【作者】 王幼奇；

【导师】 邵明安；

【作者基本信息】 西北农林科技大学 ， 环境科学， 2011， 博士

【摘要】 陕北地区是我国生态环境最为脆弱的地区,研究流域尺度植被耗水特征及环境因子影响,对合理规划、利用水土资源及建设可持续发展的生态系统有重要意义。本文通过监测流域内不同土地利用方式土壤水分,系统掌握了不同植被类型耗水规律,并分析了小流域土壤理化性质的空间分布特征。本研究的主要结论如下:(1)不同土地利用方式土壤水分垂直变化上可分为土壤水分活跃层(0~40 cm)、土壤水分利用层(40~160 cm)和土壤水分调节层(160 cm以下)。变异系数和平均含水量呈线性负相关,极差和平均含水量呈线性正相关。不同土地利用储水量季节变化规律基本一致,5、6月份土壤储水量最少,8、9月份土壤储水量最高。2009年生长期降水量相对充足为421.5 mm,裸地、农地和灌木在雨季末土壤水分略有盈余。2010年生长季降水量较少为364.7 mm,导致所有土地利用方式土壤水分在雨季末均呈现出亏缺状态。2009年耗水量依次为:林地>草地>灌木>农地>裸地,2010年耗水量依次为:草地>灌木>林地>农地>裸地。(2)通过累积概率、相对偏差和spearman相关系数分析可知,LA、LB和LC子区域内点LA17、LB13和LC8与平均储水量间的相关性好、差异性小,可以分别代表3个区域剖面平均土壤储水量。LA和LB子区域3~4 m土壤储水量比其它土层稳定,子区域LC土壤储水量时间稳定性特征与LA和LB存在明显差异性。土壤储水量与spearman秩相关系数存在密切关系,土壤储水量变化较大时,spearman秩相关系数较小,时间稳定性差。而储水量保持稳定时,spearman秩相关系数明显较大。(3)不同处理3种植被株高和茎粗在生育期内的变化呈现出“慢-快-慢”的趋势,保水剂处理植被生长状况最好,株高、茎粗比秸秆覆盖和对照处理都要好。不同处理土壤蒸发量和植被蒸散耗水量日变化过程受降水量影响明显。秸秆覆盖明显减少了土壤蒸发量和植被蒸散耗水量,但对植被生物量和土壤性质的影响不明显,保水剂处理可减少土壤和植被的耗水量,同时对植被生物量和土壤性质的改善比较明显。保水剂和对照处理日实际蒸散耗水量与气象因素有明显的相关性,而秸秆覆盖条件下日蒸散耗水量与气象因子之间的相关性较弱。(4)经典统计表明流域内土壤颗粒分形维数、土壤pH、总孔隙度和容重属于弱变异,粘粒、粉粒、砂粒、有机碳、全磷、全氮、铵态氮、有效钾、有效磷、有效铜、有效锌、有效铁、有效锰和毛管孔隙度均属于中等变异,硝态氮和饱和导水率为强变异。单因素方差分析的结果表明不同土地利用方式显著影响土壤理化性质在流域内的分布状况。小流域内的分形维数值在1.37~2.66之间,分形维数与粘粒、粉粒和沙粒分别呈曲线相关。地统计学研究表明,流域内不同土壤理化性质含量及其它属性的空间变异结构存在较大差异,且土壤理化性质的空间自相关距变化较大,空间自相关范围差异显著。土壤理化性质插值图可以获取准确的土壤性质空间信息及分布规律,为该地区实施精准农业提供帮助和依据。(5)主成分人工神经网络模型对于4站点日ET0值的模拟能力非常好,特别在训练样本阶段主成分人工神经网络模型对榆林、太原和西安日ET0值模拟效果明显好于传统BP人工神经网络模型,且与Penman-Monteith公式计算值的相关性好、差异性小(R2 > 0.95)。与传统BP人工神经网络模型比较发现,经主成分简化的人工神经网络模型具有结构简单、精度高和降低噪音的效果。更多还原

