Reconstruction of Missing Hourly Precipitation Data to Increase Training Data Set for ANN

This paper investigates the hourly precipitation estimation capacities of ANN using raw data and reconstructed data using proposed Precipitation Sliding Window Period (PSWP) method. The precipitation data from 11 Automatic Weather Station (AWS) of Delhi has been obtained from Jan 2015 to Feb 2016. The proposed PSWP method uses both time and space dimension to fill the missing precipitation values. Hourly precipitation follows patterns in particular period along with its neighbor stations. Based on these patterns of precipitation, Local Cluster Sliding Window Period (LCSWP) and Global Cluster Sliding Window Period (GCSWP) are defined for single AWS and all AWSs respectively. Further, GCSWP period is classified into four different categories to fill the missing precipitation data based on patterns followed in it. The experimental results indicate that ANN trained with reconstructed data has better estimation results than the ANN trained with raw data. The average RMSE for ANN trained with raw data is 0.44 and while that for neural network trained with reconstructed data is 0.34.

Improving Hourly Precipitation Estimates for Flash Flood Modeling in Data-Scarce Andean-Amazon Basins: An Integrative Framework Based on Machine Learning and Multiple Remotely Sensed Data

Accurate estimation of spatiotemporal precipitation dynamics is crucial for flash flood forecasting; however, it is still a challenge in Andean-Amazon sub-basins due to the lack of suitable rain gauge networks. This study proposes a framework to improve hourly precipitation estimates by integrating multiple satellite-based precipitation and soil-moisture products using random forest modeling and bias correction techniques. The proposed framework is also used to force the GR4H model in three Andean-Amazon sub-basins that suffer frequent flash flood events: upper Napo River Basin (NRB), Jatunyacu River Basin (JRB), and Tena River Basin (TRB). Overall, precipitation estimates derived from the framework (BC-RFP) showed a high ability to reproduce the intensity, distribution, and occurrence of hourly events. In fact, the BC-RFP model improved the detection ability between 43% and 88%, reducing the estimation error between 72% and 93%, compared to the original satellite-based precipitation products (i.e., IMERG-E/L, GSMAP, and PERSIANN). Likewise, simulations of flash flood events by coupling the GR4H model with BC-RFP presented satisfactory performances (KGE* between 0.56 and 0.94). The BC-RFP model not only contributes to the implementation of future flood forecast systems but also provides relevant insights to several water-related research fields and hence to integrated water resources management of the Andean-Amazon region.

Reconstructing Precipitation Events Using Co-located Soil Moisture Information

Complete and accurate precipitation records are important for developing reliable flood warning systems, streamflow forecasts, rainfall-runoff estimates, and numerical land surface predictions. Existing methods for flagging missing precipitation events and filling gaps in the precipitation record typically rely on precipitation from neighboring stations. In this study, we investigated an alternative method for back-calculating precipitation events using changes in rootzone soil water storage. Our hypothesis was that using a different variable (i.e., soil moisture) from the same monitoring station will be more accurate in estimating hourly precipitation than using the same variable (i.e., precipitation) from the nearest neighboring station. Precipitation events were estimated from soil moisture as the sum of hourly changes in profile soil water storage. Hourly precipitation and soil moisture observations were obtained for a mesoscale network in the central U.S. Great Plains from May 2017 to December 2020. The proposed method based on soil moisture had a minimum detectable limit of 7.6 mm (95th percentile of undetected precipitation events) due to canopy and soil interception. The method was outperformed by the nearest neighbor (NN) interpolation method when neighboring stations were at distances of <10 km. However, the proposed method outperformed the NN method in 22 out of 27 stations when nearest stations were at distances >10 km. Using changes in soil water storage resulted effective in flagging and reconstructing actual missing precipitation events caused by pluviometer malfunction, highlighting new opportunities for using readily available in situ soil moisture information for operational quality control in mesoscale environmental monitoring networks.

Variation of Hourly Extreme Precipitation in the Three Gorges Reservoir Region, China, from the Observation Record

