Application of the relevance vector machine to canal flow prediction in the Sevier River Basin

This work addresses management of water for irrigation in arid regions where significant delays between the time of order and the time of delivery present major difficulties. Motivated by improvements to water management that will be facilitated by an ability to predict water demand, it employs a data-driven approach to developing canal flow prediction models using the relevance vector machine (RVM), a probabilistic kernel-based learning machine. A search is performed across model attributes including input set, kernel scale parameter and model update scheme for models providing superior prediction capability using the RVM. Models are developed for two canals in the Sevier River Basin of southern Utah for prediction horizons of up to 5 days.     

移动式绿篱修剪机设计

针对绿篱修剪机的发展现状,设计开发了一种小型自走式绿篱修剪机,包括动力系统设计和行走装置设计、修剪装置设计,并且利用三维建模软件建立了移动式绿篱修剪机的三维模型。所设计的移动式绿篱修剪机在所查阅的文献中尚未发现有相同机型的报道,属原创型产品。相比现在小型绿篱修剪机以手持式为主的状况,该绿篱修剪机能够有效降低绿篱修剪的劳动强度,提高修剪质量,有比较广阔的市场前景。     

面向加工的轴流式水轮机叶片数字化建模方法研究

对轴流式水轮机叶片的数字化建模方法进行研究。从木模图提供的原始数据出发,结合叶片各组成曲面的型面特点和加工工艺要求,提出了一种用"类似流线"的描述方式来定义叶片的方法,并利用Unigraphics NX(软件构建了叶片空间曲面。叶片表面的光滑性检查和合理性检验结果验证了该方法的可行性,完整的叶片模型为叶片后续的数控加工奠定了基础。     

基于深度卷积神经网络的小麦赤霉病高光谱病症点分类方法

为快速、高效地利用高光谱成像技术诊断小麦赤霉病病症,分析了卷积层结构与光谱病症特征的关联性,并重点研究了高光谱的像元分类建模方法。首先,基于深度卷积神经网络的2种典型结构,构建了不同深度的卷积神经网络,比较了小麦赤霉病高光谱数据点集的训练和测试结果。结果显示:Visual Geometry Group(VGG)结构随着网络深度的增加,模型损失值不断下降;残差神经网络(ResNet)结构随着深度增加,损失值没有明显降低,说明ResNet网络的深度与模型性能无关。从测试集评测模型泛化性可知,具有4个基础单元模块的22层VGG网络在所有深度卷积模型中最优,其建模和验证准确率远高于传统的支持向量机(SVM),分别为0.846和0.843,测试集准确率为0.742。以VGG为基础单元构建的深度神经网络,能有效提取小麦赤霉病病症的高光谱特征。研究结果可为大尺度小麦赤霉病的智能成像诊断提供理论基础。     

Cardinal temperatures variability within a tropical japonica rice diversity panel

Air temperature is one of the most critical climatic factors controlling rice growth, development, and production in current and future climatic scenarii predicting increasingly frequent situations of extreme and/or fluctuating temperatures. With its large spectrum of geographical origins and cropping areas, one can credit tropical japonica rice subspecies of a probable genetic diversity of its response to air temperature, which is of major interest for the breeding of better adapted rice varieties. A panel of 195 rice accessions (175 japonica plus 20 reference cultivars) was studied in controlled environment to estimate cardinal (base, optimum, and maximum) temperatures based on the monitoring of the elongation rate (LERmax) of the sixth leaf on the main stem in response to six fixed thermal treatments ranging from 16 to 35 °C. A dedicated statistical framework was elaborated for estimating LERmax, cardinal temperature and related uncertainties. Developed statistical framework enhanced the precision of cardinal temperatures estimated compared to previously reported methods, especially for base temperature. Maximum temperature was trickier to estimate and will require further studies. A significant genotypic variability for base and optimal temperature was pointed out, suggesting tropical japonica subspecies represents a relevant genetic pool to breed for rice genotypes adapted to various thermal situations. These results also suggested that using genotype-dependent cardinal temperature values should enhance the way crop growth models account for genotype?×?environment interactions hence their predictive value in current and future climatic conditions.     

Site index models for Calabrian pine in the central Mediterranean region of Turkey

Stem analysis data of 432 trees were obtained from even-aged, pure natural stands of Calabrian pine in the central Mediterranean Region of Turkey. Eight dynamic site equations derived with the Generalized Algebraic Difference Approach (GADA) were compared, based on autoregressive analysis and a thorough evaluation of the goodness of fit. We used generalized nonlinear least squares methods for model fitting. The adjusted coefficients of determination (0.9825–0.9842), root-mean-square errors (0.8004–0.8435 m), and Akaike’s information criterion differences (0–145) indicated a good fit of the eight site index equations. The Hossfeld equation (M3) provided the best result. The Durbin-Watson test statistic did not reveal an autocorrelation issue while the Hossfeld equation provided a satisfactory solution to the serial correlation problem in stem analysis data as time series using autoregressive modeling. This study presents new site index models for Calabrian pine forests in the central Mediterranean region of Turkey where it is the most important commercial tree species. The site index equation, based on the Hossfeld model is recommended for height growth prediction and site classification of Calabrian pine stands in the central Mediterranean region of Turkey, providing a new basis for growth prediction and yield estimation in these important forest ecosystems.     

