Coupling high-resolution satellite imagery with ALS-based canopy height model and digital elevation model in object-based boreal forest habitat type classification

We developed a classification workflow for boreal forest habitat type mapping. In object-based image analysis framework, Fractal Net Evolution Approach segmentation was combined with random forest classification. High-resolution WorldView-2 imagery was coupled with ALS based canopy height model and digital terrain model. We calculated several features (e.g. spectral, textural and topographic) per image object from the used datasets. We tested different feature set alternatives; a classification accuracy of 78.0% was obtained when all features were used. The highest classification accuracy (79.1%) was obtained when the amount of features was reduced from the initial 328 to the 100 most important using Boruta feature selection algorithm and when ancillary soil and land-use GIS-datasets were used. Although Boruta could rank the importance of features, it could not separate unimportant features from the important ones. Classification accuracy was bit lower (78.7%) when the classification was performed separately on two areas: the areas above and below 1 m vertical distance from the nearest stream. The data split, however, improved the classification accuracy of mire habitat types and streamside habitats, probably because their proportion in the below 1 m data was higher than in the other datasets. It was found that several types of data are needed to get the highest classification accuracy whereas omitting some feature groups reduced the classification accuracy. A major habitat type in the study area was mesic forests in different successional stages. It was found that the inner heterogeneity of different mesic forest age groups was large and other habitat types were often inside this heterogeneity.     

Comparing supervised and unsupervised multiresolution segmentation approaches for extracting buildings from very high resolution imagery

Although multiresolution segmentation (MRS) is a powerful technique for dealing with very high resolution imagery, some of the image objects that it generates do not match the geometries of the target objects, which reduces the classification accuracy. MRS can, however, be guided to produce results that approach the desired object geometry using either supervised or unsupervised approaches. Although some studies have suggested that a supervised approach is preferable, there has been no comparative evaluation of these two approaches. Therefore, in this study, we have compared supervised and unsupervised approaches to MRS. One supervised and two unsupervised segmentation methods were tested on three areas using QuickBird and WorldView-2 satellite imagery. The results were assessed using both segmentation evaluation methods and an accuracy assessment of the resulting building classifications. Thus, differences in the geometries of the image objects and in the potential to achieve satisfactory thematic accuracies were evaluated. The two approaches yielded remarkably similar classification results, with overall accuracies ranging from 82% to 86%. The performance of one of the unsupervised methods was unexpectedly similar to that of the supervised method; they identified almost identical scale parameters as being optimal for segmenting buildings, resulting in very similar geometries for the resulting image objects. The second unsupervised method produced very different image objects from the supervised method, but their classification accuracies were still very similar. The latter result was unexpected because, contrary to previously published findings, it suggests a high degree of independence between the segmentation results and classification accuracy. The results of this study have two important implications. The first is that object-based image analysis can be automated without sacrificing classification accuracy, and the second is that the previously accepted idea that classification is dependent on segmentation is challenged by our unexpected results, casting doubt on the value of pursuing ‘optimal segmentation’. Our results rather suggest that as long as under-segmentation remains at acceptable levels, imperfections in segmentation can be ruled out, so that a high level of classification accuracy can still be achieved.     

一种 ReliefF 和随机森林模型组合的多波束 海底底质分类方法

海底底质的快速探测和精细划分对海洋工程建设 、海洋资源开发等具有重要意义。多波束探测是目前声学底质遥测 的有效手段之一, 通常提取多波束反向散射强度图像和地形数据中的多维特征结合分类器进行底质分类。一方面, 若特征空 间维数过高, 分类效率会显著降低; 另一方面, 个别特征容易放大原始数据处理过程中仍存留的异常现象。针对这一问题, 本文提出了一种结合 Re1iefF 算法和随机森林 (Random Forest, RF) 算法的多波束底质分类方法。提取反向散射强度和地形 共 16 维特征, 利用Re1iefF 算法进行特征筛选, 排除低相关性特征, 降低特征空间维数, 结合采样点数据进行模型训练以构 建多波束底质分类模型。试验结合随机森林算法对未经特征筛选 、经主成分分析 (Principa1 Component Ana1ysis, PCA) 特征 优化后的特征进行分类实验作为对比。本文方法 Kappa 系数达到 85%, 分类总精度高于 90%, 精度具有明显优势, 耗时也 比较短。可见, 本文提出的结合 Re1iefF 和随机森林模型的多波束底质分类方法可以在保证分类精度的同时对多维特征进行 优化, 有效地提高了分类效率, 可对海底底质分类研究提供参考。     

