The goal of precipitation nowcasting is to predict the future rainfall intensity in a local region over a relatively short period of time. Very few previous studies have examined this crucial and challenging weather forecasting problem from the machine learning perspective. In this paper, we formulate precipitation nowcasting as a spatiotemporal sequence forecasting problem in which both the input and the prediction target are spatiotemporal sequences. By extending the fully connected LSTM (FC-LSTM) to have convolutional structures in both the input-to-state and state-to-state transitions, we propose the convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model for the precipitation nowcasting problem. Experiments show that our ConvLSTM network captures spatiotemporal correlations better and consistently outperforms FC-LSTM and the state-of-theart operational ROVER algorithm for precipitation nowcasting.

In this study, synoptic situations associated with extreme hourly precipitation over China are investigated using rain gauge data, weather maps, and composite radar reflectivity data. Seasonal variations of hourly precipitation (>0.1 mm h−1) suggest complicated regional features in the occurrence frequency and intensity of rainfall. The 99.9th percentile is thus used as the threshold to define the extreme hourly rainfall for each station. The extreme rainfall is the most intense over the south coastal areas and the North China Plain. About 77% of the extreme rainfall records occur in summer with a peak in July (30.4%) during 1981–2013. Nearly 5800 extreme hourly rainfall records in 2011–15 are classified into four types according to the synoptic situations under which they occur: the tropical cyclone (TC), surface front, vortex/shear line, and weak-synoptic forcing. They contribute 8.0%, 13.9%, 39.1%, and 39.0%, respectively, to the total occurrence and present distinctive characteristics in regional distribution and seasonal or diurnal variations. The TC type occurs most frequently along the coasts and decreases progressively toward inland China; the frontal type is distributed relatively evenly east of 104°E; the vortex/shear line type shows a prominent center over the Sichuan basin with two high-frequency bands extending from the center southeastward and northeastward, respectively; and the weak-synoptic type occurs more frequently in southeast, southwest, and northern China, and in the easternmost area of northeast China. Occurrences of the weak-synoptic type have comparable contributions from mesoscale convective systems and smaller-scale storms with notable differences in their preferred locations.

Hong Kong Observatory has been operating an in-house developed rainfall nowcasting system called "Short-range Warning of Intense Rainstorms in Localized Systems (SWIRLS)" to support rainstorm warning and rainfall nowcasting services. A crucial step in rainfall nowcasting is the tracking of radar echoes to generate motion fields for extrapolation of rainfall areas in the following few hours. SWIRLS adopted a correlation-based method in its first operational version in 1999, which was subsequently replaced by optical flow algorithm in 2010 and further enhanced in 2013. The latest optical flow algorithm employs a transformation function to enhance a selected range of reflectivity for feature tracking. It also adopts variational optical flow computation that takes advantage of the Horn-Schunck approach and the Lucas-Kanade method. This paper details the three radar echo tracking algorithms, examines their performances in several significant rainstorm cases and summaries verification results of multi-year performances. The limitations of the current approach are discussed. Developments underway along with future research areas are also presented.

No abstract

Resist heating is one of the major factors that causes feature size variation and pattern displacement in photomask fabrication. A number of models have been published to predict the rise in temperature during resist heating, but no transient temperature experimental results are available to verify those models. We have fabricated thin film gold/nickel thermocouples with junction areas as small as 100 nm 2 on silicon and 500 nm 2 on quartz. Microsecond scale transient resist heating measurements were obtained using these thermocouples. Irradiation by a 15 keV, 150 nA electron beam of 1.7 m radius for 100 s yielded temperature rises at the resist bottom surface of approximately 62 K on quartz substrates and of 18 K on silicon substrates. Simulation results using a multilayer Green's function model are in reasonable agreement with these experimental data for smaller temperature rises but tend to overestimate by about 10% for larger rises in temperature. In our experiments, a 100 ms exposure is equivalent to a dose of 150 C/cm 2 . Under the same electron beam conditions electron dosages of 5 and 15 C/cm 2 result in temperature rises of 25 and 40 K, respectively, on quartz substrates.