基于Landsat OLI的绿洲灌区土壤盐度最优预测尺度分析

【目的】基于遥感提取的多尺度遥感指示因子和土壤实测电导率数据,借助统计分析,试图探寻适合干旱区典型绿洲灌区土壤盐度变异的最佳观测尺度和指征变量,为快速评估绿洲土壤盐渍化提供备选方案。【方法】以新疆渭干河-库车河绿洲为研究区,以野外采集的土壤盐度数据(采集0—10、10—20、20—40和40—60 cm土层土样,制备土壤饱和溶液并测试电导率(ms·cm-1))并将其作为预测对象(n=87),借助Landsat OLI遥感影像数据,利用栅格重采样(30—1 000 m)和领域滤波(原始分辨率为30、60、90、120、150、180、210 m,滤波尺度为3×3至31×31)两种方式,生成多个尺度若干种指示因子(主成分分析、缨帽变化、植被指数、湿度指数),共计获得1 078个(其中,栅格重采样生成352个,领域滤波生成726个)环境变量。在此基础之上,利用线性和非线性曲线模型分别拟合上述两种模式下土壤盐度和环境变量之间的相关性,进而找出最优环境因子和预测尺度。【结果】栅格重采样模式下能够较好响应各层土壤变异性的皆为非线性模式。其次,该模式下,拟合精度随着空间分辨率的降低而降低。此模式下最佳推理尺度为30 m,该尺度下最佳响应变量除了40—60 cm处为三波段差分指数(Three-band Maximal Gradient Difference,TGDVI)外,其余深度皆为扩展的归一化指数(Extended Normalized Difference Vegetation Index,ENDVI)。领域滤波模式下的最佳推理尺度为180 m(滤波尺度3×3),同时,各层最佳拟合变量皆为扩展的增强型植被指数(Extented Enhanced Vegetation Index,EEVI)。相比较栅格重采样模式,该模式下的拟合精度全面优于前者,各层依次提高14.60%、34.40%、32.10%和21.70%。【结论】基于领域滤波模式下,像元分辨率为180 m,窗口大小为3×3的ENDVI指数更适合预测本研究区土壤盐度的空间变异性。     

基于GEE平台结合RF和SVM算法的茶园提取研究

科学有效的茶园遥感监测技术为土地利用管理、茶园管理、产业政策制定奠定了基础。研究基于GEE平台利用Landsat 8/OLI分别结合SVM和RF两种分类算法对云南省普洱市和西双版纳自治州的茶园进行了提取,并将2种算法的提取结果进行了对比。结果表明,RF和SVM算法的分类总体精度(OA)分别为95.61%、95.56%,Kappa系数相同,均为87%。RF算法的分类总体精度高于SVM算法,而Kappa系数相同。RF算法的制图精度(PA)为65.44%,与SVM算法(65.75%)相比相差较小,RF和SVM算法的用户精度(UA)分别为63.08%和57.37%。基于GEE平台结合RF分类算法对茶园的提取精度高于SVM算法。采用不同的传感器以及不同的分类算法可实现对茶园准确、高效的提取,对制定科学的茶园管理政策、茶园灾害预警、地表覆被变化等研究具有重要意义。     

基于遥感技术的宜良县云南松蓄积量反演

以云南省宜良森林二类调查数据为样本,基于Google Earth Engine云平台Landsat 8 OLI影像,结合植被因子、纹理特征以及K-T变化为自变量,构建了多元线性回归和随机森林的建模方法,建立了森林蓄积量反演模型.以宜良县云南松为研究对象,运用Landsat8 OLI遥感影像数据结合地面角规控制样地调查数...     

基于主成分回归及遥感的贵阳市林分郁闭度估测

以贵阳市为研究区,以Landsat-8 OLI为遥感信息源,通过偏相关分析,选择了与郁闭度相关的12个遥感因子作为自变量因子。为克服自变量因子间严重的多重共线性和模型的不稳定性,采用主成分回归分析法建立郁闭度估测模型,并对模型进了检验及精度验证。结果表明,回归方程调整后的R2=0.756,模型的拟合效果较好,说明模型在数学上是可行的;利用15个实测样地对模型进行精度验证,估测精度的平均水平为78.80%,说明该模型可为区域林分郁闭度估测提供参考。     

基于地形校正后Landsat 8的土壤重金属反演研究

为快速地了解玛纳斯流域土壤中重金属的污染状况和潜在生态风险，在Landsat 8 OLI的基础上引入DEM数据进行地形校正，同时对地形校正后的反射率进行倒数、导数和对数等数学变换，从每种变换中筛选出与土壤各重金属相关性最高的波段构建土壤各重金属PLSR预测模型，并对研究区土壤重金属分布情况进行探索，并利用生态风险评价方法对研究区进行预警。结果表明：在Landsat 8的基础上，引入DEM数据对反射率进行地形校正，以B1波段反射率和重金属Cu为例，经过地形校正后的反射率值与实测土壤表层Cu含量的R²从0.46提高至0.52，表明地形校正后的表观反射率能够更好地反映土壤重金属状况；利用土壤各重金属的最佳预测模型分别反演相应的土壤重金属含量，并引入土壤重金属生态风险指数用于评价研究区的土壤重金属生态风险，研究表明土壤重金属风险等级总体上呈现从西南方向至东北方向逐渐减弱的趋势，其生态风险排序为恢复区(C区)＞退化区(B区)＞湖泊入湖口(A区)；为了验证基于遥感的土壤重金属生态风险预警的预测精度，将研究区土壤重金属含量实测数据也通过重金属生态风险指数进行计算，两者结果较为一致，表明可以用遥感的手段来反演该研究区的重金属分布情况，同时研究区土壤重金属污染总体上处于轻警以上级别，生态服务功能已开始退化，应该加强对该地区的重金属污染进行治理。     

