382 square kilometres and 1,896 square kilometres experienced three, two and one flood, totaling 2,282 square kilometres of farmland, mainly from the Poyang Lake and Dongting Lake basins.
This study, which includes a high proportion of moderately damaged farmland, was 29 percent larger in 2020 than the largest flood in history between 2015 and 2019, and is expected to inform rapid regional flood hazard assessment and provide services for flood mitigation.
The application of remote sensing technology in near-real-time flood monitoring is related to earth sciences
As global floods are on the rise, timely and accurate flood information is particularly important as it provides the necessary information for disaster prevention and recovery strategies in the agricultural sector, which in turn contributes to food security.
Although flood identification based on traditional river measurements can predict numerical changes in flood area at a national scale, it does not provide a spatial pattern of flood impacts.
Studies on flood mapping that provide spatial distribution information can generally be divided into post-flood event, real-time and near-real-time mapping, depending on when the hydrological data used was acquired.
Post-flood event flood mapping uses multitemporal data and various hydraulic flood models to assess long-term damage to severe flooding after a flood event that is only available for small or medium-sized flood mapping and short river segments with a few tributaries.
An Afghan flood hazard map has been proposed to extract the extent of flooding nationwide, but it does not provide a spatial pattern of flood dynamics.
Real-time mapping uses data from water level sensors and timely official flood reports to generate flood probability maps, and due to the limitations of real-time hydrological observation data, it is difficult to extract timely flood information over a large area.
Quasi-real-time flood maps enable flood monitoring shortly after a flood event, providing a feasible method for timely and accurate assessment of large-scale flood damage.
In recent decades, the use of remote sensing technology in near-real-time flood monitoring has increased, and since flooding is often accompanied by precipitation and cloud cover, active satellite-based sensors such as synthetic aperture radar.
More suitable for near-real-time flood monitoring than optical sensors because they allow Earth observation in a variety of weather conditions, striving to extract flood information based on SAR images.
These images can be divided into three categories: flood mapping based on a single image, flood mapping based on change detection, and multi-temporal flood mapping, compared to single-image-based methods.
The method based on change detection uses a decrease in reflectivity compared to a permanent water body to distinguish between transient floods and permanent water bodies, since there is no significant change in permanent water bodies during this period.
Therefore, these methods are able to overcome the shortcomings of backscatter heterogeneity due to dry and submerged vegetation confusion, but the accuracy of the methods is highly dependent on the choice of non-flood reference images.
The change detection-based method cannot depict the entire large-scale flood mapping using only a single image during the flood, and the mapping using multi-temporal data can achieve comprehensive large-scale flood mapping.
Providing a better understanding of the seasonal behavior of different land cover types, time series analysis relies on the availability of high-frequency SAR data and complex processing.
The Google Earth Engine platform constantly updates Sentinel-1SAR data and hosts time series data processing algorithms, which makes it possible to handle near-real-time flood monitoring.
Near-real-time flood monitoring and flood impact assessment in Yangtze River Plain based on multi-temporal data
Comprehensive analysis of the scope, intensity and frequency of floods can help with precise management and prevention, especially in the case of damage to farmland that is severely affected by flooding.
Previous studies have focused on the extent of floods in a single period, the frequency of multi-year floods, or the frequency of multi-year floods, with little attention to large-scale flood events.
Southern China has suffered severe flooding, mainly in the middle and lower reaches of the Yangtze River Plain, where water levels in the Yangtze River Plain and major lakes continued to exceed warning levels in July 2020 due to continued heavy rainfall.
Lake Taihu water levels exceeded their maximum levels, with some sites in Poyang Lake reaching record highs, and in August 2020, the Minjiang and Jialing rivers in the upper reaches of the Yangtze experienced another huge flood.
According to the Ministry of Emergency Management of the People's Republic of China, flooding in the Yangtze River Plain in the summer of 2020 affected 34.2 million people in 11 provinces, with direct economic losses of 132.2 billion yuan, in this case.
Near-real-time monitoring of floods and flood dynamics in the Yangtze River Basin is urgently needed, with existing studies focusing on parts of the Yangtze River Plain, such as the Poyang Lake and Chaohu Basins, while few studies have covered flood monitoring and flood severity assessment across the Yangtze River Plain.
In this context, this paper aims to combine change-based methods and multi-temporal data to achieve rapid and accurate monitoring of flooding in the Yangtze River Plain and provide a comprehensive assessment of flood severity.
And to analyze the impact of flooding on farmland, based on the S1 SAR multi-temporal data provided by the GEE platform, the synergy of the two change monitoring methods was used to obtain the flood map of MLYP from 2015 to 2020 and the flooding map of farmland from 2020.
Designed to answer the following scientific question: Is the synergy of CDAT and NDFI algorithms suitable for mapping flood floods for MLYP? What is the specific manifestation of the severity of flooding in 2020 compared to previous years.
And how did the 2020 floods affect farmland? The proposed method allows the construction of accurate, consistent flood maps with an overall accuracy of 96.5% and an F-score value of 0.93.
To assess the ability of flood maps to distinguish between floods and permanent water bodies, a consistent flood map integrating NDFI and CDAT algorithms is shown, and the zoomed area gives a reasonable flood area compared to a partial reference image of the East Dongting Lake Plain.
