【作者】 王改改；

【导师】 魏朝富； 谢德体；

【作者基本信息】 西南大学 ， 农业资源利用， 2009， 博士

【摘要】 土壤水分是作物生长、植被恢复以及生态环境建设的主要限制性因素。如何提高降雨的资源化水平,充分利用有限的水资源,降低农业生产的需水量和耗水量已成为一个国际性的课题。近年来,随着地统计学、分形理论、小波分析等新技术的应用,土壤水分的时空变异及农业生态系统水分运移模型等方面均取得了大量的研究成果,为农田水分优化管理与调控技术提供了基础。但是,这些研究成果主要集中在半干旱、半湿润气候带的华北平原和黄土高原,而在南方季节性干旱的丘陵山地研究较少。丘陵山地具有独特的水文过程,不同水文要素间相互影响相互作用导致土壤水分空间变异的复杂性,也影响到普适论的应用,已经成为研究的热点和难点之一。重庆区域面积的90%以上为丘陵山地,耕地面积的的50%左右为15°以上的坡耕地,该区也是我国水土流失和季节性缺水最严重的地区之一。因此,在重庆地区开展丘陵山地土壤水力特性和水分变异的研究,对重庆山区农业水资源调控、降水资源的集约利用、季节性干旱问题的解决具有重要科学和现实意义。以为区域水资源优化调控提供科学依据为出发点,围绕重庆地区土壤水力特性与土壤水分的时空变异规律,在重庆不同区县采集30个土壤剖面90个土壤样品,通过室内测定土壤的质地、有机质、土壤水分特征曲线、扩散数据探讨13个传递函数模型在重庆地区的适用情况,筛选出重庆地区的最佳点状模型和参数传递函数模型,然后利用筛选出的传递函数结合现有土壤普查资料、土壤质地数据推求重庆地区土壤水力参数,并分析土壤水力参数的空间变异。为深入分析重庆丘陵山地土壤水分时空变异状况,利用重庆地区169个土壤水分动态监测站点3个土层（0-10cm、10-20cm、20-40cm）的54756个动态监测资料（2006年、2007年3月-10月每隔5日1次）探讨土壤水分的时空演变规律及其内蕴稳定性。主要研究结果如下:（1）土壤持水能力差异大:土壤质地粗,土壤释水快,持水性能低;部分土壤质地粘重,有效水范围窄,易干旱。土壤水分特征曲线通过常用的幂函数表示土壤水分随压力的变化特征,土壤水分在低压力下下降速度较快,在5×10⁵Pa后高压力段下降非常缓慢,接近平缓。当土水势由-0.10～-1×10⁵Pa降至-1～-5×10⁵Pa段,再降至-5～-10×10⁵Pa段,比水容量有10^-1个数量级降至10^-2个数量级,再至10^-3个数量级。在土水势低于-5×10⁵Pa时,土壤的释水量低,是土壤容易发生旱灾的重要原因。土壤扩散系数随土壤水分变化表现出在低水分含量时,其值较低,当含量接近饱和时,其值急剧增加,可用指数形式D=aθ^eb表达,非饱和导水率的变化也表示出相同的变化规律,壤质土壤较高的持水能力和低土壤基质势较高的非饱和导水率,能提高土壤水分的有效性。设定l=0.5,采用Van Genuchten方程拟合土壤水分特征曲线和土壤扩散率的水力参数,结果发现:土壤质地越粗,θr和θs的值就较低,a、n值反而较大;饱和导水率变异大,难以如实的反映当地的实际情况。土壤颗粒组成、容重、有机质、孔隙率影响土壤持水性能。砂粒与各压力下水分含量显著负相关（P＜0.01）,粘粒与各压力下土壤水分显著正相关（P＜0.01或P＜0.05）,有机质与＜0.33×10⁵Pa压力下水分含量显著正相关（P＜0.05）,容重、孔隙率的相关性不显著,但在低压力段正相关,高压力段负相关。土壤的砂粒、粘粒是影响土壤持水能力的重要因素。（2）EPIC和神经网络传递函数模型可用来预测该区水力参数,水力参数空间变异存在尺度效应。土壤水力参数空间变异的研究是进行土壤水分时空变异分析的前提条件,利用SOILPAR软件的12个传递函数模型和RETC模型估计了土壤田间持水量和萎蔫含水量,并采用IRENE软件中的SB、RMSE、EF、MBE、FB、MAE和1∶1线性方程的坡度（Slope）、截距（intercept）对模型的性能进行分析,结果发现:相对于其他传递函数,在重庆地区更适合用EPIC和RETC软件中的神经网络模型来估测土壤水力参数和特定压力下的土壤水分。基于传递函数模型进行土壤水力参数空间变异的分析发现:θ_r在0-20cm和20-40cm可分别采用指数模型和直线模型进行拟合,其块金值为分别为0.41和0.87;采用球状+指数套合结构模型对2个层次拟合的参数均为正值,R²在0.87以上,存在小尺度变异。0-20cmθ_s用直线模型进行拟合,20-40cmθ_s在整个空间存在漂移,局部稳定性用直线模型进行拟合,R²为0.77;0-20cm和20-40cmθs的块金值分别为0.34和0.71,均不为0,存在小尺度变异。0-20cmα直线模型拟合参数的纯块金值非常接近0;20-40cmα及lnα不具备空间结构特征。0-20cm和20-40cm的n及ln（n）在整个空间发生了漂移,未呈现良好的空间结构,但ln（n）存在局部稳定性,可用球状模型拟合。（3）土壤水分的季节分配、剖面分布、空间格局多为中等程度变异,但仍为平稳时间序列。土壤水分时间稳定性分析是进行土壤水分空间分析的基础,采用Pachepsky et al提出的稳定性指数方法利用重庆地区2年的监测数据分析了采样点的时间稳定性指数,结果发现不同土壤层次的时间稳定性指数的幅度为0.22-0.34,表层时间稳定性指数最高,稳定性最差;采用柯尔莫哥洛夫-斯米诺夫方法对间隔5d、10d、15d的采样频率下的时间稳定性进行分析,发现高间隔频率仍具有时间稳定性。采用非相关参数的spearman相关参数发现3-5月各采样时间间的相关性高于水分剧烈波动6-9月的相关性。土壤水分时间稳定性的存在有助于区域水分平均值代表站点的选择,发现采样点黑水、木洞的水分最能代表重庆地区水分的平均值,间隔5天、10天、15天的监测频率下,区域水分平均值的代表监测样点发生变化。土壤水分动态的研究是定量理解植被对水分胁迫响应、土壤养分循环的水文控制、植物水分竞争等生态系统动态的关键。利用2006年-2007年每隔5天的土壤水分监测数据,研究了重庆代表墒情站点的木洞地区土壤水分的动态。结果表明:在湿润气候条件下,干旱年份（06年）,土壤水分的年内变异越大;平水年（07年）土壤水分的年内变异较小。其季节动态可划分3-5月蒸发消耗期,6-9月剧烈波动期。土壤水分的随机动态特征为,06年各层土壤水分概率分布的峰值出现在80%左右,07年出现在90%左右,无论在枯水年、平水年各层土壤水分峰值出现的位置随着土壤深度的增加。各层土壤水分垂向变异并非完全随着土壤深度的增加而增加,不同时间尺度下土壤水分的剖面分布分为4种构型即:稳定型、增长型、波动型、降低型。根据各层季节变异的方差和变异系数,0-40cm土壤水分的垂直层次相应可以分为0-10cm活跃层和10-40cm次活跃层。土壤水分的剖面分布具有相似的变点分布及周期性特点,0-10cm的变点分布最多,土壤各层的互谱系数接近1,层次间水分的上、下运行因果关系明确。平面变异上,土壤水分的结构系数C₀/C+C₀在50%左右,中等变异,即空间自相关部分引起的空间变异占整个空间变异的50%以上。土壤水分的空间分布格局采用普通克里格法插值法,可反应土壤水分分布格局季节动态的时间稳定性;以地形为辅助变量的回归克里格方法,其预测的0-10cm、10-20cm、20-40cm土层土壤水分平均误差分别比普通克立格法降低了0.35%、24.29%、17.71%,而均方根误差则分别比普通克立格法降低了0.51%、11.35%、11.40%,显著提高了预测精度,其空间插值更能反映土壤水分随地形的变化规律。土壤水分的时空动态与土壤水分库容关系密切,通过对丘陵山地土壤库容各组分的分析发现,丘陵山地土壤的通透库容变异超过了70%,为强变异;总库容、有效贮水库容、无效贮水库容的变异幅度为10%-50%,为中等程度变异。土壤水分的空间变差结构存在方向变异,不同库容的变异方向并不相同,可采用球状模型和指数模型模拟,R²均在0.6以上;土壤水库的各组分中仅20-40cm无效贮水库容的空间异质的61.88%是空间自相关索引起的,其他库容的空间异质中由空间自相关部分引起的占到75%以上,空间相关性较强。（4）多元时空模型可模拟生长季节土壤水分动态,剧烈波动时期模拟效果较差。采用多元线性模型的逐步筛选法建立3个土层的3个时空预测模型。模型0-10cm、10-20cm的截距与本层土壤水分平均值基本相等,20-40cm的截距为3.48,低于该层平均水分;各层模型R²修正值在0.5以上,即模型各要素能解释50%以上土壤水分的变化,预测模型的最优性、稳定性在3月、4月、5月较好,但是在6、8、9均较差,绝对无偏性在8月、9月模型的也较差。因此模型在土壤水分波动剧烈时期的预测性能较差。总之,重庆山丘陵山地持水能力变化大,比水容量小,有效水范围较窄,土层浅薄,抗旱能力差,土壤水分垂向运移速度快,是该区土壤易于发生季节性、区域性干旱的根本原因。土壤水分空间变异在大区域范围仍具有时间上的稳定性,3-5月生长季节,均受作物耗水量影响,土壤水分空间分布格局基本相同;6-9月受高温伏旱影响,土壤水分空间分布受地形的影响较大,甚至表现出随海拔的增加而增加的趋势,土地利用方法、降水季节分布不均和山地地形的影响也是诱发该区发生土壤水分季节性、区域性干旱的重要原因。因此,在重庆丘陵山地大力推行工程、农艺、生物、化控等措施相结合的集雨增效旱作农业技术,提高降雨转化效率,增强土壤水库在农田水分的调蓄是提高降水资源化效率的必然要求。更多还原