基于太赫兹成像技术的大豆叶片水分含量测定

为了实现叶片水分含量的快速、精准检测,提出一种基于太赫兹成像技术的大豆叶片水分含量测定方法。利用太赫兹光谱成像系统获取96份大豆叶片太赫兹图像,采用干燥法测量叶片含水率,通过主成分分析(PCA)提取出水分敏感特征波段0.557、1.098、1.163 THz,对这3个特征波段下的叶片图像采用自适应阈值分割法,将其分为叶脉图像与叶肉图像,分别求取各自的图像灰度特征,并分为叶片特征组(G1)、叶脉特征组(G2)和叶肉特征组(G3)。分别采用多元线性回归(MLR)、反向传播神经网络(BP-ANN)和最小二乘支持向量机(LS-SVM)算法,以上述3个特征组作为输入,构建出9种大豆叶片水分预测模型。对比分析各模型性能,发现基于G3的LS-SVM模型预测结果最好,校正集和预测集的决定系数和均方根误差分别为0.967 8、0.963 2,0.057 8、0.046 5。试验结果表明,利用太赫兹成像技术来检测叶片中的水分含量具有非常高的预测精度,为作物叶片水分含量测定提供了一种行之有效的检测手段。     

Suitability of common models to estimate hydrology and diffuse water pollution in North-eastern German lowland catchments with intensive agricultural land use

Various process-based models are extensively being used to analyze and forecast catchment hydrology and water quality. However, it is always important to select the appropriate hydrological and water quality modeling tools to predict and analyze the watershed and also consider their strengths and weaknesses. Different factors such as data availability, hydrological, hydraulic, and water quality processes and their desired level of complexity are crucial for selecting a plausible modeling tool. This review is focused on suitable model selection with a focus on desired hydrological, hydraulic and water quality processes (nitrogen fate and transport in surface, subsurface and groundwater bodies) by keeping in view the typical lowland catchments with intensive agricultural land use, higher groundwater tables, and decreased retention times due to the provision of artificial drainage. In this study, four different physically based, partially and fully distributed integrated water modeling tools, SWAT (soil and water assessment tool), SWIM (soil and water integrated model), HSPF (hydrological simulation program– FORTRAN) and a combination of tools from DHI (MIKE SHE coupled with MIKE 11 and ECO Lab), have been reviewed particularly for the Tollense River catchment located in North-eastern Germany. DHI combined tools and SWAT were more suitable for simulating the desired hydrological processes, but in the case of river hydraulics and water quality, the DHI family of tools has an edge due to their integrated coupling between MIKE SHE, MIKE 11 and ECO Lab. In case of SWAT, it needs to be coupled with another tool to model the hydraulics in the Tollense River as SWAT does not include backwater effects and provision of control structures. However, both SWAT and DHI tools are more data demanding in comparison to SWIM and HSPF. For studying nitrogen fate and transport in unsaturated, saturated, and river zone, HSPF was a better model to simulate the desired nitrogen transformation and transport processes. However, for nitrogen dynamics and transformations in shallow streams, ECO Lab had an edge due its flexibility for inclusion of user-desired water quality parameters and processes. In the case of SWIM, most of the input data and governing equations are similar to SWAT but it does not include water bodies (ponds and lakes), wetlands and drainage systems. In this review, only the processes that were needed to simulate the Tollense River catchment were considered, however the resulted model selection criteria can be generalized to other lowland catchments in Australia, North-western Europe and North America with similar complexity.     

Simulating bimodal tall fescue growth with a degree‐day‐based process‐oriented plant model

Plant growth simulation models have a temperature response function driving development, with a base temperature and an optimum temperature defined. Such models function well when plant development rate shows a continuous change throughout the growing season. This approach becomes more complex as it is extended to cool‐season perennial grasses with a dormant period and bimodal growth curves. The objective of this study was to develop such a bimodal growth model for tall fescue (Schedonorus arundinaceus (Schreb.) Dumort) in the Midwest USA based on multiyear measurement trials. Functions for bimodal growth were incorporated into the ALMANAC model and applied to tall fescue using published tall fescue yields for a variety of sites and soils. Fields of cultivars “Kentucky 31” and “BarOptima Plus E34” were divided into paddocks and sampled weekly for dry‐matter accumulation. These biomass estimates were used to derive weekly growth values by differences between sequential weekly samplings. The measured values were compared to a single tall fescue simulation each year on one soil. Using these results, the ALMANAC model was modified and tested against mean reported tall fescue yields for 11 sites, with one to three soils per site. When we introduced midsummer dormancy into ALMANAC, we assumed dormancy began on the longest day of the year and lasted until the photoperiod was 0.68 hr shorter than the longest. ALMANAC simulated previously reported tall fescue yields well across the range of sites. Thus, ALMANAC shows great promise to simulate bimodal growth in this common cool‐season grass.