基于Landsat 8的节水改造背景下盐渍化土壤含盐量反演

为探明沈乌灌域节水改造后因渠道衬砌、引排水量减少引起的土壤含盐量时空分布特征及变化规律,采用区域土壤信息定点监测,结合经典统计学、空间插值以及机器学习建模反演等技术手段,利用Landsat 8卫星获取光谱数据,通过对实测土壤含盐量、光谱指数及波段反射率进行处理,运用Adaboost回归、BP神经网络回归、梯度提升树回归、KNN回归、决策树回归、随机森林回归方法构建了沈乌灌域土壤含盐量空间反演模型。采用最优反演模型对沈乌灌域土壤含盐量空间分布特征进行了遥感反演。结果表明： 通过全变量单一回归法筛选出相关系数大于0.55的9个光谱因子,使用SPSS PRO软件构建6种机器学习反演模型,对比6种反演模型精度,验证集决定系数R2由大到小依次为随机森林回归、梯度提升树回归、Adaboost回归、KNN回归、决策树回归、BP神经网络回归。其中随机森林回归模型的拟合精度最佳,训练集与验证集的决定系数R2分别为0.834和0.86,说明随机森林回归模型的反演效果较好。反演结果表明：节水改造后非盐渍土面积增加391.7km2,占灌域总面积的21%,中度盐渍土面积、重度盐渍土面积、盐土面积分别减少95.61、63.37、45.7km2,分别占灌域总面积的5%、3%、2%。综上所述,节水改造工程完成后,沈乌灌域土壤盐渍化程度减轻,作物生长安全区面积增加,但由于渠道衬砌以及引排水量减少,土壤盐分淋洗效果减弱,土壤盐分在灌域内部运移,整体土壤环境得到改善,局部地区出现盐分聚集。     

基于Sentinel-2破碎化地块灌区作物种植结构的提取

【目的】探究基于Sentinel-2遥感影像的决策树分类模型提取破碎化地块灌区作物种植结构的适用性。【方法】选取新疆阿拉沟灌区为研究区,以2021年覆盖作物全生育期的Sentinel-2遥感影像为数据源,结合田间调查和Google高清影像目视解译采样,基于主要作物物候信息、NDVI时序特征等分析确定作物识别的关键期阈值,构建决策树模型进行灌区主要作物分类,并对分类结果精度验证。【结果】基于Sentinel-2提取的灌区种植结构分布图地块纹理清晰,能够满足灌区用水管理需要;构建的决策树分类模型可在灌区尺度实现作物分类,方法简便易行,总体精度达到81.56%,Kappa系数为0.716 6。【结论】采用Sentinel-2遥感影像和决策树分类方法识别破碎化地块灌区复杂作物分类是可行的,可为灌区输配水决策和农业用水精细化管理提供基础信息。     

灌区种植结构时空变化及其与地下水相关性分析

利用数据融合算法(ESTAFM)对Landsat7 ETM+和MODIS影像进行融合,构建了高时空NDVI数据集;结合地面实体作物NDVI变化特征和光谱耦合技术(SMT),提取了解放闸灌域2000—2015年间种植结构空间信息;在此基础上分析了其时空特征变化以及地下水对种植结构调整的制约性。利用2015年实地调查数据对遥感监测结果进行评价,灌域主要作物类型玉米、向日葵、小麦以及套种的纯像元分类精度分别达到了95%、88%、91%和90%,总体精度达到了91%;多年遥感监测结果与历史统计数据相吻合。研究结果表明:玉米和小麦种植面积不断增加,玉米种植面积增加尤为明显,由2000年的0.83万hm2(占灌溉面积的5.80%)增加到2015年的4.02万hm2(占灌溉面积的28.31%);向日葵种植面积由下降变为上升趋势,其种植规模受市场因素主导;因土地流转、农村劳动力的外流以及套种种植模式劳动力成本高,使得农户转向单一作物种植模式,套种模式种植面积逐年下降,由2000年的4.30万hm2(占灌溉面积的30.32%)减少到2015年的0.41万hm2(占灌溉面积的2.91%)。种植结构空间格局与地下水埋深分布相似,其空间的相对差异并未随时间发生明显变化。向日葵在地下水埋深较浅、盐碱化偏高的区域分布集中,玉米、小麦及套种在地下水埋深大、盐碱化偏弱的区域分布集中,反映出向日葵耐盐性更高,更能适应盐碱化程度高的土壤;作物这一空间分布格局不仅受水土环境影响,而且与作物生理特征密切相关。灌区节水改造以来,在净引水量减小的情况下,农田蒸散发量并未减小,而是表现为稳中有升的趋势,这与高耗水作物玉米种植面积逐年增加有直接关系;地下水位整体下降,由1.76 m下降到2.33 m,土壤盐碱化得到改善,表明大型灌区节水续建配套项目实施以来,节水改造对水循环和水土环境起到了积极的影响。     