Comprehensive analysis of the scope, intensity and frequency of flooding in the Yangtze River Plain and assessment of its impact on farmland
The comparison of the two maps is spatially consistent in the flooded areas, and the Sentinel-2-based flood map extracts additional flood areas on the Dongting Lake plain.
Most of the points are scattered along the 1:1 line with a coefficient of determination of 0.94, indicating a relatively good correlation between the two maps, however, flood maps based on S2 overestimate flooding conditions for the following reasons.
The spatial resolution of 30 meters leads to severe mixing pixel phenomena, which cannot exclude non-flooded roads and houses, increases the uncertainty of flood identification, and cloudy weather reduces image quality and interferes with flood extraction.
Dividing rice fields into flood zones also overestimates flood zones, with flood affected areas of 3,249 square kilometers, 3,652 square kilometers and 2,450 square kilometers in three phases, in 380 counties.
The flood affected 111, 133 and 124 counties in three phases, and the second phase of the flood affected most counties, which is the flood control period that government departments need the most attention to.
Counties and cities near the Dongting Lake Plain and Poyang Lake Plain were the hardest hit areas, with Jiujiang and Yugan counties in Jiangxi Province being among the five hardest hit counties in the three phases of flooding.
Due to the severe shrinkage of the lake area, especially Dongting Lake, the flood control capacity of the lake has been weakened, and due to the soil erosion in the upper reaches of the river, a large amount of sediment has been generated, which silts up in the midstream basin, and the lake area has been shrinking year by year.
Lake shore reclamation in the last century undermined the lake's function as a flood control reservoir, and the area of 1896 square kilometers, 1382 square kilometers and 1658 square kilometers experienced one, two and three floods respectively in terms of flood intensity.
Accounting for 38%, 34% and 28% of the area affected by floods, respectively, Figures 4A and 4B show the spatial distribution pattern of average flood intensity at the 10 km and county scales, with large areas of high flood intensity located around lakes or along major waterways.
This is due to the rapid rise in water levels and the low intensity of flooding in sporadic inland flood areas, possibly due to stagnant water caused by continuous rainfall during the rainy season.
As for flood affected areas of varying intensity at the county scale, Yuanjiang County and Yueyang County in the Dongting Lake Plain and Poyang County in the Poyang Lake Plain were severely affected by flooding, resulting in large-scale inundation areas.
In all three counties, the area of high-intensity areas was about twice that of low-intensity areas, suggesting that most flood-affected areas suffered repeated flooding.
Farmland and artificial infrastructure are constantly damaged, and in low-lying areas close to lakes or rivers, some heavily inundated areas are areas used as flood storage areas37 to prevent urban flooding.
Areas frequently affected by flooding are located near lakes or rivers, with the Dongting Lake Plain and Poyang Lake Plain each experiencing six floods, representing each of these areas from the 2015 to 2020 flood season.
The Yangtze River and its tributaries are regularly affected by floods, with three floods near major watercourses, 1-2 major tributaries, and 29% of the largest floods between 2015 and 2019 in 2020.
It can be observed that the flood area near the main stream of the Yangtze River in 2020 is more extensive than in previous years, and Yueyang County and Poyang County, located in the plain of two lakes, are the farmland areas most affected by flooding compared to other counties.
The 2020 floods affected the largest area of farmland at 2,282 square kilometers, or 46 percent of the total flood area, and floods in 2016 ranked second with 1,372 square kilometers, or 57 percent of the total flood area.
The floods caused varying degrees of damage to farmland in different regions, with the severely damaged areas concentrated in the Dongting Lake Plain and the Poyang Lake Plain, accounting for 25%.
The moderately damaged areas were mainly distributed in the main plains of the Yangtze River in Hubei and Anhui, accounting for 47%; The area of minor damage is located around the tributary plains of Hubei and Jiangxi provinces, accounting for 28%.
Nearly half of the farmland was moderately damaged, including 953 square kilometres of rice fields and 101 square kilometres of other crops, calculated on the basis of flood intensity and vegetation indices reflecting vegetation growth.
This pattern of distribution implies that farmland has suffered a lower number of inundations and moderate growth damage, and that farmers can compensate by replanting suitable crops, such as vegetables, as soon as the water recedes.
To reduce economic losses and avoid regional poverty, this study proposes a framework for MLYP near-real-time flood monitoring between 2015 and 2020 based on cloud computing based on the integration of CDAT and NDFI algorithms, all available S1 imagery, and the GEE platform.
Based on the flood result map, the extent, intensity and interannual variation of floods were comprehensively compared, and the impact of flooding on farmland was quantified by combining the resulting flood map with high-resolution farmland data, and the most affected farmland area in 2020 was discovered.
summary
An accurate and timely description of flood severity can help implement flood prevention strategies, reduce economic losses, and protect flood-prone farmland, unlike existing large-scale flood mapping studies, which mostly describe flood events on a single map.
Splitting the flood period into three time periods to analyze dynamic patterns and inundation intensity to support precise management and prevention in severely flood-affected areas, with interannual comparisons reflecting inundation frequency and flood hotspots.
Unlike other studies that only provide maps of flood frequency, 27,30 studies analyzed newly inunded areas of flooding in 2020 to measure the severity of floods and future trends affecting new areas, and these comprehensive results of near-real-time high-resolution flood mapping can be explained by the following facts.