Meta-analysis of influential factors on crop yield estimation by remote sensing

Abstract

Crop yield estimation by remote sensing is an important task in food security. Despite numerous studies in this field, there is a lack of a systematic analysis on the past studies to enable researchers to link individual case studies together. In this article, meta-analysis is adopted to pool various studies around the world and reveal the influences of the factors (such as crop types, soil orders, and climate systems) on the correlations between crop yield and remote-sensing data. The meta-analysis synthesizes the data on the effectiveness of NDVI (normalized difference vegetation index) on crop-yield estimation and it has been found that the correlation coefficient (r) varies significantly. Different correlation values are found in different crop types, with the highest 0.88 in cotton and the lowest 0.79 in sugar cane. They are found to be related to the single-leaf blades of the crops except for sugar cane. In addition, the meta-analysis results show that the correlations are also affected by the soil orders and climates, as is evident from the positive correlation with fertility and the negative correlation with precipitation and temperature, respectively. The mean correlation in Mollisols is stronger than that in Oxisols, and the value in hot and humid climates (e.g. humid subtropical climate) is lower than that in dry and cold climates (e.g. temperate continental climate). The study provides useful information for future individual case studies and meta-analysis in this field.

Acknowledgement

The first author acknowledges the support provided by China Scholarship Council during a visit to the University of Bristol.

Funding

This work was supported by the National Social Science Foundation of China [grant number 12&ZD214]; Project of Science and Technology of Yunnan Province [grant number 2010CA013]; and Research Innovation Programme for College Graduates of Jiangsu Province [grant number CXLX12_0259].