【Abstract】 It has become an international topic how to improve the utilization of rainfall and limited water resource,to reduce the water requirement and consumption by agriculture.Soil moisture plays a very important role in crop growth,vegetation recovery and ecological construction.Recently,with the application of new technologies and method,such as geo-statistical technologies and fractal theory,wavelet analysis in China have made great achievements in the fields of modeling water transport in agro-ecosystem,temporal and spatial variation,optimizing and regulating field water in agriculture and so on.However,most of current the researcher focus on north China plain and the loess plateau,the semiarid zone of China.And do apoor job in hills and mountain area,especially the study of the areas of southern China with seasonal drought in Chongqing.Topographic attributes are also useful indicators of hillslope and mountain hydrological dynamics.Hillslope hydrology remains challenging because a number of processes interact at different scales,significantly contributing to the complexity of the systems that hampers the possibility theory.Hills and mountain area was in an absolute dominance with the proportion of 90%of study are in Chongqing.Besides the complicated topography,the slope cultivated land which slope is more than 15°is about 50%proportion of cultivated land area.And in this region soil erosion and water shortage is serious and seriously limits the sustainable development of agriculture.In this research,a typical hill and mountain area of Chongqing in China is selected as an experiment area,where the soil is used to analyze the soil moisture dynamics and their environmental factors.The research provides a scientific basis for demonstrating and spreading of anti-drought technologies and soil moisture utilizes efficiently.This study focused on providing scientific basis for solving the season drought problem in hill and mountain area in Chongqing.According to the spatio-temporal variation of soil moisture in research region, 90 soil sample was sampled in different county in Chongqing and soil OM.soil particle distribution、soil water retention curve,soil diffusivity was measured,and utilized these data to select out a best point and parameter PTFs model to predict soil hydraulic parameters in Chongqing,and then collected 160 soil profile data,such as soil bulk density,soil particle distribution and soil OM to estimate soil field water capacity,wilt point and hydraulic parameter in research region,based on these estimation,Spatial variation of soil hydraulic parameters was analyzed.Spatial characteristics of unsaturated soil water movement parameter are the base and precondition of scientific understanding soil moisture dynamistic variation of large-scale.In order to further study on temporal and spatial variation of soil moisture in hilly and mountain in Chongqing,54756 data for soil moisture in 84 times were collected from March to September in 2006 and 2007.The soil moisture in layers of 0-10cm、10-20cm、20-40cm was measured at an interval of 5 days. All the data were analyzed by methods of the traditional statistics,geo-statistics,rank