1    Introduction

A flash flood is a sudden and intense flow of water, usually occurring in a confined channel or across a sloping surface[1]. It originates due to extreme rainfall in upstream catchment areas, which are generally surrounded by mountainous regions[2]. Flash floods particularly pose a significant risk to the exposed community due to the severe damage associated with the lack of sufficient lead time-based forecast because of their instantaneous rainfall-runoff response[3]. Approximately 90% of fatalities resulting from flash floods, particularly in tropical regions, are attributed to drowning caused by individuals becoming drowned by swiftly rising waters [4]. A total of 1,185 fatalities occurred during 32 occurrences of flash floods in the United States of America from 1969 to 1981, resulting in an average of 37 deaths in each event observed when dams collapsed due to intense rainfall [5]. Furthermore, The Indian Meteorological Department (IMD) issued a warning for severe rainfall in Uttarakhand on August 23rd and 24th, 2023, following previous incidents where 80 people perished due to flash floods and landslides induced by extreme precipitation in the Himalayan part of India. Subsequently, these floods resulted in extensive property damage and widespread displacement of people [6]. The conventional structural measures of flood mitigation alone have shown limited functionality against flash floods; however, the combination of structural and non-structural flood mitigation methods is performing better worldwide [7]. Consequently, flash flood forecasting mechanisms, the main constituent of nonstructural flood mitigation measures, are being evolved for accuracy with prolonged lead time.
Globally, there are several methods employed for flash flood forecasting using different indicators. Firstly, hydrologic modeling followed by hydrodynamics modeling is utilized for the estimation of water depth in the stream where rainfall data gauges are scant. Secondly, accumulated rainfall in different timeframes is utilized as the indicator of upcoming flash floods. Thirdly, risk-based flash flood forecasting considers the vulnerabilities of the exposed population for the upcoming flash floods. The evolution of techniques for computing flash flood early-warning indices in China can be categorized into three distinct phases[8], [9]. During the early phase, an empirical method was employed where the prewarning varied from 1 to 24 hours. method. However, this method operates on the assumption that the frequency of rainfall is equivalent to that of flooding, and it does not account for the influence of antecedent rainfall. Later, they proposed warning indexes based on China's national context and specific characteristics: framework for verifying, calibrating, and reevaluating warning indexes at the provincial level. On the other hand, the Indian Meteorological Department (IMD) [10] is responsible for the dissemination of Flash Flood warnings in India, which are communicated to the public via the National Bulletin. Based on the analysis of Merged Mean Areal Precipitation and actual rainfall. In addition, the provision of Flash Flood Risk products is available for durations of 12 to 36 hours, categorized into regions with varying levels of threat: High, Moderate, and Low. Furthermore, the Indian Meteorological Department (IMD) is currently developing a novel Flash Flood Guidance System. This system is designed to monitor many factors including soil moisture, soil temperature, soil saturation, topography, and real-time rainfall in order to enhance the prediction and detection of flash floods.
Various strategies have utilized rainfall occurring in the upstream of a watershed as the trigger for flash floods. A new rainfall triggering indicator, denoted as β, has been proposed to categorize floods resulting from intense intra-day rainfall and substantial cumulative rainfall. They stated that the Rainfall Triggering Index (RTI) is more effective in cases where the flood is caused by a significant amount of accumulated rainfall. Alfieri [11] provided a concise summary of three flash flood warning techniques that make use of observed rainfall data from gauge stations, high-resolution radar, and rainfall forecasts. Flash flood forecasting based on Numerical Weather Prediction (NWP) has challenges such as accurately determining the exact position and time of extreme precipitation events. Moreover, the limited accessibility to data from transboundary catchments may impede the utilization of these models. Weishuai [12] conducted a comprehensive analysis of the current approaches used in China to calculate rainfall thresholds. The study revealed that the relationship between rainfall thresholds and the severity of floods is not well-established. Furthermore, it emphasized the need to resolve the uncertainty associated with rainfall threshold calculations. However, the lack of data sharing for transboundary catchments restricts the utilization of these models.  In addition, satellite-based gridded rainfall products such as Global Precipitation Model (GPM), Asian Precipitation - Highly-Resolved Observational Data Integration Towards Evaluation (APHRODITE), Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), and others provide global coverage. However, they are unable to account for orographic rainfall due to its intricate nature. Bharti [13] identifies a height of 3100m as the threshold at which the satellite begins to overestimate rainfall amounts for elevation ranges below this threshold and underestimate them for elevation ranges above it. The connection between individual rainfall events is weak, with a value of approximately 0.23. However, the correlation improves to around 0.67 when considering areal-averaged precipitation values. Salio [14] demonstrated a notable underestimation of precipitation that was evident in all estimations over the examined period.  The investigation reveals a significant correlation between errors and terrain height for all techniques that integrate infrared and passive microwave data. Climatic drivers like Atmospheric River can be an alternative in this regard as it generate extreme orographic rainfall events in different parts of the world.