Impact-based forecasts provide information on exposure, vulnerability, and risk, which are essential for quantifying risks and facilitating timely evacuation. For small- and medium-sized rivers, accurate deterministic flood forecasting more than three hours in advance is quite difficult owing to forecast errors in the location and intensity of precipitation systems. Operational weather forecasting services employ ensemble prediction systems for early flood warning, although the number of ensemble members is limited to a few dozen. From the perspective of probability prediction with minimized sampling errors, increasing the number of ensemble members for flood forecasting is an important research topic. This study investigated the impacts of the number of ensemble members on flood prediction. The precipitation dataset was 100 and 1000-member local ensemble transform Kalman filter (LETKF100 and LETKF1000, respectively) and 21-member operational weather forecast (MEPS). The flood forecasting model was the Japan Meteorological Agency's (JMA) operational flood forecasting system, “Runoff Index Model (RIM)”, which incorporates a tank model and the Manning equation. The case study was the extreme flooding caused by record-breaking rainfall in the Kuma River Basin (1880 km²), Japan, that occurred in July 2020. Using LETKF1000, the RIM successfully forecasted a high flood risk in the flood damaged area, with a probability of 60 % half a day ahead; six hours earlier than the JMA's operational flood forecast. Furthermore, we investigated the number of members required for ensemble flood forecasting. The prediction accuracy for the occurrence of risks during the flood forecasting period was similar between LETKF1000 and LETKF100. Whereas, LETKF1000 has higher prediction accuracy than LETKF100 regarding the timing of flood peaks. When selecting 500 members from the initial 1000 members, results nearly identical to those from the 1000 members were obtained. These results demonstrate that the LETKF1000 has the potential to provide valuable information for facilitating early evacuation.

 Four-dimensional variational data assimilation (4DVar) has been used as widely as ensemble Kalman filters (EnKFs) in meteorology and oceanography. Unlike EnKFs, 4DVar can be applied to strongly nonlinear regimes in data assimilation. A problem with 4DVar is that the cost function may have multiple minima, and that it can be difficult to find the global minimum using a gradient descent method. Quantum annealing can find the global minimum via quadratic unconstrained binary optimization (QUBO). This study proposes a method of searching for the global minimum of the 4DVar cost function by combining a second-order incremental approach and quantum annealing, in which the latter provides guidance on where to explore in state space by minimizing an approximated cost function. This approximated cost function is constructed in low-dimensional space by expanding state variables up to the second order around a basic state. If the global minimum cannot be reached after a couple of updates of the basic state, the 4DVar analysis is replaced by an EnKF analysis in assimilation cycles. Data assimilation experiments using the Lorenz-63 model were conducted as a proof of concept of the proposed method. The results revealed that the proposed method significantly reduced the frequency of falling into local minima, and that the benefit of extending the length of the assimilation window was realized even in strongly nonlinear regimes. Data assimilation experiments in which simulated annealing was adopted instead of quantum annealing showed that quantum annealing exhibited comparable or better performance compared to simulated annealing.

 Numerical weather prediction (NWP) centers around the world operate a variety of NWP models. In addition, recent advances in AI-driven NWP models have further increased the availability of NWP outputs. While this expansion holds the potential to improve forecast accuracy, it raises a critical question: which prediction is the most plausible? If the NWP models have comparable accuracy, it is impossible to determine in advance which one is the best. Traditional approaches, such as ensemble or weighted averaging, combine multiple NWP outputs to produce a single forecast with improved accuracy. However, they often result in meteorologically unrealistic and uninterpretable outputs, such as the splitting of tropical cyclone centers or frontal boundaries into multiple distinct systems.

 To address this issue, we propose DeepMedcast, a deep learning method that generates intermediate forecasts between two or more NWP outputs. Unlike averaging, DeepMedcast provides predictions in which meteorologically significant features—such as the locations of tropical cyclones, extratropical cyclones, fronts, and shear lines—approximately align with the arithmetic mean of the corresponding features predicted by the input NWP models, without distorting meteorological structures. We demonstrate the capability of DeepMedcast through case studies and verification results, showing that it produces realistic and interpretable forecasts with higher accuracy than the input NWP models. By providing plausible intermediate forecasts, DeepMedcast can significantly contribute to the efficiency and standardization of operational forecasting tasks, including general, marine, and aviation forecasts.

 This study reports the correction methods of a newly introduced upper-air radiosonde instrument, “Storm Tracker” (ST), with more than one thousand co-launches of ST and Vaisala RS41-SGP (VS) data in field observations in the Taiwan area during 2016-2022. The co-launches provided more than a million comparable independent observations of wind, pressure, temperature, and humidity (PTU) data. Using the VS measurements as the reference, we use the statistical models, including the cumulative distribution function (CDF) matching method and generalized linear model (GLM), to correct the temperature and moisture fields of the ST sounding. Both approaches yield similar results. With a sounding-by-sounding comparison, the CDF-corrected ST soundings have a 1-K temperature and 7 % relative humidity root mean square difference from the VS soundings. These error differences can be reduced to 0.66-K and 4.61 % below the 700-hPa height. The GPS estimated a 0.05 ms⁻¹ ST wind difference from the VS sounding. The biases of the corrected ST observations are slightly larger than the random errors, which were 0.24 K and 2.21 % in the laboratory and 0.52 K and 2.23 % in the field. The lower atmosphere in a region of complex terrain may have large wind, temperature, and moisture variations. With the relatively low cost, a high proportion of successful launches, and accuracy of wind, temperature, and moisture, ST can complement regular upper-air radiosonde observations for high-resolution observations in the lower troposphere. The high-resolution lower troposphere observation is important for severe weather research in East Asia.