基于NNARMAX模型的地下水位预报研究

区域地下水系统是一个受多种因素影响的复杂的非线性动态系统 ,应用遗忘因子的递归最小二乘( RLS)算法进行训练 ,采用通过减少网络节点间的连接权值 ,进而裁减冗余节点来选择适宜的网络拓扑结构 ,建立了非线性自回归滑动平均神经网络 ( NNARMAX)模型 ,地下水位预报结果表明 :通过对网络结构的优化达到了自动确定非线性自回归模型阶数和影响因素的选择 ,能有效地表示区域地下水动态系统内部及其外部诸多影响因素间的不确定关系 ,是预测区域地下水动态变化较为有效的方法之一。     

利用改进BP神经网络预测初产奶牛产奶量的方法

针对传统BP算法的缺陷，提出了一种采用L-M训练法的BP神经网络。在此基础上建立了基于改进BP神经网络非线性系统的奶牛305d产奶量预测模型，此模型可以提前215d预测初产奶牛305d产奶量，从而实现提早进行选择，加速奶牛育种工作进程。并通过具体的实验验证了改进BP神经网络预测模型的有效性。     

CO 2与土壤水分交互作用的番茄光合速率预测模型

为了实现不同土壤水分管理下的CO 2气肥精细控制,建立了番茄作物不同生长阶段的光合速率预测模型。实验设置了4个CO 2浓度与3个土壤水分条件的交互处理,利用无线传感器网络长期实时监测温室内环境信息,采用LI-6400XT型光合速率仪定时采集作物净光合速率信息;并用BP神经网络分别建立了番茄苗期、花期和果期的光合速率预测模型。预测模型的验证结果表明,对于苗期预测模型,预测值与实测值之间的决定系数 R 2为0.925;花期预测模型的决定系数 R 2为0.920,果期预测模型的决定系数 R 2为0.958;番茄各生长期的光合速率预测模型均具有较高的预测精度。在不同土壤水分条件下改变CO 2浓度,得到的CO 2浓度与光合速率预测曲线与实测值相近,可反映实际土壤水分管理下的CO 2浓度最优值,对指导不同土壤水分条件下CO 2气肥的精细调控具有重要意义。     

灰色动态模型群法在河流水质预测中的应用初探

在对原始数据序列对数变换的基础上，依据灰色系统理论，构造了由6个GM（1，1）模型组成的灰色动态模型群，并用于淮河干流枯水期氨氮浓度变化趋势预测。研究表明，灰色动态模型群法能够充分利用近期水质资料信息预测水质变化趋势；相对单个GM（1，1）模型，灰色动态模型群法能有效改善随机波动数据序列的拟合效果，提高预测精度。     

灌溉预报与节水灌溉决策专家系统研究

从节水灌溉的基本原理入手，分析灌溉预报中的各个参量，提出了节水灌溉决策的依据、方法及评价标准。介绍了灌溉预报决策专家系统中由框架库、规则组库和黑板3部分组成的综合知识体结构的知识表示技术。     

中国野生葡萄抗霜霉病相关基因VpPR4b及其启动子的克隆和功能分析

以中国野生葡萄(Vitis piasezkii)‘留坝-8’(抗霜霉病)和欧亚种葡萄(V.vinifera)‘黑比诺’(易感霜霉病)为材料,克隆得到抗霜霉病相关基因VpPR4b和VvPR4b。序列分析发现VpPR4b和VvPR4b核苷酸序列相似性为98.61%。VpPR4b编码区为432 bp,编码143个氨基酸,含有BARWIN保守结构域,属于典型的Ⅱ型PR4蛋白。VpPR4b和VvPR4b均可在霜霉菌侵染后诱导表达。亚细胞定位表明VpPR4b定位于细胞膜上。分别从‘留坝-8’和‘黑比诺’中克隆得到了VpPR4b和VvPR4b的启动子,其序列相似性高达95.65%,并且含有相似的顺式作用元件。通过酵母单杂交验证了转录因子WRKY40和WRKY75可以结合VpPR4b启动子。本研究的结果表明PR4b可能在葡萄抵御霜霉病侵染过程中发挥一定作用。     