Canadian Prairie Rangeland and Seeded Forage Classification Using Multiseason Landsat 8 and Summer RADARSAT-2

Rangeland and seeded forage in Canada’s Prairie provinces represent productive landscapes that provide multiple ecosystem services. Past efforts to map these resources at regional scales have not achieved consistently high accuracies as they are spatially variable in both ecology and management. In particular, Agriculture and Agri-food Canada needs to distinguish these land use classes from each other and from cropland in its annual national agricultural land cover inventory. Given the potential to distinguish these classes based on seasonal phenological differences, this study used multi-season Landsat 8 top-of-atmosphere reflectance data and derived vegetation and phenological indices, as well as mid-summer RADARSAT-2 data in random forest classification of two ecoregions in Alberta and Manitoba. Classification accuracy was compared for single and multi-date Landsat 8 variables, the vegetation index and phenological variable groups, RADARSAT-2 VV and VH backscatter intensity, and combined datasets. Variable importance analysis showed that spring Landsat 8 reflectance generally contributed most to class discrimination, but accuracy improved with the addition of Landsat 8 data from the other seasons. Vegetation indices and phenological variables produced similar accuracies and were deemed to not warrant the additional processing effort to derive them. RADARSAT-2 VH backscatter was the most important variable for the Manitoba study area, which is wetter with more vegetation structure variability than the Alberta study area. Backscatter intensity significantly increased overall accuracy when it was combined with one or two-season Landsat 8 data. The best overall accuracy was achieved using the three seasons of Landsat 8 and mid-summer RADARSAT-2 data, but it was not significantly better than that for two season Landsat 8 + RADARSAT-2. The methods presented in this paper provide a process for accurate and efficient classification of seeded forage, rangeland and cropland that can be applied over large areas in operational agricultural land cover inventory.     

基于林分结构响应的PALSAR森林结构参数估测

为提高ALOS PALSAR数据估测森林结构参数的精度,引入代表林分结构复杂程度的调整熵值(ENTadj)参与估测,以消除林分结构对雷达后向散射系数的干扰。首先利用野外样地实测的树高计算林分的调整熵值,与Landsat8 OLI第6波段建立线性回归模型,获得基于像元的调整熵值。一般森林结构参数与ALOS PALSAR后向散射系数之间的关系可以用对数模型模拟。引入基于像元的调整熵值作为自变量对原始对数模型进行改进,分别对林分平均高、林分平均胸径、林分蓄积量建立了3种形式的改进模型。利用原始模型和改进模型分别对杉木林、马尾松林、阔叶林和针阔混交林的上述森林结构参数进行估测。最后比较模型拟合精度筛选出3项森林结构参数在各类森林中的最优模型,共计12个。结果表明:考虑林分结构干扰后,雷达估测森林结构参数模型的拟合精度R2均得到了提高。马尾松林各项森林结构参数模型的拟合度提高最大。精度检验结果表明:林分平均高估测精度(RMSE为0.74~2.51 m)、林分平均胸径估测精度(RMSE为2.61~5.61 cm)和林分蓄积量估测精度(RMSE为21.71~30.92 m3/hm2)都比较理想。本研究探讨了林分结构信息应用于合成孔径雷达后向散射系数反演森林结构参数方面的潜力,提高了光学数据结合雷达数据估算森林结构参数的能力。      

Quantification of carbon sequestration by urban forest using Landsat 8 OLI and machine learning algorithms in Jodhpur,India

Urban forests play a significant role in carbon cycling. Quantification of Aboveground Biomass (AGB) is critical to understand the role of urban forests in carbon sequestration. In the present study, Machine learning (ML) based regression algorithms (SVM, RF, kNN and XGBoost) have been taken into account for spatial mapping of AGB and carbon for the urban forests of Jodhpur city, Rajasthan, India, with the aid of field-based data and their correlations with spectra and textural variables derived from Landsat 8 OLI data. A total of 198 variables were retrieved from the satellite image, including bands, Vegetation Indices (VIs), linearly transformed variables, and Grey Level Co-occurrence textures (GLCM) taken as independent input variables further reduced to 29 variables using Boruta feature selection method. All the models have been compared where with RF algorithm, R² = 0.83, RMSE = 16.22 t/ha and MAE = 11.86 t/ha. For kNN algorithm R² = 0.77, RMSE = 28.04 t/ha and MAE = 24.24 t/ha and SVM where R² = 0.73, RMSE = 89.21 t/ha and MAE = 74.22 t/ha and the best prediction accuracy has been noted with XGBoost algorithm (R² = 0.89, RMSE = 14.08 t/ha and MAE = 13.66 t/ha) with predicted AGB as 0.51−153.76 t/ha. The study indicates that ML-based regression algorithms have great potential over other linear and multiple regression techniques for spatial mapping of AGB and carbon of urban forests for arid regions.