基于随机森林模型的黄河流域城市建设用地结构时空演化及其驱动机制研究

城市是区域人口、产业等要素集聚的主要空间载体,系统揭示黄河流域城市建设用地结构演化及其驱动机制对于深入推进黄河流域生态保护和高质量发展战略落实具有重要意义。基于黄河流域2007—2019年城市建设用地数据,论文运用信息熵模型分析了黄河流域城市建设用地结构的时空演化特征,并用随机森林模型探讨其驱动机制。结果表明：(1) 2007—2019年,黄河流域城市建设用地面积快速增长,年均增速达5.34%,但城市建设用地结构信息熵不断降低,经历了“高位无序—相对稳定—低位有序”的变化过程;(2)黄河上、中、下游城市建设用地规模增速依次加大,年均增速分别达到3.91%、5.06%、6.22%;中游地区城市建设用地结构信息熵下降最多,上游地区和下游地区降幅相近;(3)黄河流域各类建设用地主要配置在省会及经济发达城市;黄河流域城市用地结构信息熵存在空间异质性,省会城市及经济发达城市建设用地结构信息熵存在降低趋势,而欠发达城市的信息熵逐步增大;(4)城市用地规模和城市人口规模对城市建设用地结构演化的作用最强,产业结构升级和财政支出规模的作用强度次之,经济发展水平和政策支持力度的作用较小,上中下游影响因子的...     

Spatial Differentiation of the Coupling Characteristics of Soil Carbon and Nitrogen on Mulberry Plantations in China

Soil is the most important carbon pool of the mulberry plantation ecosystem, so understanding the characteristics of the soil carbon pool in mulberry plantations provides an important basis for the research of carbon sinks in economic forest ecosystems and farmland ecosystems. In order to explore the spatial differentiation pattern of the relationship between carbon and nitrogen in mulberry plantation soil, this study analyzed the organic carbon content and total nitrogen content of the surface soil layer (0-20 cm) and the subsurface soil layer (20-40 cm) of 475 mulberry plantations in five major regions of China, Southwest China (SWC), Central South China (CSC), East China (EC), North China (NC), and Northwest China (NWC). The research showed seven key aspects of this system. (1) The soil organic carbon of mulberry plantations was significantly different at the two soil depths. The average content of organic carbon in the surface layer of mulberry plantation soil was 10.71±7.01g kg^?1, which was 37.13% higher than that of the subsurface layer. (2) The soil organic carbon of mulberry plantations had significant differences in spatial differentiation, which was manifested as SWC>CSC>EC>NC>NWC. (3) The total nitrogen content in mulberry plantation soil had significant responses to the region, the soil layer depth, and the interaction between the region and soil layer depth. Among the regions, NWC had no significant difference between the surface layer and subsurface layer of the soil. EC had the maximum difference in total nitrogen content, with the total nitrogen content in the surface soil layer being 56.68% higher than that of the subsurface soil layer. The total nitrogen contents of the surface soil layers in the SWC and NC were 34.27% and 20.79% higher than those of the respective subsurface soil layers. (4) The mulberry plantation soil C/N ratios had a significant response to regional differences, as NWC>SWC> EC>CSC>NC, but this ratio had no significant response to soil depth. (5) Soil pH had significant spatial differentiation in relation to soil organic carbon and total nitrogen content in mulberry plantations. NWC had no significant correlation between pH and organic carbon or total nitrogen content, while CSC had a significant positive correlation between pH and both soil organic carbon and total nitrogen content. Other regions showed significant negative correlations between pH and both organic carbon and total nitrogen content. (6) There was a significant negative correlation between the C/N ratio of the surface soil layer and pH in mulberry plantations, which was mainly contributed by SWC, while the other regions’ surface soil layers had no significant correlations between C/N ratio and pH. (7) There was no significant correlation between the C/N ratio and pH in the subsurface soil layer in mulberry plantations. These results reveal that in either the research on mulberry plantation carbon pools or the innovation of green and low-carbon planting technology in mulberry plantations, the spatial differentiation characteristics of soil must be considered. Furthermore, the spatial differentiation of soil organic carbon can be used as the basic foundation for the planning and design of mulberry afforestation or ecological restoration projects.     

Developing a gross primary production model for coniferous forests of northeastern USA from MODIS data

Accurate estimation of ecosystem carbon fluxes is crucial for understanding the feedbacks between the terrestrial biosphere and the atmosphere and for making climate-policy decisions. A statistical model is developed to estimate the gross primary production (GPP) of coniferous forests of northeastern USA using remotely sensed (RS) radiation (land surface temperature and near-infra red albedo) and ecosystem variables (enhanced vegetation index and global vegetation moisture index) acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor. This GPP model (called R-GPP-Coni), based only on remotely sensed data, was first calibrated with GPP estimates derived from the eddy covariance flux tower of the Howland forest main tower site and then successfully transferred and validated at three other coniferous sites: the Howland forest west tower site, Duke pine forest and North Carolina loblolly pine site, which demonstrate its transferability to other coniferous ecoregions of northeastern USA. The proposed model captured the seasonal dynamics of the observed 8-day GPP successfully by explaining 84–94% of the observed variations with a root mean squared error (RMSE) ranging from 1.10 to 1.64 g C/m²/day over the 4 study sites and outperformed the primary RS-based GPP algorithm of MODIS.     