stability and wavelet analysis to investigate the spatial-temporal variation of soil moisture.The water holding capacity of soil, precipitation transport on sloping land,soil moisture dynamics and its affecting factor had been discussed in this paper.The main achievements of this study are as follows:（1）There is a big difference for soil water capacity,when the soil texture is coarse,soil reduces water discharge quickly,and soil texture is heavy,a narrow range of the available soil moistureThe specific water capacity decreases to the scale of 10^-2from 10^-1,and to the scale of 10^-3from 10^-2 when the water potential decreases from-0.l0～1×10⁵Pat o-l×10⁵Pa～-5×l0⁵Pa,and then to-5×10⁵Pa～-10×10⁵Pa The soil water content decreases very slowly and the soil water discharge rapidly with a small amount when the suction exceeds 5.0×10⁵kPa..This is the reason that soil is frequently influenced by seasonal droughts.The relationship between soil moisture diffusivity and soil water content in every soil layer were sharp positive exponential function as well as between soil moisture conductivity and soil water content,and these relationship were significant by statistical analysis.The van Genuchten（l=0.5,m=1-1/n）hydraulic parameters was fitted to soil water retention and soil diffusivity,the results are as follows,when soil texture is coarser,the value ofθ_r andθ_s is lower,but the value of a、n is higher.So the available soil water which soil is heavy texture is lower than which in median texture,the hydraulic parameters K_s cannot be used as a matching point for the hydraulic parameter in field or laboratory experiments.Then,the bulk density,porosity,the type of pore,particle constitute and organic carbon in the soil affected the soil moisture diffusivity.The water-holding capacity of soil is negatively related to contents of sand fractions with the size from 2mm to 0.05mm,there is a significant positive correlation between clay content and soil water under different tension,soil organic matter and soil water under＜0.33×10⁵Pa,there is no significant positive correlation between bulk density,porosity and soil water under different tension. Sandy particle and clay particle is the important factors influenced soil water capacity.（2）Pedotransfer function of EPIC and RETC model is the best model to predict soil hydraulic parameters,based on these research,the spatial variation of soil hydraulic parameters occurs scale effect. Temporal and spatial variation of soil moisture is the based on the spatial variation of soil hydraulic parameters.Thirteen pedotransfer functions（PTFs）,namely RERC,Brakensiek.Rawls,British Soil Survey Topsoil,British Soil Survey Subsoil,Mayr-Jarvis,Campbell,EPIC,Manrique,Baumer, Rawls-Brakensiek,Vereecken,and Huston were evaluated for accuracy in predicting the soil moisture contents at field capacity（FC） and wilting point（WP） of hills and mountain chongqing in china.PTFs were evaluated on the basis of SB,RMSE,EF,MBE,FB,MAE and 1:1 line in software of Irene.The result shows that in the case of the hills and mountain of Chongqing city.the EPIC and RETC were found to be the best methods more than the others.The hydraulic parameters are evaluated using PTF,based on these study,Spatial variation was analysized using Geo-statistics,the results showed that linear variogram model was fitted toθ_r in 0-20cm when the adjusted coefficient of determination of is 0.7;Accordingly,the nugget is 0.41 and 0.87 in 0-20cm and in 20-40cm respectively;based on these results,spherical and exponential were used to fit toθ_r.the parameters are all positive