An index of non-sampling error in area frame sampling based on remote sensing data.

Chunyang Zhang,Pengyu Hao,Yuchu Qin,Wang Li,Mingquan Wu,Chenghai Yang,Zheng Niu,Dailiang Peng

doi:10.7717/peerj.5824

Abstract

Agricultural areas are often surveyed using area frame sampling. Using non-updated area sampling frame causes significant non-sampling errors when land cover and usage changes between updates. To address this problem, a novel method is proposed to estimate non-sampling errors in crop area statistics. Three parameters used in stratified sampling that are affected by land use changes were monitored using satellite remote sensing imagery: (1) the total number of sampling units; (2) the number of sampling units in each stratum; and (3) the mean value of selected sampling units in each stratum. A new index, called the non-sampling error by land use change index (NELUCI), was defined to estimate non-sampling errors. Using this method, the sizes of cropping areas in Bole, Xinjiang, China, were estimated with a coefficient of variation of 0.0237 and NELUCI of 0.0379. These are 0.0474 and 0.0994 lower, respectively, than errors calculated by traditional methods based on non-updated area sampling frame and selected sampling units.

Highlights

In many countries, accurate and timely national-level agricultural statistics are provided by agricultural statistics services such as the US Department of Agriculture’s National Agricultural Statistics Service (NASS), the Italian National Statistical Institute (Istat), the Agrifood and Fishery Information Service of Mexico (SIAP), the National Bureau of Statistics of China (NBS), and the Iranian Ministry of Agrculture’s Agricultural Statistics Information Division (ASID; Allen, 1990; Alonso, Soria & Gozalo, 1991; Gallego, 1999; Kussul et al, 2012; Pradhan, 2001; Wu & Li, 2004)
With the area sampling frame updated and the sampled grids re-sampled and surveyed by 2 m high spatial resolution remote sensing imagery with areas of interest (AOIs) based on ground survey data, the total cotton area in the study area and associated errors were successfully estimated (Table 3)
We determined an approach for calculating the non-sampling errors arising from land use changes, which were readily extracted from remote sensing data

Summary

Introduction

The traditional method of producing agricultural statistics (Benedetti et al, 2010; Tsiligirides, 1998) is, firstly, to create a census. The list of sampling units is usually not updated for around 5–10 years. Data from selected sampling units are collected by face-to-face interviews, telephone interviews, or emails. Overall, this method is very costly and, difficult for developing countries to use. Agricultural statistics are often produced by aggregating administrative data, which results in low data quality and quantity (World Bank, FAO & United Nations Statistical Commission, 2011; FAO, World Bank & United Nations Statistical Commission, 2012)

Objectives

Methods

Results

Discussion

Conclusion