Search

Arabian Sea Mid-Tropospheric Cyclones (MTCs), which are responsible for extreme rainfall events in western India, often coincide with monsoon low-pressure systems (LPS) over the Bay of Bengal. However, the role of Bay of Bengal LPSs in the formation of Arabian Sea MTCs remains unclear. This study utilizes the Weather Research and Forecasting Model (WRF) to investigate the atmospheric connection between the two basins. By introducing a balanced bogus vortex over the Bay of Bengal, cyclonic systems are induced over the Arabian Sea in the majority of ensemble members, exhibiting characteristics consistent with observations. As the Bay of Bengal vortex moves westward, the middle tropospheric trough deepens, horizontal wind shear increases, the low-level Arabian Sea stable inversion layer weakens, and the middle troposphere moisture content increases over western India and the northeast Arabian Sea. Subsequently, MTC genesis occurs along the western edge of the trough within 2-4 days of model integration over the northeast Arabian Sea. Vorticity budget analysis highlights the critical role of vorticity advection and tilting during the initial 24 hours of MTC genesis, while vortex stretching becomes the dominant vorticity source during rapid intensification. To further substantiate these findings, a mechanism denial experiment is conducted using a real-world instance of coexistent Arabian Sea MTC and Bay of Bengal LPS, which was replicated in the model. In the experiment, conditions unfavorable for LPS genesis are created by cooling and drying the Bay of Bengal. The results demonstrate that when the Bay of Bengal LPS does not develop or intensify, the Arabian Sea MTC fails to form. This study presents compelling evidence for the significant influence of Bay of Bengal low-pressure systems on the formation of severe weather-inducing MTCs over the Arabian Sea and western India., Comment: 32 pages 18 figures

Deep Learning (DL) based downscaling has become a popular tool in earth sciences recently. Increasingly, different DL approaches are being adopted to downscale coarser precipitation data and generate more accurate and reliable estimates at local (~few km or even smaller) scales. Despite several studies adopting dynamical or statistical downscaling of precipitation, the accuracy is limited by the availability of ground truth. A key challenge to gauge the accuracy of such methods is to compare the downscaled data to point-scale observations which are often unavailable at such small scales. In this work, we carry out the DL-based downscaling to estimate the local precipitation data from the India Meteorological Department (IMD), which was created by approximating the value from station location to a grid point. To test the efficacy of different DL approaches, we apply four different methods of downscaling and evaluate their performance. The considered approaches are (i) Deep Statistical Downscaling (DeepSD), augmented Convolutional Long Short Term Memory (ConvLSTM), fully convolutional network (U-NET), and Super-Resolution Generative Adversarial Network (SR-GAN). A custom VGG network, used in the SR-GAN, is developed in this work using precipitation data. The results indicate that SR-GAN is the best method for precipitation data downscaling. The downscaled data is validated with precipitation values at IMD station. This DL method offers a promising alternative to statistical downscaling.

The formation of Middle Tropospheric Cyclones (MTCs) that are responsible for a large portion of annual precipitation and extreme rainfall events over western India is studied using an unsupervised machine learning algorithm and cyclone tracking. Both approaches reveal four dominant weather patterns that lead to the genesis of these systems; specifically, re-intensification of westward moving synoptic systems from Bay of Bengal (Type 1, 51%), in-situ formation with a coexisting cyclonic system over the Bay of Bengal that precedes (Type 2a, 31%) or follows (Type 2b, 10%) genesis in the Arabian Sea, and finally in-situ genesis within a northwestward propagating cyclonic anomaly from the south Bay of Bengal (Type 2c, 8%). Thus, a large fraction of rainy middle tropospheric synoptic systems in this region form in association with cyclonic activity in the Bay of Bengal. The four variants identified also show a marked dependence on large-scale environmental features with Type 1 and Type 2a formation primarily occurring in phases 4 and 5, and Type 2b and Type 2c in phases 3 and 4 of the Boreal Summer Intraseasonal Oscillation. Further, while in-situ formation with a Bay of Bengal cyclonic anomaly (Type 2a and 2b) mostly occurs in June, downstream development is more likely in the core of the monsoon season. Out of all categories, Type 2a is associated with the highest rain rate (60 mm/day) and points towards the dynamical interaction between a low pressure system over the Bay of Bengal and the development of MTCs over western India and the northeast Arabian Sea. This classification, identification of precursors, connection with cyclonic activity over the Bay of Bengal and dependence on large-scale environment provides an avenue for better understanding and prediction of rain-bearing MTCs over western India.

This survey focuses on the current problems in Earth systems science where machine learning algorithms can be applied. It provides an overview of previous work, ongoing work at the Ministry of Earth Sciences, Gov. of India, and future applications of ML algorithms to some significant earth science problems. We provide a comparison of previous work with this survey, a mind map of multidimensional areas related to machine learning and a Gartner's hype cycle for machine learning in Earth system science (ESS). We mainly focus on the critical components in Earth Sciences, including atmospheric, Ocean, Seismology, and biosphere, and cover AI/ML applications to statistical downscaling and forecasting problems.