and the coefficient of determination was more than 0.87, this implies that spatial variance ofθ_r operate at multiple spatial scales.Linear variogram models are fitted to the hydraulic parametersθ_s in 0-20cm,the whole spatial variation ofθ_s in 20-40cmdon not reach stable state,but the local variation is stable,linear model are fitted to the local stability,and the coefficients of determination is 0.77.and it validates the accuracy of the linear model to fit the local stability.The nugget ofθ_s are 0.34 and 0.71 in 0-20cm and 20-40cm respectively,that is all different from 0,spatial variation occurred at smaller scale less than the sample length.The change of semivarigram ofαis as well as a stage change,the linear model is fitted to the change.the nugget of model close to 0.there occurred a slight spatial variation in smaller than the sample scale;αand lnαin 20-40cm varies with lag length change and the curve is significantly concavo,thus there occurs no spatial structure.The hydraulic parameters of n and ln（n） had no evident spatial structure in all layers,but ln（n） in 20-40cm is stable in local range,the spherical variogram models are fitted to the spatial variation.（3） There are median variations degrees within seasonal,vertical and spatial distribution of soil moisture,but still contribute to stable temporal series.When a field or a small watershed is repeatedly surveyed for soil water content,locations can often be identified where soil water contents are either consistently larger or consistently less than the study area average.This phenomenon has been called temporal stability,time stability,temporal persistence,or rank stability in spatial patterns of soil water contents.Temporal stability is of considerable interest in terms of facilitating upscaling of observed soil water contents to obtain average values across the observation area, improving soil water monitoring strategies,and correcting the monitoring results for missing data.The objective of this work was to contribute to the existing knowledge based on temporal stability in soil water patterns using frequent multi-depth measured with micro-wave oven。Water contents at 0-10cm、10-20cm、20-40cm depths were measured every 5 days for 7 months of observation from March to September in 2006 and 2007.Temporal stability are analyzed using temporal stability index suggested by Pachepsky, The temporal stability index ranges from 0.22-0.34 and decreases with depths,the temporal stability increase with depth.The statistical hypothesis could not be rejected that data collected each five days,each ten days,each fifteen days had the same temporal stability.The spearman correlation parameters aim to describe the temporal variation of temporal persistence.The coefficient of correlation from March to May is higher than the value from June to September.The locations of which were best for estimating the average water contents were different for different depths.The best three locations for the whole observation period were the same as the best locations for a month of observation in about 60%of the case. Temporal stability for a specific location and depth could serve as a good predictor of the utility of this location for estimating the area-average soil water content for that depth.Temporal stability could be efficiently used to correct area-average water contents for missing data.Soil water contents can be upscaled and efficiently monitored using the temporal stability of soil water content patterns.The study of the temporal and spatial dynamics of Soil moisture is critical to understanding