2.2 Flood Forecasting Evaluation
The 3-day and 5-day lead time-based flood forecasting from BWDB is evaluated by comparing with the observed water level from 10 to 20 June, the timeframe within the extreme rainfall at Cherapunjee of India. The timeline of overtopping the water level above the assigned danger level is also observed for all the stations to observe the period of flooding. The anomaly of the forecasts is calculated to determine the effect of extreme rainfall in the BWDB stations situated in the upstream and the downstream regions of the portion of the Meghna Basin in Bangladesh.
2.3 Atmospheric River Detection
IPART technique [48] will be utilized to identify AR during periods of heavy rainfall in the Meghalaya region. Previously, The AR identification methods[49], [50] relied on the selection of a threshold of a parameter, such as IVT, or fixed value, such as 250 kg/m.s, which need to be exceeded within the region that satisfies specific geometric criteria, such as a length of 2000 km. Some contemporary studies used the surpassing of relative, spatial- and/or seasonally influenced values (specifically, the 85th percentile of Integrated Vapor Transport) as a means to detect ARs. The method will use the Integrated Vapor Transport (IVT) data at 6-hour intervals for June 2022 as input.
Unlike the abovementioned methods, IPART does not necessitate a particular threshold IVT value to define AR. The method is based on the image-processing technique known as "top hat by reconstruction" (THR), which involves subtracting a "greyscale reconstruction by dilation" image from the original one. Considering the detection of AR, the extent of AR is those AR candidates identified from the IVT anomaly that follow the geometric requirements. To generate the IVT anomaly, at first, the collected IVT data are gone through the grayscale reconstruction by dilation process consisting of two components for creating background IVT. Firstly, raw IVT data is gone through the grayscale erosion. Greyscale erosion can be comprehended through the use of similar methods of the moving average. Erosion displaces the central value with the minimum ones within the neighborhood instead of calculating the mean value among a neighborhood. Secondly, the erosion raster goes through the dilation process where the values are substituted with the maximum value for the reconstruction. Structuring Element E is implemented to characterize the neighborhood that works with the spatio-temporal scale and is representative of the physical processes associated with atmospheric rivers (ARs). After the processing of background IVT, the anomaly IVT is calculated through the subtraction of Background IVT from the original IVT data. Finally, a contiguous area where the THR Anomaly exceeds zero is characterized as the candidate of Atmospheric River. Nonetheless, the candidate considers both components of IVT (reconstruction and anomaly). Detected candidates are subsequently subjected to geometric filtering., specifically, the length of the object must be greater than 2000 km and the ratio of length to breadth must be greater than 2. Every identified AR instance is documented with an interval of 6 hours. The methodological framework is illustrated in Figure 2.