Extreme hourly precipitation is amongst the most prominent driving factors of flash floods and geological disasters. Based on the hourly precipitation data of 35 stations in the Three Gorges Reservoir Region (TGRR) from 1998 to 2020, we analyzed the spatiotemporal variation characteristics of hourly extreme precipitation indexes. The selected indicators included the frequency, intensity, period, annual maximum, trend of hourly heavy precipitation (20–50 mm/h) and hourly extreme heavy precipitation (≥50 mm/h) in the TGRR. Closely related climatic factors such as the Western Pacific Subtropical High Intensity (WPSHI) were also discussed. The results showed that in 2010–2020, the cumulative frequency of heavy precipitation magnitude between 25 and 40 mm/h slightly increased, while the corresponding frequency for magnitudes ≥50 mm/h decreased. In summer, the frequency of both heavy and extreme heavy precipitation increased in June and decreased in August, indicating a shift of extreme events to an earlier time in the flood season. The cumulative frequency of heavy precipitation in July had a period of about 7a, and that of extreme heavy precipitation had a period of 3a. The annual average intensity of heavy precipitation and extreme heavy precipitation in the TGRR was 28.9 mm/h and 61.4 mm/h per station, respectively, and both fluctuated and insignificantly decreased from 1998 to 2020. The annual maximum hourly precipitation center in the TGRR moved downstream from west to northeast. The frequency of heavy precipitation was relatively small along the main stream of the river valley. Both the frequency and total amount of heavy precipitation in southeast of the TGRR were significantly higher than those in other regions. Heavy precipitation in the majority of stations with high elevation (higher than 500 m) showed a decreasing trend. The cumulative frequency of precipitation with an intensity of 20–50 mm/h was closely correlated with the Western Hemisphere Warm Pool (WHWP) Index in February and the WPSHI Index in January, and especially, the abnormal large annual frequency (top 20%) showed strong correlation with the two indexes, implying highly predictable factors for extreme events. The frequency of precipitation intensity above 50 mm/h was correlated with the Western Pacific Warm Pool (WPWP) Area Index in January and the WPWP Intensity Index in November of last year. The research results provide a strong and refined factual basis for the assessment and prediction of extreme precipitation, and for disaster prevention and mitigation, in the TGRR.

The hourly precipitation frequencies in the tropical-belt version of WRF: sensitivity to cumulus parameterization and radiative schemes

The sensitivity of hourly precipitation to cumulus parameterizations and radiative schemes is explored by using the tropical-belt configuration of the Weather Research and Forecasting Model. The domain covers the entire tropical region from 45°S to 45°N with a grid spacing of about 45 km. A series of 5-year simulations with four cumulus parameterizations (New Tiedtke, NT; Kain-Fritsch, KF; New SAS, NS; and Tiedtke, TK) and two radiative schemes (RRTMG, CAM) are carried out. We focus on the frequencies of hourly precipitation above three thresholds (0.02 mm h-1, light drizzle rate; 0.2 mm h-1, moderate rate; and 2 mm h-1, heavy rate) between the observed CMORPH products and simulations. The sensitivity is higher for precipitation frequency than amount, and frequency is dominated by the cumulus parameterization. Frequencies above the moderate rate are well-reproduced, while frequencies above the other two rates present large deviations. No combination of physical schemes is found to perform best in reproducing the frequencies above all thresholds. Simulations using the NT and NS schemes show higher (lower) precipitation frequencies above the light drizzle rate (heavy rate) than those simulations using the KF and TK schemes. Precipitation frequency is higher reproduced by experiments using the RRTMG scheme than those using the CAM scheme, except for frequencies above the light rate over oceans. The overestimation of frequency is mainly caused by too-frequent convective rainfall. The results imply that the triggering based on the vertical velocity may increases the occurrence of rain event, while CAPE-based closure maybe increase the heavy precipitation frequency in the cumulus parameterizations.

Assessment of a Gauge-Radar-Satellite Merged Hourly Precipitation Product for Accurately Monitoring the Characteristics of the Super-Strong Meiyu Precipitation over the Yangtze River Basin in 2020

The recently developed gauge-radar-satellite merged hourly precipitation dataset (CMPAS-NRT) offers broad applications in scientific research and operations, such as intelligent grid forecasting, meteorological disaster monitoring and warning, and numerical model testing and evaluation. In this paper, we take a super-long Meiyu precipitation process experienced in the Yangtze River basin in the summer of 2020 as the research object, and evaluate the monitoring capability of the CMPAS-NRT for the process from multiple perspectives, such as error indicators, precipitation characteristics, and daily variability in different rainfall areas, using dense surface rain-gauge observation data as a reference. The results show that the error indicators for CMPAS-NRT are in good agreement with the gauge observations. The CMPAS-NRT can accurately reflect the evolution of precipitation during the whole rainy season, and can accurately capture the spatial distribution of rainbands, but there is an underestimation of extreme precipitation. At the same time, the CMPAS-NRT product features the phenomenon of overestimation of precipitation at the level of light rain. In terms of daily variation of precipitation, the precipitation amount, frequency, and intensity are basically consistent with the observations, except that there is a lag in the peak frequency of precipitation, and the frequency of precipitation at night is less than observed, and the intensity of precipitation is higher than observed. Overall, the CMPAS-NRT product can successfully reflect the precipitation characteristics of this super-heavy Meiyu precipitation event, and has a high potential hydrological utilization value. However, further improvement of the precipitation algorithm is needed to solve the problems of overestimation of light rainfall and underestimation of extreme precipitation in order to provide more accurate hourly precipitation monitoring dataset.