A modified spectral remote sensing index to map plastic greenhouses in fragmented terrains

Plastic greenhouse (PG), as a new type of modern agricultural measure, has been used widely due to its significant benefits for agricultural production. However, it also raises concerns about its potential environmental impact. Monitoring of PG is necessary for the agricultural sustainability. However, extracting PGs in fragmented terrains based on remote sensing images is difficult due to the variety of types of PGs and high environmental heterogeneity. In this study, a modified plastic greenhouse index (MPGI) was proposed to monitor PG based on the differences on spectral signatures using Landsat-8 Operational Land Imager. Four study sites, including Weifang (China), Nantong (China), Kunming (China), and Dalat (Vietnam), were selected for index applications. And the effectiveness and robustness of the MPGI were examined by comparing with the exiting PG indices. The results indicated that MPGI improved extraction accuracy in fragmented terrains. The F1 scores for MPGI classification accuracy ranged from 85.7 % to 87.9 %, while other PG indices demonstrated between 67.0 % and 86.4 %. The MPGI demonstrated its capability across various season and datasets, highlighting it has the potential for the PGs mapping in heterogeneous regions. This index is capable of effecting a transformation of greenhouses from "vague agricultural facilities" into computable and manageable spatial decision-making units. In establishing an underlying data foundation for smart agriculture development, it serves to reduce the workload of manual labor.