of several ecological hydrological processes,e.g.,water stress to plant,hydraulic control of soil nutrient cycle,plant water competition,and also is focus of ecological hydrology.Based on the data of every five days for soil moisture from 2006Y to 2007Y at mountain Mudong sites in Chongqing where soil moisture can be serve as a good predictor of the utility of this location for representing the area-average soil water content for the whole Chongqing,the dynamics of soil moisture were analyzed.The results showed soil moisture values observed at all depths are higher in rainy year than in dry year,in contrast to this,the variance is lower in dry year than in rainy year under subtropical climate.The seasonal dynamics of soil moisture can be divided into two stages:less unstable stage（from Mar to May）,and unstable stage（from June to Sept）.Soil moisture stochastic dynamics at a point is studied in detail utilizing the probability distributions.The peak value of soil moisture lied in about〈s〉=80%in 06 year,while in 07 year,the peak value lies in about〈s〉=90%。Soil water peak increases with soil depth in different years.Soil water do not increase with depth is increasing completely and the profile distribution can be devided into four structure（soil is stable with depth,increase with depth,fluctuate with depth,decrease with depth）in different temporal scales.By using standard deviation and variation coefficient,the vertical layers of soil water content can be divided into two layers,active layer（0-l0cm）and sub-active layer（10-40cm）.There are similar saltation point distribution and periodic span between the adjacent layers. There are the most salsation points in 0-10cm,the mutual spectral of coefficient in different layer closes to 1,that implies that the mutual relationships are evident in soil vertical transmission.Variation coefficient of soil moisture range from 15.42%to 23.15%,the degree of variation is medium.From the C₀/C+C₀ comparable values of soil three layers,the value is about 50%,spatial autocorrelation account for the 50% of the level of system variation.Spatial distribution is interpolated using ordinary kriging,to some extent; spatial distribution exhibit similar distribution according to seasonal division and agree with temporal stability of soil moisture dynamics.Mean error predicted with regression kriging by using terrain as ancillary variable reduced respectively by 0.34%,24.29%and 17.71%in 0-10cm 10-20cm and 20-40cm compared to spatial distribution with ordinary kriging.Root mean error reduced by 0.51%、11.35%、11.40% respectively,the spatial distribution of soil moisture which predicted with regression kriging varies typically with the terrain characters.Temporal and spatial of soil moisture closely related to soil water storage.Spatial correlations are analyzed with geo-statistics.In research site,the variance coefficient of transmission volumetric capacity outperformed 70%,degree of variance is very strong,while variance coefficient of total soil reservoir capacities,available soil reservoir and unavailable soil reservoir range from 10%to 50%,degree of variance is medium.Spatial semivarigram structures of soil water are anisotropy,spherical and exponential variance model are fitted to soil water reservoir.Spatial autocorrelations account for 61.88%of total variance of unavailable soil reservoir,while the residuals spatial autocorrelations accounts for 75%of total spatial variance,there is a very strong spatial autocorrelations.Soil water reservoir exhibited spotted spatial distribution by interpolating with ordinary kriging in Chongqing.