关键词:耕地质量;监测;抽样方法;精度比较;黄骅市 Abstract Cultivated land monitoring is an essential way to obtain information about the change in land quality and productivity in time. A scientific and rational sampling method can ensure monitoring precision and reduce monitoring costs. We took Huanghua in Hebei as the study area,illustrated the implementation steps of different sampling methods in detail; drew samples in different sampling methods based on cultivated land quality condition; listed the formulas of sample variance and maximum relative error of the sampling mean;calculated and compared the maximum relative error of the sampling mean and the relative sampling precision of different sampling methods. Our results show that sampling precision increases from fast to slowly when the sample size increases. In the matter of sampling precision,compared with the simple random sampling,system stratification sampling and system isometric sampling have overwhelming advantages which increase obviously when the sample size increases. The grade stratification sampling has a little greater advantage which changes smoothly when the sample size increases. The town stratification sampling has a razor-thin advantage,but the grid stratification sampling does not have any advantage,instead,it may cause precision loss. When sampling precision is the same,the sample sizes of different sampling methods are significantly different,so on the condition that different sampling methods achieve the same sampling precision,a reasonable one can obviously reduce sample size and save monitoring costs. This study aims to provide effective methodological guidance for the precision estimation and comparison of different sampling methods,which helps to select the most reasonable sampling method in cultivated land monitoring.
Keywords:cultivated land quality;monitoring;sampling method;comparison of precision;Huanghua City -->0 PDF (1174KB)元数据多维度评价相关文章收藏文章 本文引用格式导出EndNoteRisBibtex收藏本文--> 殷守强, 王鑫, 贺文龙, 门明新, 张利. 耕地质量监测中不同抽样方法的精度比较——以河北省黄骅市为例[J]. , 2016, 38(11): 2049-2057 https://doi.org/10.18402/resci.2016.11.04 YINShouqiang, WANGXin, HEWenlong, MENMingxin, ZHANGLi. Comparison of precision among different sampling methods in cultivated land quality monitoring: a case study of Huanghua City in Hebei Province, China[J]. 资源科学, 2016, 38(11): 2049-2057 https://doi.org/10.18402/resci.2016.11.04
为了比较不同的抽样方法在相同样本容量下的抽样精度,通过上一章节的精度计算公式,计算得到简单随机抽样、系统等距抽样、等别分层抽样、乡镇分层抽样和系统分层抽样这5种抽样方法在样本容量分别为20、40、60、80、100、200、300和400个的抽样均值的最大相对误差(表1)。因为网格分层抽样法须在ArcGIS软件中通过一系列操作才能得到监测网格的数量,即网格分层抽样法根据实际抽样结果才能确定样本容量,所以网格分层抽样无法像其他抽样方法在抽样前可以精确控制样本的容量,这导致了网格法分层抽样与其他抽样方法在确定的样本容量上的不同。因此,本文把网格法分层抽样中不同网格边长对应的样本容量以及抽样均值的最大相对误差单独列出(表2)。 Table 1 表1 表15种抽样方法的不同样本容量下抽样均值的最大相对误差 Table 1The maximum relative error of sample mean in different sample sizes among 5 sampling schemes
样本容量 /个
简单随机抽样 /%
系统等距抽样 /%
等别分层抽样 /%
乡镇分层抽样 /%
系统分层抽样 /%
20
4.90
2.20
2.25
4.48
1.09
40
3.36
1.21
1.48
3.08
0.56
60