Accurate short range weather forecasting has significant implications for various sectors. Machine learning based approaches, e.g., deep learning, have gained popularity in this domain where the existing numerical weather prediction (NWP) models still have modest skill after a few days. Here we use a ConvLSTM network to develop a deep learning model for precipitation forecasting. The crux of the idea is to develop a forecasting model which involves convolution based feature selection and uses long term memory in the meteorological fields in conjunction with gradient based learning algorithm. Prior to using the input data, we explore various techniques to overcome dataset difficulties. We follow a strategic approach to deal with missing values and discuss the models fidelity to capture realistic precipitation. The model resolution used is (25 km). A comparison between 5 years of predicted data and corresponding observational records for 2 days lead time forecast show correlation coefficients of 0.67 and 0.42 for lead day 1 and 2 respectively. The patterns indicate higher correlation over the Western Ghats and Monsoon trough region (0.8 and 0.6 for lead day 1 and 2 respectively). Further, the model performance is evaluated based on skill scores, Mean Square Error, correlation coefficient and ROC curves. This study demonstrates that the adopted deep learning approach based only on a single precipitation variable, has a reasonable skill in the short range. Incorporating multivariable based deep learning has the potential to match or even better the short range precipitation forecasts based on the state of the art NWP models., Comment: 14 pages, 14 figures. The manuscript is under review with journal 'Transactions on Geoscience and Remote Sensing'

The formation of precipitation in state-of-the-art weather and climate models is an important process. The understanding of its relationship with other variables can lead to endless benefits, particularly for the world's monsoon regions dependent on rainfall as a support for livelihood. Various factors play a crucial role in the formation of rainfall, and those physical processes are leading to significant biases in the operational weather forecasts. We use the UNET architecture of a deep convolutional neural network with residual learning as a proof of concept to learn global data-driven models of precipitation. The models are trained on reanalysis datasets projected on the cubed-sphere projection to minimize errors due to spherical distortion. The results are compared with the operational dynamical model used by the India Meteorological Department. The theoretical deep learning-based model shows doubling of the grid point, as well as area averaged skill measured in Pearson correlation coefficients relative to operational system. This study is a proof-of-concept showing that residual learning-based UNET can unravel physical relationships to target precipitation, and those physical constraints can be used in the dynamical operational models towards improved precipitation forecasts. Our results pave the way for the development of online, hybrid models in the future.

Mid-Tropospheric Cyclones (MTCs) are moist synoptic systems with distinct mid tropospheric vorticity maxima and weak signatures in the lower troposphere. Composites and statistics of MTCs over the tropics are constructed and compared with monsoon lows and depressions (together, lower troposphere cyclones; LTCs). We begin with South Asia, where tracking reveals that MTCs change character during their life, i.e., their track is composed of MTC and LTC phases. The highest MTC-phase density and least motion is over the Arabian Sea, followed by the Bay of Bengal and South China Sea. A MTC-phase composite shows an east-west tilted warm above deep cold-core temperature anomaly with maximum vorticity at 600 hPa. While the LTC-phase shows a shallow cold-core below 800 hPa and a warm upright temperature anomaly with lower tropospheric vorticity maximum. Apart from South Asia, systems with similar morphology are observed over the west and central Africa, east and west Pacific in boreal summer. In boreal winter, regions that support MTCs include northern Australia, the southern Indian Ocean, and South Africa. Relatively, the MTC fraction is higher equatorward, where there is a cross-equatorial low-level jet that advects oppositely signed vorticity. Whereas the LTCs are more prevalent further poleward. Finally, a histogram of differential vorticity (difference between middle and lower levels) versus the height of peak vorticity for cyclonic centers is bimodal. One peak, around 600 hPa, corresponds to MTCs while the second, around 900 hPa, comes from LTCs. Thus, moist cyclonic systems in the tropics have a natural tendency to reside in either the MTC or LTC category preferentially

Deep and shallow convection calculations occupy significant times in atmosphere models. These calculations also present significant load imbalances due to varying cloud covers over different regions of the grid. In this work, we accelerate these calculations on Intel{\textregistered} Xeon Phi{\texttrademark} Coprocessor Systems. By employing dynamic scheduling in OpenMP, we demonstrate large reductions in load imbalance and about 10% increase in speedups. By careful categorization of data as private, firstprivate and shared, we minimize data copying overheads for the coprocessors. We identify regions of false sharing among threads and eliminate them by loop rearrangements. We also employ proportional partitioning of independent column computations across both the CPU and coprocessor cores based on the performance ratio of the computations on the heterogeneous resources. These techniques along with various vectorization strategies resulted in about 30% improvement in convection calculations.