（4）The performance of soil moisture prediction using multiple-linear model exhibits more suitable in cropping growth than in drought period.Three spatial-temporal prediction model was established respectively by using stepwise regression of multi linear model in 0-10cm,10-20cm and 20-40cm,and the model included 6 or 9 index which refer to meteorological factors,terrain factor,soil attributes and land use type.In which the intercept of model in 0-10cm and 10-20cm almost equals to mean soil moisture,while the intercept in 20-40cm is only 3.48 and is far lower than mean soil water accordingly soil layer.Regression coefficient of model combined with specialist’s knowledge may analyze the relationship between the independent variable and target variables. The adjusted value of model is more than 0.5 and showed that factors of model can explain less than 50% the change of soil water.F value is significantly（P＜0.01）different.And residuals are normal distribution, and the error value ranges from-2 to 2.This showed the model established can be exploited to simulate spatial and temporal of soil moisture.In order to evaluate the performance of model in detail.the performance of model was analyzed in terms of fitness,Optimum,absolute bias and stability based on the 11 performance indices from March to September,while the model are more optimum and stabile from March to May,but the value is higher in June.August,September as well as the MAE in August,September. Therefore,the performance of model prediction is poor in soil moisture actively fluctuant state.In summary,it can be seen from all above that the soil in mountain has a bigger variance,which leads to which water-holding capacity is variable,a limited specific water capacity,an extremely narrow range of available soil moisture,a thinner soil depth,a poor capacity of anti-drought and a limitation of available soil water in top soil.It is the basic reason that the soil in hills and mountains area of Chongqing is easily to be attacked by seasonal drought and localized drought.Spatial and temporal variation of soil moisture occurs temporal stability in a larger study region,The temporal span of soil water can be devided into two periods according the variability of soil moisture and effect factors:crop growth from March to May. because soil depletion is affected by crop consumption,spatial distribution is similar.Soil moisture changes,the reason are follows as:it was affected by high temperature,serious drought and terrain attributes,even soil moisture increasing with the hight of elevation is increase,thus land use type, seasonal distribution of precipitation and mountain topography which lead to seasonal and regional drought in research region.Based on these researches,soil moisture dynamics was nominated using muti-regression model,Due to a short monitoring span,there occurs obviously seasonal limitation of soil moisture simulation and the accuracy of simulation is lower in drought season than in cropping season. Therefore,it is necessary to take an action to start the engineering construction for available rainfall collection and extend the drying farming technologies combined with agricultural,biological measure as well as the measure for soil and water conversation as to maximize the utilization of rainfall and optimize the effect of soil reservoir on the field water supply.Besides these,In order to solve seasonal and localized drought,it is important to strengthen the dynamical monitoring in regional scale,and applied scientific management technology to forecasting soil drought.更多还原