【Abstract】 The nature ecological environment is very fragile in the northern loess plateau of china. Study on characteristics of vegetation water consumption and its influencing factors in a small watershed, will help to utilize water-soil resources and establish sustainable development ecosystem. In this study, in order to understand characteristics of different vegetations water consumption, the dynamics of soil moisture and water consumptions of vegetations on different land uses were analyzed. In addition, spatial variability of soil properties was investigated in a watershed on the Loess Plateau of China. The main conclusions of this study were as follows:（1） The results showed that the vertical change of soil moisture could be divided into three layers: weak absorbing layer, absorbing layer, and regulating layer. There were significant linear negative correlation between coefficient of variations and mean soil moisture, and significant linear positive correlation between extreme differences and mean soil moisture. The seasonal changes of soil water storage in different land uses were similar. In May and June the soil water storages were low, and in September and October were high. The total precipitation in the growing season was 421.5 mm in the year of 2009, which was sufficient for the growing of plants. The soil moisture was surplus for bare land, farm land, and shrub land in the end of growing season. The total precipitation was 364.7 mm in the year of 2010 which was 56.8 mm less than the year of 2009. In the end of growing season in 2010 the soil moisture was deficit. The orders of different land uses water consumptions in the year of 2009 were: wood land > grass land > shrub land > farm land > bare land. The orders of different land uses water consumptions in the year of 2010 were: grass land > shrub land > wood land > farm land > bare land.（2） The sites of LA17, LB13, and LC8 could represent the mean SWS of LA, LB, and LC subregions, respectively. The 3⁴ m SWS had higher time stabilities than other soil depths at LA and LB subregions. The time stability of SWS in LC subregion was obviously different from LA and LB subregions. The spearman rank correlation coefficients were small when the SWS varied greatly, but increased when the SWS remained more or less stable.（3） The change of vegetations height and stem diameter complied with“slow-quick-slow”curve during the different growing stages. The biomasses under PAM treatment were higher than straw mulch treatment and CK. The daily water consumption variations of the three types of vegetations were significantly affected by precipitation, indicating that evapotranspiration processes were controlled by the soil water moisture in the arid region. PAM and straw mulch could reduce the water consumption for bare land and three kinds of vegetations. PAM could increase the biomasses and improve soil properties; however straw mulch had no influence on them. There were significant correlation between the water consumption and meteorological factors on PAM treatment and CK, and no correlation on straw mulch treatment.（4） The classical statistics indicated that fractal dimension （D）, pH, total porosity, and bulk density were weak variability, and ammonium nitrogen （NH₄^`+-N）, extractable soil potassium （K）, phosphorus （P）, copper （Cu）, zinc （Zn）, ferrum （Fe）, and manganese （Mn） were medium variability, while nitrate nitrogen （NO₃^--N） and soil hydraulic conductivity （KS） had strong variability. Soil properties were mainly correlated to land uses within the research area. The D values in the research area ranged from 1.37 to 2.66. There was a significant sigmoidal correlation between D values and the contents of clay, silt, and sand fractions. Geostatistical analyses showed that the spatial autocorrelation of all soil properties were strong. The ordinary kriging maps could provide useful information for the development and application of precision agriculture in wind-water erosion crisscross region on the Loess Plateau of China.（5） The performance of the principal components BP neural network model was well in simulating daily ET0 on the Loess Plateau. The simulation ability of principal components BP neural network model was better than tradition BP neural network model during the training stage for Yulin, Taiyuan, and Xi’an stations. The ET0 values simulated by three principal components BP neural network were consistent with the Penman-Monteith method （R2>0.95）. It clearly demonstrated that high precision, simple structure and minimum error for the principal components BP neural network model in estimating ET0 compared to the tradition BP neural network model.更多还原