Impact of feature selection on the accuracy and spatial uncertainty of per-field crop classification using Support Vector Machines

Crop mapping is one major component of agricultural resource monitoring using remote sensing. Yield or water demand modeling requires that both, the total surface that is cultivated and the accurate distribution of crops, respectively is known. Map quality is crucial and influences the model outputs. Although the use of multi-spectral time series data in crop mapping has been acknowledged, the potentially high dimensionality of the input data remains an issue. In this study Support Vector Machines (SVM) are used for crop classification in irrigated landscapes at the object-level. Input to the classifications is 71 multi-seasonal spectral and geostatistical features computed from RapidEye time series. The random forest (RF) feature importance score was used to select a subset of features that achieved optimal accuracies. The relationship between the hard result accuracy and the soft output from the SVM is investigated by employing two measures of uncertainty, the maximum a posteriori probability and the alpha quadratic entropy. Specifically the effect of feature selection on map uncertainty is investigated by looking at the soft outputs of the SVM, in addition to classical accuracy metrics. Overall the SVMs applied to the reduced feature subspaces that were composed of the most informative multi-seasonal features led to a clear increase in classification accuracy up to 4.3%, and to a significant decline in thematic uncertainty. SVM was shown to be affected by feature space size and could benefit from RF-based feature selection. Uncertainty measures from SVM are an informative source of information on the spatial distribution of error in the crop maps.     

Performance of dense digital surface models based on image matching in the estimation of plot-level forest variables

Recent research results have shown that the performance of digital surface model extraction using novel high-quality photogrammetric images and image matching is a highly competitive alternative to laser scanning. In this article, we proceed to compare the performance of these two methods in the estimation of plot-level forest variables. Dense point clouds extracted from aerial frame images were used to estimate the plot-level forest variables needed in a forest inventory covering 89 plots. We analyzed images with 60% and 80% forward overlaps and used test plots with off-nadir angles of between 0° and 20°. When compared to reference ground measurements, the airborne laser scanning (ALS) data proved to be the most accurate: it yielded root mean square error (RMSE) values of 6.55% for mean height, 11.42% for mean diameter, and 20.72% for volume. When we applied a forward overlap of 80%, the corresponding results from aerial images were 6.77% for mean height, 12.00% for mean diameter, and 22.62% for volume. A forward overlap of 60% resulted in slightly deteriorated RMSE values of 7.55% for mean height, 12.20% for mean diameter, and 22.77% for volume. According to our results, the use of higher forward overlap produced only slightly better results in the estimation of these forest variables. Additionally, we found that the estimation accuracy was not significantly impacted by the increase in the off-nadir angle. Our results confirmed that digital aerial photographs were about as accurate as ALS in forest resources estimation as long as a terrain model was available.     

基于MAPGIS地理信息平台城镇地籍数据库建库探讨——以神农架林区松柏镇为例

以MapGIS地理信息平台为基础、以SQL SERVER2000数据库为后台,神农架林区松柏镇为研究对象,阐述了城镇地籍建库的方式、建库流程,再列举了在建库过程中遇到常见问题及解决方案,从而建成符合城镇地籍数据库标准MAPGIS地理信息平台的城镇地籍数据库成果。最后介绍了MapGIS地理信息平台的下数据库实现功能如地籍数据格式转换功能、图形与信息检索查询、ACESS数据导入与影像挂接、统计分析与成果输出、日常办公及查询统计等功能。     

Fallow,labour and livelihood in shifting cultivation: implications for deforestation in northern Lao PDR

Shifting cultivation is often blamed for deforestation in tropical upland areas. Based on a case study of three villages in northern Lao PDR, this paper combines household surveys with a remote sensing based analysis of forest cover, covering the period 1989–1999, in order to analyse changes in shifting cultivation practices and livelihood strategies and the impact of these on deforestation. Due to population pressure and relocation of villages, easily accessed land is increasingly scarce and fallow periods have been shortened during the 1990s. A net annual deforestation of about 1% was found in the area during the study period. This deforestation rate reflects shorter fallow periods in secondary forests rather than encroachment on mature forests, which are not used for cultivation by the farmers in the three villages. Farmers rate scarce labour as a major constraint on shifting cultivation; nonetheless, a tendency towards lower labour input with shorter fallow periods is observed, contradicting conventional intensification theory. Livelihoods are diversifying through the establishment of plantations, cultivation of wet rice and adoption of animal husbandry, but given the socio-economic conditions in the area, shifting cultivation is likely to remain the most suitable farming system in the near future.