2.72
0.95
1.19
2.47
0.38
80
2.34
0.60
1.04
2.17
0.25
100
2.09
0.42
0.91
1.93
0.15
200
1.46
0.22
0.63
1.31
0.08
300
1.19
0.17
0.51
1.06
0.06
400
1.03
0.13
0.43
0.91
0.05
新窗口打开 用SPSS软件按照每种抽样方案重复进行100次抽样实验验证计算结果,计算每次实验的样本均值,进而计算该值与抽样均值的相对误差。经过验证,发现样本均值的相对误差均小于该抽样方法对应的抽样均值的最大相对误差,表明表1和表2中计算的抽样均值最大相对误差精准可靠,符合实际,能够准确预测实际的抽样限差。6种抽样方案的样本容量和对应均值最大相对误差的散点分布比较见图2。 Table 2 表2 表2网格分层抽样方法中不同样本容量下的均值最大相对误差 Table 2The maximum relative error of sample mean in different sample sizes among grid stratification sampling
网格边长 /km
网格法样本容量 /个
最大相对误差 /%
2
399
0.98
3
185
1.52
4
114
2.13
5
74
2.59
6
59
2.95
7
40
3.41
8
30
4.12
9
24
4.57
10
20
5.18
新窗口打开 从表1、表2和图2中可以看出,不同抽样方法的均值最大相对误差随着样本容量的增加而减少,速度由快变慢。但另一方面,由于抽样成本也会随着样本容量的增加线性地提高,所以要找出既能满足精度要求又能节约成本的抽样方法。样本容量相同的情况下,网格分层抽样、简单随机抽样和乡镇分层抽样的抽样均值的最大相对误差大体相当,表明这3种抽样方法的抽样精度差别不大;而等别分层抽样、系统等距抽样、系统分层抽样的均值最大相对误差依次减小,均比另外3种抽样方法小,表明这3种抽样方法相对于另外3种抽样方法的抽样精度依次提高,监测样点对总体的估计精度和代表水平依次增加。 显示原图|下载原图ZIP|生成PPT 图26种抽样方法中不同样本容量下的均值最大相对误差比较 -->Figure 2Comparison of the maximum relative error in different sample sizes among 6 sampling methods -->
把相同样本容量下简单随机抽样的抽样均值最大相对误差与某种抽样方法的抽样均值最大相对误差的比值称为此样本容量下该抽样方法的相对抽样精度。某种抽样方法的相对抽样精度越大,表明在相同样本容量下,该抽样方法比简单随机抽样对总体的估计误差和造成的精度损失越小,对总体的估计精度和代表性越高,抽样效率也越高。依据相对抽样精度的含义,可知某种抽样方法的相对抽样精度均是以简单随机抽样方法为基准计算的,所以简单随机抽样方法的相对抽样精度可看作始终为1的常数,只需用折线图展现其他抽样方法相对抽样精度的变化情况。通过计算除简单随机抽样以外的其他5种抽样方法的相对抽样精度,绘制不同样本容量下的相对抽样精度变化曲线(图3),更加直观地比较不同抽样方法的抽样精度和抽样效率。 显示原图|下载原图ZIP|生成PPT 图35种抽样方法中不同样本容量下的相对抽样精度比较 -->Figure3Comparison of the relative sampling precision in different sample sizes among 5 sampling methods -->
从图3中可以看出,在样本容量小于400的范围内,系统分层抽样的相对抽样精度最高,其次为系统等距抽样,而且两者均远远大于1,并随样本量的不断提高而提升,速度由快变慢,这说明与简单随机抽样相比,这两种抽样方法在提高抽样精度方面的优势极大,并随着样本量的增加而扩大。而等别分层抽样的相对抽样精度由2.18提高到2.40,变化平稳,说明与简单随机抽样相比,该种抽样方法在提高抽样精度方面的优势较大并且平稳。乡镇分层抽样的相对抽样精度在1.08到1.13之间,曲线变化平稳,说明与简单随机抽样相比,该种抽样方法在提高抽样精度方面的优势不大。网格分层抽样的相对抽样精度在0.89和1.05之间,说明与简单随机抽样相比,该种抽样方法在提高抽样精度方面不但没有优势,反而可能造成精度的损失。 根据样本量计算公式、表1和表2中的统计结果,得到在95%置信度水平下抽样均值最大相对误差等于1%时的最小样本容量(表3)。对于缺少该值的抽样方法,可通过线性插值的方法获取。 由表3可知,当抽样均值最大相对误差为1%的时,系统分层抽样、系统等距抽样、等别分层抽样、简单随机抽样、乡镇分层抽样和网格法抽样所需样本容量分别为23、56、84、218、337和386个,相应的抽样比分别为0.61%、1.48%、2.22%、5.77%、8.92%和10.21%。其中,简单随机抽样与等别分层抽样、系统等距抽样和系统分层抽样所需的样本量的比值分别是2.6、3.9和9.46。可见,不同的抽样方法在满足相同的监测精度时所需的样本量差异显著,选择合理的抽样方法在满足监测精度的同时,可以明显地减少样本量,从而节约监测成本。 Table 3 表3 表36种抽样方法满足精度要求的最小样本容量 Table 3The minimum sample sizes meeting the precision requirement among 6 sampling methods
(1)他人的研究一般是在预估抽样精度未知的情况下布设监测样点,再通过辅助变量或者研究者的经验进行样点的补充和修正,主观性较强,对监测样点布设的合理性解释较差,可能提高监测的成本或达不到监测的精度。本研究定量计算了不同抽样方法的样本容量与抽样精度之间的关系,科学地预估不同布样方案的抽样误差,并通过比较不同抽样方案的估计精度,确定合理的抽样方法和样本容量,避免了布设监测样点的盲目性,提高了样点的全局代表性,在保证抽样精度的前提下最大化地节约了抽样成本。 (2)本研究提出的抽样方法不仅在本区域可行,在其他区域也具有可操作性,旨在从统计学角度对总体的监测精度进行估计,从而选取满足精度要求的抽样方法,为耕地质量监测样点布设提供普遍适用的方法指导。可根据抽样方法进行抽取监测样点,再根据抽取的样点进行全区的插值,然后比较不同方法的插值精度,这是下一步需要研究的方向。另外,本研究仅仅考虑了耕地自然等指数的估计精度,但是由于耕地自然等指数相同的耕地单元,其分等因素值存在不同的情况,所以为了精确监测耕地分等因素等别变化情况,进行抽样时还须考虑各分等因素的抽样精度。对于抽样方法的选取,本文只提出了6种,实际上远不止这些。因此,可根据区域实际状况探索新的抽样方法。 The authors have declared that no competing interests exist.