基于GIS的河套灌区井渠结合分布区的确定方法

实施井渠结合,合理开发利用地下水是河套灌区节水和控制土壤盐碱化的重要途径。运用ArcGIS分析了地下水矿化度和土壤岩性特征,采用空间分层的处理方法构建了地下水矿化度分区图,并结合灌区土地利用状况和农业灌溉用水要求确定了井渠结合区分布以及面积。结果表明,将矿化度数据分3个空间层处理是可行的,建立的矿化度分区图呈一定的空间变化特征。井渠结合区主要分布在乌兰布和灌域非沙漠带、解放闸灌域和永济灌域南部、义长灌域北部及其与乌拉特灌域、乌梁素海交界附近区域、乌拉特灌域尾部、灌区北部沿狼山山脉部分区域。井渠结合区约占整个灌区控制面积的45%。     

不同土质灌区土壤盐分与物理特性空间变异规律及协同关系研究

为探究土壤盐分空间特性,以鄂尔多斯黄河南岸灌区不同土质的两个灌域为研究对象,采用经典统计学与地质统计学相结合的方法对0～20 cm土层土壤电导率与土壤物理特性参数间的关系进行分析研究。研究结果表明:两灌域0～20 cm土层土壤电导率均可以采用高斯模型进行半方差分析拟合。砂土灌域土壤电导率为中等变异程度,中等空间相关性,电导率与黏粒含量、土壤容重、含水量、导热率及热容量显著正相关,相关范围约2～6 km;与砂粒含量显著负相关,相关范围约2.5～4 km。砂壤土灌域土壤电导率为中等偏强变异程度,较强空间相关性,电导率与土壤物理特性空间相关性较低,土壤盐渍化程度高于砂土灌域。     

河套灌区解放闸灌域盐荒地积盐量分析

以解放闸灌域为背景对盐荒地水分和盐分的变化展开研究.在田间试验的基础上,通过遥感技术对TM数据解译统计盐荒地面积,利用HYDRUS-1D模型对盐荒地土壤盐分的积聚量与流失量进行了模拟和分析.结果表明,盐分在表层土壤中积聚,解放闸灌域总面积为2 156.47 km²,盐荒地达到了总面积的16.32%.灌域盐荒地1 m深土壤层中,在作物生育期积聚盐量为10.32×10⁶ kg,在秋浇期流失盐量为9.60×10⁶ kg,占积盐量的93.01%;年度积盐量为0.72×10⁶ kg,占积盐量的6.99%.经检验,HYDRUS-1D模型对土壤水分和盐分在垂直方向运移的模拟具有较高精度,反映出盐分积聚和流失规律.盐荒地盐分的积聚与流失对灌区盐分具有重要调节作用,可为灌区水盐运移的研究提供科学依据.     

植被覆盖条件下的解放闸灌域土壤盐分卫星遥感估算模型

为探究植被覆盖时的土壤盐分反演,以河套灌区解放闸灌域为研究区域、GF-1号影像为数据源,将盐分指数(SI2、S2、S3)、增强植被指数(EVI)和近红外NIR波段作为输入因子,分别利用多元线性回归(multivariable linear regression,MLR)、分位数回归(quantile regression,QR)和BP神经网络(back propagation neural network,BP)三种方法建立0～60 cm深度下土壤盐分反演模型。研究结果表明,MLR模型与QR模型均具有较高精度,能够较好的反演植被覆盖时的土壤盐分,其中QR模型验证精度最高,建模和验证的决定系数(coefficient of determination,R~2)分别达到0.627与0.636,均方根误差(root mean square error,RMSE)为0.249,平均绝对误差(mean absolute error,MAE)为0.235,是本次土壤盐分估算的最优模型。BP模型效果相对较差,建模与验证R~2为0.605和0.558。采用QR模型反演研究区土壤盐分,发现模型反演的盐分趋势符合实际情况;灌区主要分布非盐土和轻度盐渍化土壤,灌域南部地区土壤盐渍化程度低,约占32%;盐渍化程度较高的区域约占灌域总面积的19%。研究为探讨植被覆盖时的土壤盐分反演提供了思路。     

