CAJ | 학술논문

耕地质量监测是保障耕地资源的永续利用,实现耕地产能提升、加强耕地资源的管理、保护、合理利用的重要措施,对实现持续粮食安全具有重要意义.该文提出了基于空间模拟退火算法的耕地质量布样优化方法,以空间模拟退火算法为基础生成一组最优样本,构成基础监测网络,在此基础上,通过多期耕地等级成果数据提取属性发生变化的分等因素和对应发生变化的区域,生成潜在变化区,并结合研究区实际情况辅以专家知识和异常监测点,对基础样本点进行增加、删除、替换等优化操作,生成最终监测样点.以北京市大兴区为例,最终确定布设55个监测样点,结果表明,该方法布设的样点在耕地质量预测方面的精度高于传统的随机抽样和分层抽样方法,能有效地预测县域耕地质量并监控耕地质量的变化情况.
경지질량감측시보장경지자원적영속이용,실현경지산능제승、가강경지자원적관리、보호、합리이용적중요조시,대실현지속양식안전구유중요의의.해문제출료기우공간모의퇴화산법적경지질량포양우화방법,이공간모의퇴화산법위기출생성일조최우양본,구성기출감측망락,재차기출상,통과다기경지등급성과수거제취속성발생변화적분등인소화대응발생변화적구역,생성잠재변화구,병결합연구구실제정황보이전가지식화이상감측점,대기출양본점진행증가、산제、체환등우화조작,생성최종감측양점.이북경시대흥구위례,최종학정포설55개감측양점,결과표명,해방법포설적양점재경지질량예측방면적정도고우전통적수궤추양화분층추양방법,능유효지예측현역경지질량병감공경지질량적변화정황.
M Monitoring points in country area are the foundation to reflect changes of cultivated land quality, which directly affect the result of farmland grading and its accuracy. Through the monitoring network for cultivated land quality in county area, the distribution and changing trend of the cultivated land quality can be reflected. Besides, the quality of non-sampled locations should also be estimated with the data of sampling points. Due to the correlation among spatial samples, the traditional methods such as simple random sampling, stratified sampling and systematic sampling are inefficient to accomplish the task above. Thus, we propose a new spatial sampling and optimizing method based on the spatial simulated annealing (SSA). This paper presents a pre-processing method to determine the number of sampling points, including preprocessing the data of cultivated land quality before sampling, exploring the spatial correlation and spatial distribution pattern of cultivated land quality, and computing the appropriate quantity of sampling points by analyzing the change trend of sampling number and sampling precision, and on this basis we propose the extended spatial simulated annealing method to optimize spatial sampling design for obtaining the minimal Kriging variance. The main steps for computing the optimal sampling design can now be summarized as follows: 1) calculate the semi-variogram of cultivated land quality and determine the parameters of ordinary Kriging interpolation; 2) identify the quantity of samples, choose a set of cultivated land map spots randomly as an initial design, and compute the associated fitness function; 3) given one design,construct a candidate new sampling design by random perturbation; 4) compute the fitness function for the new design, and if it is smaller than or equal to that for the original design, accept the original design, or else accept the new design with an acceptance probability. If the new design is accepted, the estimated point (j) is returned to zero, or else increased by 1; 5) ifj is smaller than or equal to a threshold value of continuous rejections, increasei (representing monitoring point) by 1, or else stop the iteration and current design is the best. Designs by simulated annealing that reduce the average Kriging standard error are always accepted, and designs that worsen the interpolation effect are accepted with a certain probability, which decreases to zero as iterations proceed. However, there are integrated factors such as soil organic matter content, topsoil texture, profile pattern, salinization which affect arable land quality change over time and space, and are taken as potential change factors to detect potential change areas. Under the guidance of expert knowledge, the sampling points are set up through spatial simulated annealing algorithm and adjusted based on potential change areas, rivers, roads and abnormal monitoring points. We illustrate this new method using Daxing District, Beijing City as a case study. Spatial overlay analysis of potential change factors and geostatistics method of GIS are employed to test this method. The spatial variability of cultivated land quality is simulated using natural quality indices and a specified number of network locations is defined which can be used to adequately predict the quality of cultivated land. The experimental results of Daxing District, Beijing City show that 55 monitoring reference sample units are finally deployed, and the average ordinary Kriging standard error with this method is 131.78, which is smaller than the simple random sampling (134.97) and stratified sampling (134.93) when the quantity of samples is the same. Besides, sampling accuracy and cost are both considered and reach a certain balance in this method. This method is suited for counties which have carried out several surveys of cultivated land quality, or counties whose grading factors have certain changes. Besides, it is also suitable for counties which have some prior knowledge but never have conducted a survey of cultivated land quality.