Deep learning-based multi-source precipitation merging for the Tibetan Plateau

Due to its complex and diverse terrain, precipitation gauges in the Tibetan Plateau(TP) are sparse, making it difficult to obtain reliable precipitation data for environmental studies. Data merging is a method that can integrate precipitation data from multiple sources to generate high-precision precipitation data. However, the more commonly used methods, such as regression and machine learning, do not usually consider the local correlation of precipitation, so that the spatial pattern of precipitation cannot be reproduced, while deep learning methods do incorporate spatial correlation. To explore the ability of using deep learning methods in merging precipitation data for the TP, this study compared three methods: a deep learning method—a convolutional neural network(CNN) algorithm, a machine learning method—an artificial neural network(ANN) algorithm, and a statistical method based on Extended Triple Collocation(ETC) in merging precipitation from multiple sources(gauged, grid,satellite and dynamic downscaling) over the TP, as well as their performance for hydrological simulations. Dynamic downscaling data driven by global reanalysis data centered on the TP were introduced in the merging process to better reflect the spatial variability of precipitation. The results show that:(1) in terms of the meteorological metrics, the merged data perform better than the gauge interpolation data. By using data merging, the error between the raw multi-source and gauged precipitation can be reduced, and the precipitation detection capability can be greatly improved;(2) The merged precipitation data also perform well in the hydrological evaluation. The Xin’anjiang(XAJ) model parameter calibration experiments at the source of the Yangtze River(SYR) and the source of the Yellow River(SHR) were repeated 300 times to remove uncertainty in the model parameter results. The median Kling-Gupta Efficiency Coefficients(KGE) of simulated runoff from the merged data of the ANN, CNN and ETC methods for the SYR and the SHR are 0.859, 0.864, 0.838 and 0.835, 0.835, 0.789, respectively. Except for the ETC merging data at the SHR, the performance of other merged data was improved compared to the simulation results of the gauged precipitation(KGE=0.807 at the SYR, KGE=0.828 at the SHR);and(3) In contrast to the machine learning ANN method and the statistical ETC method, the deep learning method, CNN, consistently showed better performance.

Flight dynamics modeling and analysis for a Mars helicopter

Flight dynamics modeling for the Mars helicopter faces great challenges.Aerodynamic modeling of coaxial rotor with high confidence and high computational efficiency is a major difficulty for the field.This paper builds an aerodynamic model of coaxial rotor in the extremely thin Martian atmosphere using the viscous vortex particle method.The aerodynamic forces and flow characteristics of rigid coaxial rotor are computed and analyzed.Meanwhile,a high fidelity aerodynamic surrogate model is built to improve the computational efficiency of the flight dynamics model.Results in this paper reveal that rigid coaxial rotor can bring the Mars helicopter sufficient controllability but result in obvious instability and control couplings in forward flight.This highlights the great differences in flight dynamics characteristics compared with conventional helicopters on Earth.

城乡融合视角下黄河流域镇域经济的空间格局及其影响因素

基于城乡融合视角,通过多种渠道搜集镇域农民人均纯收入数据,分析黄河流域镇域经济的空间格局及其影响因素。具体而言,以农民人均纯收入为基础指标,以国镇比(全国城镇居民的平均收入与乡镇农民人均纯收入之比)、省镇比(省域城镇居民的平均收入与乡镇农民人均纯收入之比)、市镇比(市域城镇居民的平均收入与乡镇农民人均纯收入之比)、县镇比(县域城镇居民的平均收入与乡镇农民人均纯收入之比)为城乡融合下的镇域经济测度指标,分析黄河流域7098个镇域单元组成的经济空间格局,进而探讨其影响因素。结果表明:(1)从国镇比下的城乡融合水平看,大多数地区的城乡收入差距较大,表明流域整体的城乡融合水平普遍较低。从空间分布看,低融合水平区主要分布在内蒙古以南、河南以西的广大地区并形成连绵集聚片区,高融合水平区主要分布在山东大部、河南中部及西部、内蒙古局部。省镇比、市镇比、县镇比下的空间分异格局与国镇比保持较高的一致性,不同的是随着参照指标的进一步缩小,各水平区集聚的程度有所弱化。(2)从空间关联格局看,国镇比、省镇比、县镇比下城乡融合发展水平的空间集聚效应明显,主要以显著高-高(HH)区、显著低-低(LL)区为关联类型。国镇比下的显著LL区分布在山东大部、河南中部及北部、内蒙古局部,而显著HH区集中出现在山西外围、甘肃、青海南部。(3)基于最小二乘法、空间滞后模型、空间误差模型等定量分析发现,整体分异格局主要受二三产业从业人员、建镇区人口占比、人均工业产值、二三产业从业人员占比的影响最为显著,工业生产总值、建镇区面积占比仅起基础性影响。

Construction of ground-state preserving sparse lattice models for predictive materials simulations

First-principles based cluster expansion models are the dominant approach in ab initio thermodynamics of crystalline mixtures enabling the prediction of phase diagrams and novel ground states.However,despite recent advances,the construction of accurate models still requires a careful and time-consuming manual parameter tuning process for ground-state preservation,since this property is not guaranteed by default.In this paper,we present a systematic and mathematically sound method to obtain cluster expansion models that are guaranteed to preserve the ground states of their reference data.The method builds on the recently introduced compressive sensing paradigm for cluster expansion and employs quadratic programming to impose constraints on the model parameters.The robustness of our methodology is illustrated for two lithium transition metal oxides with relevance for Li-ion battery cathodes,i.e.,Li_(2x)Fe_(2(1−x))O_(2) and Li_(2x)Ti_(2(1−x))O_(2),for which the construction of cluster expansion models with compressive sensing alone has proven to be challenging.We demonstrate that our method not only guarantees ground-state preservation on the set of reference structures used for the model construction,but also show that out-of-sample ground-state preservation up to relatively large supercell size is achievable through a rapidly converging iterative refinement.This method provides a general tool for building robust,compressed and constrained physical models with predictive power.