基于Google Earth Engine的黄土高原覆膜农田遥感识别

为了建立覆膜农田遥感识别技术体系,本研究选取甘肃省定西市安定区团结镇作为黄土高原地膜覆盖旱作农业代表性区域,基于Google Earth Engine云平台和Landsat-8反射率数据,采用特征重要性分析优选纹理特征,利用参数优化后的随机森林算法提取覆膜农田区域并选出最佳特征组合方案,最后通过对比随机森林、支持向量机、决策树和最小距离分类4种算法的分类结果来评价不同分类算法的性能。结果表明:优化关键参数后的随机森林算法能够显著提高遥感影像的分类精度;单一特征方案中,基于光谱特征的分类精度最高,且加入指数和纹理特征可提高总体识别精度;利用随机森林特征重要性分析选取的优选纹理特征分类性能优于全部纹理特征,基于"光谱+指数+优选纹理"特征方案的识别结果最佳,总体精度和Kappa系数达95.05%和0.94;与支持向量机、决策树和最小距离分类相比,随机森林优势明显,总体精度分别高3.10、7.74、50.78个百分点。本研究实现了对地形复杂地区覆膜农田空间分布较为精准的识别。     

基于遥感的宝鸡峡灌区作物面积信息提取

灌区中不同作物的种植面积是分配灌溉用水的重要依据,遥感方法不仅可以提取各作物种植面积,而且可以获取其空间分布情况。基于2009年Landsat5-TM影像,综合分析宝鸡峡灌区农作物种植特点和光谱特征,运用决策树分类法获取灌区各作物面积。并由地面实际调查数据和高分辨率影像对分类精度进行评价,各地物精度均达85%以上。结果表明该方法适合灌区的作物分类。     

基于Sentinel-2的青铜峡灌区春小麦和苜蓿早期识别

【目的】基于决策树分类算法对青铜峡灌区春小麦和苜蓿进行早期识别。【方法】在实地调研的基础上,结合Sentinel-2遥感影像,构建能够不断融入遥感数据的决策树分类算法,对青铜峡灌区春小麦和苜蓿进行早期识别。【结果】4月上旬,由于春小麦和苜蓿生长特征较为相似,春小麦和苜蓿的总体分类精度分别为69%和75%。此后,随着观测数据的不断融入,分类精度逐步提升,5月14日二者的分类精度即可达到90%以上。5月31日利用全部5期卫星影像数据时,春小麦的总体精度可达到94%,Kappa系数为0.75,苜蓿的总体精度可达到97%,Kappa为0.86。2023年青铜峡灌区春小麦种植面积为24 000 hm²,苜蓿种植面积为2 000 hm²。其中,春小麦种植结构较为复杂,既存在集中连片的大规模种植,也存在大量面积较小的碎片化种植带。【结论】基于Sentinel-2遥感数据的决策树分类方法,可在4月上旬获取初步分类结果,在5月中旬苜蓿第1次收割之前对青铜峡灌区春小麦和苜蓿进行较为准确的识别。     

不同灌溉制度对根层土壤盐分影响的模拟

以河套灌区沙壕渠灌域为例,采用SaltMod模型探讨了不同灌溉制度对作物根层土壤盐分的影响.利用2008-2010年基础资料对模型进行率定和验证并对研究区的根层土壤盐分进行模拟和预测.结果表明:在现有灌排条件下,沙壕渠灌域的盐渍化程度基本达到较为稳定的水平,且有轻微脱盐趋势,未来10 a后作物根层土壤盐分降低3%.生育期土壤水盐垂直交换运动强烈,对于控制土壤盐分而言,灌水量越小越好;作物根层土壤盐分随冬灌灌溉定额的增大而减小,不同冬灌灌溉定额对应的根层土壤盐分最初的增加量均较大,但增加的趋势随时间逐渐降低.根据当地作物种植结构,综合考虑节水灌溉、作物产量和根层土壤水盐环境,建议研究区较优的作物生育期综合净灌溉定额为2 700~3 500 m³/hm²,冬灌净灌溉定额为2 700 m³/hm².     