中红外光谱技术在牛奶营养物质预测及奶牛相关特性分析上的应用

近年来,许多研究表明采用中红外光谱(MIRS)分析技术能够对牛奶中各营养物质(如脂肪酸、蛋白质和矿物质)及潜在的有害物质(如掺杂物、抗生素)进行实时、快速和准确地定量分析,并以此建立预测模型来对奶牛的营养(如饲料转化率、能量利用效率和甲烷排放)、健康(如乳房炎和代谢性疾病)和生殖生理与繁殖状况进行鉴定、评估和筛选,从而为优化畜牧养殖业生产方式,发展创新畜牧养殖业循环经济模式,实现低碳、健康和可持续发展提供有效的技术保障。因此本文主要就近年来国际上最新的关于MIRS分析技术在牛奶营养物质测定及奶牛相关特性分析方面的研究进展进行综述。     

Soil organic matter fractionation as a tool for predicting nitrogen mineralization in silty arable soils

After decades of searching for a practical method to estimate the N mineralization capacity of soil, there is still no consistent methodology. Indeed it is important to have practical methods to estimate soil nitrogen release for plant uptake and that should be appropriate, less time consuming, and cost effective for farmers. We fractionated soil organic matter (SOM) to assess different fractions of SOM as predictors for net N mineralization measured from repacked (disturbed) and intact (undisturbed) soil cores in 14 weeks of laboratory incubations. A soil set consisting of surface soil from 18 cereal and root‐cropped arable fields was physically fractionated into coarse and fine free particulate OM (coarse fPOM and fine fPOM), intra‐microaggregate particulate OM (iPOM) and silt and clay sized OM. The silt and clay sized OM was further chemically fractionated by oxidation with 6% NaOCl to isolate an oxidation‐resistant OM fraction, followed by extraction of mineral bound OM with 10% HF (HF‐res OM). Stepwise multiple linear regression yielded a significant relationship between the annual N mineralization (kg N/ha) from undisturbed soil and coarse fPOM N (kg N/ha), silt and clay N (kg N/ha) and its C:N ratio (R² = 0.80; P < 0.01). The relative annual N mineralization (% of soil N) from disturbed soils was related to coarse fPOM N, HF‐res OC (% of soil organic carbon) and its C:N ratio (R² = 0.83; P < 0.01). Physical fractions of SOM were thus found to be the most useful predictors for estimating the annual N mineralization rate of undisturbed soils. However, the bioavailability of physical fractions was changed due to the disturbance of soil. For disturbed soils, a presumed stable chemical SOM fraction was found to be a relevant predictor indicating that this fraction still contains bio‐available N. The latter prompted a revision in our reasoning behind selective oxidation and extraction as tools for characterizing soil organic N quality with respect to N availability. Nonetheless, the present study also underscores the potential of a combined physical and chemical fractionation procedure for isolating and quantifying N fractions which preferentially contribute to bulk soil N mineralization. The N content or C:N ratio of such fractions may be used to predict N mineralization in arable soils.     

Soil Properties Degradation Assessment Using Infrared Scanning Technique of Soils （Case Study： Eastern of Congo）

Recently, near infrared reflectance （NIR） and mid-infrared （MIR） spectroscopy techniques are increasingly introduced as convenient and simple non-destructive techniques for quantifying several soil properties. This study uses MIR method to predict pH, soil organic C, total N, AI, Ca, Mg and K, CEC and soil texture for soil samples collected in Sud-Kivu, Congo. A total of 536 composite soil samples were taken from two locations （Burhale and Luhihi） at two depths （0-20 cm and 20-40 cm） using a spatially-stratified random sampling design within an area of 200 km2. Differences in characteristics were evaluated between the two locations, land use （cultivated vs. non-cultivated land） with soil depths. A random subset of the samples （10%） were analyzed using standard wet chemistry methods, and calibration models developed by MIR data to estimate soil properties for the full soil sample set. Partial least squares regression （PLS） method gave acceptable coefficients of determination between 0.71 and 0.93 for all parameters. Soil organic matter levels were higher in cultivated plots in Luhihi （3.9% C） than in Burhale （3.0% C）, suggesting lower levels of soil fertility in the later area. This indicates high levels of acidity, which are likely to limit crop production in the area. Phosphorus deficiency is acute in Burhale （2.4 mg P/kg） but less in Luhihi （5.4 mg P/kg）. In both locations, low levels of Ca and Mg indicate that soils may be susceptible to deficiencies in both elements.These findings provide new opportunities for monitoring soil quality in the region which can benefit multiple actors and scientists involved in the agricultural and environmental sectors.