【水科學(xué)講壇】第25講：黃河勘測(cè)規(guī)劃設(shè)計(jì)研究院有限公司董事長(zhǎng)、

黨委書記張金良教高學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

首先介紹黃河調(diào)水調(diào)沙的基本情況，對(duì)調(diào)水調(diào)沙的緣由、定義、指導(dǎo)思想和總體目標(biāo)以及調(diào)水調(diào)沙模式進(jìn)行講解。其次報(bào)告黃河調(diào)水調(diào)沙方案制定關(guān)鍵技術(shù)，黃河調(diào)水調(diào)沙21年以來取得的效果以及當(dāng)前調(diào)水調(diào)沙存在的問題。最后展望黃河調(diào)水調(diào)沙前景。

愛爾蘭都柏林圣三一大學(xué)殷碩教授學(xué)術(shù)報(bào)告通知

報(bào)告簡(jiǎn)介：

高性能金屬構(gòu)件是航空、航天、交通、能源、軍工等現(xiàn)代工業(yè)的基石，且高端裝備的服役性能很大程度上取決于構(gòu)件的高性能。金屬增材制造技術(shù)是近年來興起的一種尖端制造技術(shù)，其獨(dú)特的零件生產(chǎn)過程與傳統(tǒng)的打印過程十分類似，因此又被稱為金屬3D打印技術(shù)。增材制造技術(shù)的誕生，為生產(chǎn)可定制化、高性能、結(jié)構(gòu)復(fù)雜構(gòu)件的制造提供了完美的解決方案。增材制造過程幾乎不受零件復(fù)雜程度的影響，對(duì)于單件小批量生產(chǎn)和具有較高幾何復(fù)雜性的零件，增材制造具有顯著的競(jìng)爭(zhēng)優(yōu)勢(shì)。傳統(tǒng)的零件的制造受到零件本身復(fù)雜性的限制，往往在設(shè)計(jì)過程中并未完全實(shí)現(xiàn)功能優(yōu)先的設(shè)計(jì)，結(jié)構(gòu)上有很多冗余，浪費(fèi)材料。增材制造可以通過對(duì)結(jié)構(gòu)進(jìn)行拓?fù)鋬?yōu)化設(shè)計(jì)，實(shí)現(xiàn)構(gòu)件的結(jié)構(gòu)功能一體化、輕量化、高強(qiáng)度、耐極端載荷、超強(qiáng)散熱、復(fù)雜內(nèi)部流道、功能梯度等功能，極大的滿足現(xiàn)代工業(yè)對(duì)難加工金屬構(gòu)件短周期、高精度、高性能制造的重大需求。

美國(guó)勞倫斯利弗摩爾國(guó)家實(shí)驗(yàn)室潘寶祥研究員學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

The earth climate system is featured by the chaotic geophysical fluid dynamics and the complicated interaction among various subsystems. This chaoticity and complexity raise the need to disentangle internal climate variability noise, external forcing, and model formulation deficiencies to answer climate-relevant questions, such as weather variability and climate adaptation. This talk discusses a self-supervised adversarial learning method for merging climate models and climate observations to disentangle different sources of uncertainties in climate prediction, therefore diagnosing, correcting, and improving our modeling of the earth system. We discuss the limitations of supervised (deep) learning in climate applications, and highlight the necessity of shifting toward novel learning paradigms to realize the power of modern machine learning techniques. We believe by replacing human model diagnosis experts with tireless machine "nitpickers" and "cleaners", we may soon reach a true "digital twin" of the earth climate system.

美國(guó)勞倫斯利弗摩爾國(guó)家實(shí)驗(yàn)室潘寶祥研究員學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

Deep neural networks, operate with large, high quality data, which together with proper computation resources, motivate an ongoing paradigm shift in scientific discovery and engineering practices. This talk is for domain experts who are interested in deep learning, and would like to apply deep learning to make predictions, explanations, or quickly explore research ideas. I will briefly review the technical history of deep learning, discuss six mindsets underpinning the data-driven modeling paradigm, use several application studies to illustrate the potential pitfalls and benefits for applying deep neural networks in specific problems. I encourage an open discussion of research frontiers, given that individuals could hardly follow the fast progress in this field. Finally, I will close the talk by providing useful resources for learning and tracking advances in this field.

美國(guó)勞倫斯利弗摩爾國(guó)家實(shí)驗(yàn)室潘寶祥研究員學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

The path toward realizing the potential of seasonal forecasting and its socioeconomic benefits relies on improving general circulation model (GCM) based dynamical forecast systems. To improve dynamical seasonal forecasts, it is crucial to set up forecast benchmarks, and clarify forecast limitations posed by model initialization errors, formulation deficiencies, and internal climate variability. With huge costs in generating large forecast ensembles, and limited observations for forecast verification, the seasonal forecast benchmarking and diagnosing task proves challenging. Here, we develop a probabilistic deep learning-based statistical forecast methodology, drawing on a wealth of climate simulations to enhance seasonal forecast capability and forecast diagnosis. By explicitly modeling the internal climate variability and GCM formulation differences, the proposed Conditional Generative Forecasting (CGF) methodology enables bypassing crucial barriers in dynamical forecast, and offers a top-down viewpoint to examine how complicated GCMs encode the seasonal predictability information. We apply the CGF methodology for global seasonal forecast of precipitation and 2 m air temperature, based on a unique data set consisting 52,201 years of climate simulation. Results show that the CGF methodology can faithfully represent the seasonal predictability information encoded in GCMs. We successfully apply this learned relationship in real-world seasonal forecast, achieving competitive performance compared to dynamical forecasts. Using this CGF as benchmark, we reveal the impact of insufficient forecast spread sampling that limits the skill of the considered dynamical forecast system. Finally, we introduce different strategies for composing ensembles using the CGF methodology, highlighting the potential for leveraging the strengths of multiple GCMs to achieve advantgeous seasonal forecast.

華東師范大學(xué)孫勛研究員學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

Corn is the 1st economic field crop in the world, whose price stability guarantees sustainable and equitable food security. Most previous farm commodity price prediction model only focus on detecting the autoregression of historical transaction, while ignoring other factors. For agricultural commodities, different climate condition leads to different harvest situation, thus bringing volatility to prices. Therefore, it is reasonable to propose a method based on climate indices to measure the degree of their influence on price fluctuation.

A multiple regression model is developed for predicting corn price movements at the nation level. The June-September season is selected to target the essential growing stages of corn which are especially sensitive to drought, high temperature stress and water stress. In order to describe the movements of price, the price difference between June and September is chosen as the dependent variable. Daily climate data are obtained from PRISM which integrates both satellite and meteorological station observation data, and monthly price data are sourced from USDA. 39-year trend from 1981-2019 is explored to construct a predictive model. The results show that the accuracy of predicting up and down of price is 85%. Specifically, temperature in July has an identifiable effect on price movements which explains 36.99% price variation. These results imply that during the key growing period, climate indices occupy an important position on improving crop price forecast ability.

中國(guó)船舶科學(xué)研究中心彭曉星研究員學(xué)術(shù)報(bào)告

報(bào)告簡(jiǎn)介：

空化的尺度效應(yīng)一般是指模型空化現(xiàn)象與原型的偏差，空化尺度效應(yīng)的存在極大地阻礙了空化研究的工程應(yīng)用。本報(bào)告首先介紹空化尺度效應(yīng)的一般概念，從空化機(jī)理角度分析空化尺度效應(yīng)的來源。然后從空化起始、云空化、空蝕等幾個(gè)方面討論空化尺度效應(yīng)的修正方法和未來研究展望。