小麦间作向日葵耗水量与优化灌溉制度研究

小麦间作向日葵是一种典型的河套灌区种植模式,研究运用WIN ISAREG模型确定了参考作物蒸发蒸腾量,结合立体种植特征,采用综合作物系数法确定小麦间作向日葵全生育期作物耗水量。同时通过实际处理结果对模型参数进行率定与验证,最终对小麦间作向日葵实际灌溉制度进行准确的评价并进行不同方案的优化,得到试验区如配有井灌等条件,优选灌溉方案为土壤含水率下降至适宜含水率70%时实施灌溉,灌水量为补充根系层水量至田间持水量85%所需水量;若灌水水源仅为黄河水,则优选灌溉方案为整个生育期灌5水,灌水量为达到根系层中总有效水量的80%。     

DRAINMOD-S: Water management model for irrigated arid lands,crop yield and applications

The primary objective of an agriculture water management system is to provide crop needs to sustain high yields. Another objective of equal or greater importance in some regions is to reduce agriculture impacts on surface and groundwater quality. Kandil et al. (1992) modified the water management model DRAINMOD to predict soil salinity as affected by irrigation water quality and drainage system design. The objectives of this study are to incorporate an algorithm to quantify the effects of stresses due to soil salinity on crop yields and to demonstrate the applications of the model. DRAINMOD-S, is capable of predicting the long-term effects of different irrigation and drainage practices on crop yields. The overall crop function in the model includes the effects of stresses caused by excessive soil water conditions (waterlogging), soil water-deficits, salinity, and planting delays. Three irrigation strategies and six drain spacings were considered for all crops. In the first irrigation strategy, the irrigation amounts were equal to evapotranspiration requirements by the crops, with the addition of a 10 cm depth of water for leaching applied during each growing season. In the second strategy, the leaching depth (10 cm) was applied before the growing season. In the third strategy, a leaching depth of 15 cm was applied before the growing season for each crop. Another strategy (4th) with more leaching was considered for bean which is the crop most sensitive to salinity. In the fourth strategy, 14 days intervals were used instead of 7 and leaching irrigations were applied: 15 cm before the growing season and 10 cm at the middle of the growing season for bean. The objective function for these simulations was crop yield. Soil water conditions and soil salinity were continuously simulated for a crop rotation of bean, cotton, maize, soybean, and wheat over a 19 years period. Yields of individual crops were predicted for each growing season. Results showed that the third irrigation strategy resulted in the highest yields for cotton, maize, soybean and wheat. Highest yields for bean were obtained by the fourth irrigation strategy. Results are also presented on the effects of drain depth and spacing on yields. DRAINMOD-S is written in Fortran and requires a PC with math-coprocessor. It was concluded that DRAINMOD-S is a useful tool for design and evaluation of irrigation and drainage systems in irrigated arid lands.     

Influence of different water quantities and qualities on lemon trees and soil salt distribution at the Jordan Valley

The influences of water quantity and quality on young lemon trees (Eureka) were studied at the University of Jordan Research Station at the Jordan Valley for 5 years (1996–2000). Five water levels and three water qualities were imposed via trickle irrigation system on clay loam soil. The primary effect of excess salinity is that it renders less water available to plants although some is still present in the root zone. Lemon trees water requirements should be modified year by year since planting according to the percentage shaded area, and this will lead into substantial water saving. Both evaporation from class A pan and the percentage shaded area can be used to give a satisfactory estimate of the lemon trees water requirement at the different growth stages. The highest lemon fruit yield was at irrigation water depth equal to evaporation depth from class A pan when corrected for tree canopy percentage area. Increasing irrigation water salinity 3.7 times increased average crop root zone salinity by about 3.8–4.1 times.The high salt concentration at the soil surface is due to high evaporation rate from wetted areas and the nature of soil water distribution associated with drip irrigation system. Then, the salt concentration decreased until the second depth, thereafter, salt concentration followed the bulb shape of the wetted soil volume under trickle irrigation. Irrigation water salinity is very important factor that should be managed with limited (deficit) irrigation. But increasing amount of applied saline water could result in a negative effect on crop yield and environment such as increasing average crop root zone salinity, nutrient leaching, water logging, increasing the drainage water load of salinity which might pollute ground water and other water sources.     

矿区农田土壤重金属分布特征与污染风险研究

对渭北旱原矿区130个农田土壤样品的Cd、Cr、Cu、Pb、Zn含量进行了测定,结果显示,Cd、Cu、Pb平均含量均高于陕西省土壤背景值,而Cr和Zn含量低于背景值。利用地统计方法得到的土壤重金属含量分布图显示,土壤各重金属含量由西向东呈下降趋势,水泥厂周边土壤重金属含量最高。相关性分析和主成分分析结果表明,5种重金属之间呈极显著正相关,说明其存在较高的同源性或复合关系。第1主成分主要由Cd构成,且主要反映了人为活动的影响,而第2主成分中的Cr所占负荷最高,体现了成土母质的作用,Cu、Pb和Zn含量受人为活动和成土母质共同影响。分别利用污染负荷指数(PLI)法和潜在生态危害指数(PER)法对研究区域土壤污染风险进行了评价,评价结果为煤矿区呈无污染或轻微到中度污染,水泥厂区土壤呈中度污染水平,单一元素污染程度由高到低依次为Cd、Pb、Cu、Cr、Zn。     

