Long-term changes in North American monsoon (NAM) precipitation intensity in the southwestern United States are evaluated through the use of convective-permitting model simulations of objectively identified severe weather events during “historical past” (1950–70) and “present day” (1991–2010) periods. Severe weather events are the days on which the highest atmospheric instability and moisture occur within a long-term regional climate simulation. Simulations of severe weather event days are performed with convective-permitting (2.5 km) grid spacing, and these simulations are compared with available observed precipitation data to evaluate the model performance and to verify any statistically significant model-simulated trends in precipitation. Statistical evaluation of precipitation extremes is performed using a peaks-over-threshold approach with a generalized Pareto distribution. A statistically significant long-term increase in atmospheric moisture and instability is associated with an increase in extreme monsoon precipitation in observations and simulations of severe weather events, corresponding to similar behavior in station-based precipitation observations in the Southwest. Precipitation is becoming more intense within the context of the diurnal cycle of convection. The largest modeled increases in extreme-event precipitation occur in central and southwestern Arizona, where mesoscale convective systems account for a majority of monsoon precipitation and where relatively large modeled increases in precipitable water occur. Therefore, it is concluded that a more favorable thermodynamic environment in the southwestern United States is facilitating stronger organized monsoon convection during at least the last 20 years.

Capsule Summary An integrated, high resolution, data-driven regional modeling system has been recently developed for the Red Sea region and is being used for research and various environmental applications.

We present a method to formulate algorithm discovery as program search, and apply it to discover optimization algorithms for deep neural network training. We leverage efficient search techniques to explore an infinite and sparse program space. To bridge the large generalization gap between proxy and target tasks, we also introduce program selection and simplification strategies. Our method discovers a simple and effective optimization algorithm, Lion (EvoLved Sign Momentum). It is more memory-efficient than Adam as it only keeps track of the momentum. Different from adaptive optimizers, its update has the same magnitude for each parameter calculated through the sign operation. We compare Lion with widely used optimizers, such as Adam and Adafactor, for training a variety of models on different tasks. On image classification, Lion boosts the accuracy of ViT by up to 2% on ImageNet and saves up to 5x the pre-training compute on JFT. On vision-language contrastive learning, we achieve 88.3% zero-shot and 91.1% fine-tuning accuracy on ImageNet, surpassing the previous best results by 2% and 0.1%, respectively. On diffusion models, Lion outperforms Adam by achieving a better FID score and reducing the training compute by up to 2.3x. For autoregressive, masked language modeling, and fine-tuning, Lion exhibits a similar or better performance compared to Adam. Our analysis of Lion reveals that its performance gain grows with the training batch size. It also requires a smaller learning rate than Adam due to the larger norm of the update produced by the sign function. Additionally, we examine the limitations of Lion and identify scenarios where its improvements are small or not statistically significant. The implementation of Lion is publicly available. 1 * Work done as a student researcher at Google Brain. † Work done while at Google.

This study analyses the connection between extreme rainfall events in Jeddah, Saudi Arabia, and synoptic-scale weather patterns over the Arabian Peninsula.Mean rainfall follows a decreasing trend; however, the number of rainy days has increased. Interestingly, extreme rainfall is becoming less frequent but shows an increased intensity. Here we utilize self-organizing maps (SOMs) to identify the weather patterns of the most intense rainy days and the synoptic systems causing extreme rainfall in the Jeddah region. Three main weather patterns that cause heavy rainfall events over Jeddah during the cooler months (November-April) are identified, all reflect tropical-extratropical interactions. Extreme events in the early period (1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998) are characterized by a stronger tropical influence and local precipitation patterns, while a stronger extratropical forcing and higher extreme rainfall amounts are spotted in the late period (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). Our results suggest that in recent decades, the mechanism causing extreme rainfall over the city of Jeddah has shifted toward a weather regime with stronger extratropical influence.

24 Transient inverted troughs (IVs) are a trigger for severe weather during the North 25 American Monsoon (NAM) in the Southwest Contiguous US (CONUS) and northwest Mexico. 26 These upper tropospheric disturbances enhance the synoptic scale and mesoscale environment 27 for organized convection, increasing the chances for microbursts, straight-line winds, blowing 28 dust, and flash flooding. This work considers changes in the track density climatology of IVs 29 between 1951 and 2010. IVs are tracked as potential vorticity (PV) anomalies on the 250 hPa 30 surface from a regional climate model that dynamically downscales the NCEP/NCAR Reanalysis 31 1. Late in the NAM season, a significant increase in IV track density over the 60-year period is 32 observed over southern California and western Arizona, coupled with a slight decrease over 33 northwest Mexico. Changes in precipitation are evaluated on days when an IV is observed and 34 days without an IV, using high-resolution model-simulated precipitation estimates and CPC 35 gridded precipitation observations. Due to changes in the spatial distribution of IVs during the 36 1951-2010 analysis period, which are associated with a strengthening of the monsoon ridge, we 37 suggest that IVs have played a lesser role in the initiation and organization of monsoon 38 convection in the southwest CONUS during recent warm seasons. 39 3

The city of Jeddah, Saudi Arabia, is characterized by a hot and arid desert climate. On occasion, however, extreme precipitation events have led to flooding that caused extensive damage to human life and infrastructure. This study investigates the effect of incorporating an urban canopy model and urban land cover when simulating severe weather events over Jeddah using the Weather Research and Forecasting (WRF) Model at a convective-permitting scale (1.5-km resolution). Two experiments were conducted for 10 heavy rainfall events associated with the dominant large-scale patterns favoring convection over Jeddah: (i) an ''urban'' experiment that included the urban canopy model and modern-day land cover and (ii) a ''desert'' experiment that replaced the city area with its presettlement, natural land cover. The results suggest that urbanization plays an important role in modifying rainfall around city area. The urban experiment enhances the amount of rainfall by 26% on average over the Jeddah city area relative to the desert experiment in these extreme events. The changes in model-simulated precipitation are primarily tied to a nocturnal heat-island effect that modifies the planetary boundary layer and atmospheric instability of the convective events.

Tropical cyclones are projected to be more intense, smaller in size, and to move faster • Landfalling locations shift poleward and further west by the end of the 21 st century.• Wind and rain impacts increase in the northern Southeast Asia, extend further inland.

Abstract:A commonly noted problem in the simulation of warm season convection in the North American monsoon region has been the inability of atmospheric models at the meso-β scales (10 s to 100 s of kilometers) to simulate organized convection, principally mesoscale convective systems. With the use of convective parameterization, high precipitation biases in model simulations are typically observed over the peaks of mountain ranges. To address this issue, the Kain-Fritsch (KF) cumulus parameterization scheme has been modified with new diagnostic equations to compute the updraft velocity, the convective available potential energy closure assumption, and the convective trigger function. The scheme has been adapted for use in the Weather Research and Forecasting (WRF). A numerical weather prediction-type simulation is conducted for the North American Monsoon Experiment Intensive Observing Period 2 and a regional climate simulation is performed, by dynamically downscaling. In both of these applications, there are notable improvements in the WRF model-simulated precipitation due to the better representation of organized, propagating convection. The use of the modified KF scheme for atmospheric model simulations may provide a more computationally economical alternative to improve the representation of organized convection, as compared to convective-permitting simulations at the kilometer scale or a super-parameterization approach.