Use of high resolution imagery and ground survey data for estimating crop areas in Mengcheng county, China

35th International Symposium on Remote Sensing of Environment

地球科学;生态环境科学

Kerdiles, H.^1 ; Dong, Q.^2 ; Spyratos, S.^1 ; Gallego, F.J.^1

Joint Research Centre (JRC) of the European Commission, Institute for Environment and Sustainability (IES), MARS Unit, Via Enrico Fermi 2749, 21027 Ispra (Va), Italy^1

Flemish Institute for Technological Research (VITO), Centre for Remote Sensing and Earth Observation, Boeretang 200, B-2400 Mol, Belgium^2

关键词: Direct expansion;    High resolution image;    High resolution imagery;    Non-agricultural lands;    North China Plain;    Regression estimators;    Remote sensing images;    Variance of estimates;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/17/1/012057/pdf
DOI  :  10.1088/1755-1315/17/1/012057

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

The use of remote sensing images in combination with ground survey data was assessed for deriving crop areas over Mengcheng County in 2011 in the North China Plain. First, a stratification of the county into arable land, permanent crops and non agricultural land was carried out by photo-interpreting a grid of points on Google Earth and a 2.5m Spot5 image from 2011. Then a sample of 83 segments was randomly selected in the arable stratum and surveyed with GPS. Two high resolution images (TM 30m and Spot5 10m) were acquired over the 2011 summer crop season and classified using maximum likelihood. The regression estimator was then applied using the surveyed segments and the classification and compared to the direct expansion estimate derived from the segments only; the calibration estimator was also tested using the same classification and the 83 arable points that served as seeds for the segments and compared to the estimate derived from the 83 points alone. The regression estimator proved to be the most efficient one in the North China Plain landscape. To reach the same variance of estimate as the regression estimator, the number of points to be surveyed for the calibration estimator should be multiplied by seven. Last pixel counting tested on the whole county and on the arable points of the grid resulted in biased estimates, in contrast to estimates based on ground data, in combination with remote sensing or not.