干旱区垄膜沟灌条件下土壤水盐空间分布特征研究

研究了垄膜沟灌条件下,不同灌溉定额下土壤水盐空间分布特征。结果表明,土壤含水率的时空分布受灌水次数和灌水定额的影响。灌溉期,在垂直方向上沟底土壤含水率较垄顶明显增加。在非灌溉期,由于强烈的蒸发蒸腾作用,垄顶土壤含水率持续降低,含水率阶段性变化明显。土壤盐分空间分布随土壤含水率的变化而变化,在灌溉期,沟底土壤脱盐深度随灌溉定额的增加呈增加的趋势。在非灌溉期,垄顶在作物生育期均发生积盐现象,且垄顶表层土壤盐分累积量高于沟底表层土壤。     

西北内陆旱区洋葱耗水规律试验研究

以白皮洋葱为试材,利用称量式蒸渗仪灌水试验,研究了不同灌水量及土壤含盐量对洋葱的耗水规律、产量和水分利用效率的影响。供试土壤分别为低盐土和高盐土,低盐土在鳞茎膨大期设置3种灌水处理,其对应的灌水下限分别为0~30 cm土层土壤含水率达到田持的80%、70%和60%,其他生长期和高盐土的全生育期灌水下限均为田持的80%。研究结果表明,鳞茎膨大期是洋葱的关键需水期;洋葱作物系数立苗期为0.50,发叶期为1.00,鳞茎膨大期为1.50,成熟期为1.25。低盐土的中水处理洋葱可获得最大产量、水分利用效率和较大的灌溉水利用效率;水分和盐分胁迫对洋葱的产量和耗水具有明显的影响。在石羊河流域洋葱鳞茎膨大期,0~30 cm土层土壤灌水下限为田持的70%左右是较适宜的灌水方案。     

不同作物种植条件下微咸水灌溉的土壤环境效应试验研究

在田间试验条件下,采用正常灌溉定额和淋洗灌溉定额2种灌溉水平,对小麦、葵花、玉米3种作物进行了不同灌溉水浓度处理的灌溉试验,对3种作物种植条件下土壤水盐运移规律进行了研究。研究发现灌溉水浓度达到某一临界值时,盐分将在土壤剖面一定深度的土层内聚集明显增大。且不同作物的临界灌水浓度值不同。     

北疆膜下滴灌棉田生育期土壤水盐的变化特征

针对北疆地区土壤盐化制约农业发展的问题,通过野外实地分层采集土样,分析北疆地区长期膜下滴灌农田种植前后土壤水分和盐分分布特征。统计学分析表明,种植期土壤含水量变异属于中等偏弱变异强度,土壤盐分的变异属于中等偏强变异程度,土壤盐分分布成自上而下先增大再减少的波动趋势。在空间分布上,土壤水分和盐分种植前后存在明显的差异,种植后在浅层土壤水分和盐分的含量成西高东低的变化分布。     

Adapting the CROPGRO model for saline soils: the case for a common bean crop

Water scarcity and severe environmental degradation are causing water managers in the Fergana Valley, Uzbekistan to re-evaluate irrigation water use. Crop models could play an important role in helping farmers decide which systems (crops and irrigation technologies) are feasible. CROPGRO is a physiologically robust agronomic model, although the current version does not consider the effects of soil salinity on crop water use or growth. CROPGRO was modified to include a salinity response function and was tested for gypsiferous soils. A qualitative analysis of the model indicated the model performed as expected under a range of atmospheric, irrigation and crop tolerance scenarios. Model simulations compared very favourably for common bean (Phaseolus vulgaris) to results obtained in the greenhouse for yield and seasonal crop evapotranspiration with values of the Willmott agreement index (i) of 0.98 for both variables evaluated at different levels of salinity and deficit irrigation. Final biomass predictions were less satisfactory, although the modified model performed as well as the original model. The modified model was successfully tested with field data on common bean from an experiment in the Fergana Valley (i of 0.75 for ET and 0.74 for final yield), although the sensitivity of the model to a soil fertility function and relative nodule number made it difficult to assess the model performance.