[YangL,Zhu AX,Qin CZ,et al. A purposive sampling design method based on typical points and its application in soil mapping [J]. ,2010,29(3):279-286.] [本文引用: 1]
[Lu YJ,Qin CZ,Qiu WL,et al. Sensitivity analysis of soil property parameter in typical-sample-based prediction model of digital soil mapping [J]. ,2011,31(12):1549-1554.]
[Sun XL,Wang HL,NingY.Application of sampling method based on representativeness grades of sampling points in topsoil organic matter mapping within hilly area [J]. ,2014,46(3):439-445.]
[Han ZW,HuangW,LuoY,et al. Application of simulated annealing method and neural network schemes based on road distribution [J]. ,2015,26(3):891-900.] [本文引用: 1]
[Zheng XQ,Wang JY.Study on the Voronoi method for the disposition of dynamic monitoring points of urban land price [J]. ,2003,20(4):263-266.] [本文引用: 1]
[Chen SY,Qu FT,Ni SX,et al. Methodology for establishment of urban land price monitoring systems based on GIS:Taking Zhenjiang City of Jiangsu Province as an example [J]. ,2005,27(6):65-70.]
[Wang LC,YangX,Song YH,et al. Study on interactive coupling mechanism of industrial structure and environmental quality:A case study of Gansu Province [J]. ,2012,31(5):103-106.] [本文引用: 1]
[Ma RH,Guo LN,Zhang FR,et al. Methods for setting up farmland grade monitoring points from standard plots in Hebei,Henan and Hubei provinces [J]. ,2009,31(11):1957-1966.] [本文引用: 1]
[Sun YB,Wu KN,Hu XT,et al. Layout method for monitoring quality level of arable land based on combination of potential index [J]. ,2013,29(4):245-254.] [本文引用: 1]
[Hu XT,Wu KN,Ma JH,et al. Control points for quality of arable land quality monitoring in Daxing district,Beijing [J]. ,2012,34(10):1891-1897.] [本文引用: 1]
[WanQ,Shang YM,FengR,et al. Study on location sampling for monitoring the quality of arable land based on variation functions:Case studies in Zhongjiang county and Daxing District [J]. ,2012,26(8):80-86.] [本文引用: 1]
[Yang JY,TangS,Yun WJ,et al. Sampling method for moni-toring classification of arable land in county area based on Kriging estimation error [J]. ,2013,29(9):223-230.] [本文引用: 1]
[XiangH,Kong XB,Chen PX,et al. Establishing monitoring samples of arable land at the county level:A case study of Dalate District in Inner Mongolia [J]. ,2014,36(6):1203-1210.] [本文引用: 1]
[Yu SQ,Zhang BB,Kong XB,et al. Layout method for monitoring sample point of arable land quality level based on combination of factors [J]. ,2014,30(24):288-297.]
[Song YH,Mao HB,Wang LC,et al. Sampling method of cultivated land quality monitoring based on the monitoring control region spatial variability [J]. ,2014,33(6):131-136.]
[Liu YH,WanP,Peng ZT,et al. Dynamic monitoring sampling point arrangement system of cultivated land quality in southwestern mountainous area-a case study of Fengdu in Chong-qing [J]. ,2014,45(4):609-614.] [本文引用: 1]
[Cai LB,Kong XB,Duan JN,et al. Layout for monitoring quality of arable land of China:Inspiration from overseas experience [J]. ,2014,30(14):192-197.] [本文引用: 1]
[Yang JY,Yue YL,Song HR,et al. Sampling and optimizing methods of cultivated land quality based on spatial simulated annealing algorithm [J]. ,2015,31(20):253-261.] [本文引用: 1]
[Zhu JX,Xu BG,Zhang LY.Optimization layout method of monitoring sample points of cultivated land quality based on semi-variance analysis and grade combination [J]. ,2015,31(19):254-261.] [本文引用: 1]
[Ma JH.Study on the cultivated land quality monitoring sample point stationing method based on the variation function and grade [J]. ,2015,(5):111-116.] [本文引用: 1]
[Yang JY,Yue YL,Song HR,et al. Sampling distribution method for monitoring quality of arable land in county area based on spatial balanced [J]. ,2015,31(24):274-280.] [本文引用: 1]