Landsat-Derived Annual Maps of Agricultural Greenhouse in Shandong Province, China from 1989 to 2018

Abstract

Agricultural greenhouse (AG), one of the fastest-growing technology-based approaches worldwide in terms of controlling the environmental conditions of crops, plays an essential role in food production, resource conservation and the rural economy, but has also caused environmental and socio-economic problems due to policy promotion and market demand. Therefore, long-term monitoring of AG is of utmost importance for the sustainable management of protected agriculture, and previous efforts have verified the effectiveness of remote sensing-based techniques for mono-temporal AG mapping in a relatively small area. However, currently, a continuous annual AG remote sensing-based dataset at large-scale is generally unavailable. In this study, an annual AG mapping method oriented to the provincial area and long-term period was developed to produce the first Landsat-derived annual AG dataset in Shandong province, China from 1989 to 2018 on the Google Earth Engine (GEE) platform. The mapping window for each year was selected based on the vegetation growth and the phenological information, which was critical in distinguishing AG from other misclassified categories. Classification for each year was carried out initially based on the random forest classifier after the feature optimization. A temporal consistency correction algorithm based on classification probability was then proposed to the classified AG maps for further improvement. Finally, the average User’s Accuracy, Producer’s Accuracy and F1-score of AG based on visually-interpreted samples over 30 years reached 96.56%, 86.64% and 0.911, respectively. Furthermore, we also found that the ranked features via calculating the importance of each tested feature resulted in the highest accuracy and the strongest stability in the initial classification stage, and the proposed temporal consistency correction algorithm improved the final products by approximately five percent on average. In general, the resultant AG sequence dataset from our study has revealed the expansion of this typical object of “Human–Nature” interaction in agriculture and has a potential application in use of greenhouse-related technology and the scientific planning of protected agriculture.

Remote sensing of plastic‐covered greenhouses and plastic‐mulched farmlands: Current trends and future perspectives

“…However, a combination of supervised classification, feature extraction, and principal component analysis (PCA) resulted in a very high accuracy (96.6%–98.54%) in few studies (e.g., Yu et al, 2017). In addition, Ou et al (2021) obtained an average users accuracy of 96.56% and a producer's accuracy of 86.64% in a recent study using supervised classification of PCG maps between 1989 and 2018 using LANDSAT imagery and GoogleEarth data. The overall accuracy of supervised classification was found to be superior to that of unsupervised classification (Arcidiacono & Porto, 2010).…”

“…RF classifiers is another supervised classification algorithm, which is an ensemble classifier that produces multiple decision‐trees from a randomly selected subset of training samples and variables (Belgiu & Dragut, 2016), that has been used widely for mapping PCGs and PMFs around the world (e.g., Acharki, 2022; Cui et al, 2022; Gonzalez‐Yebra et al, 2018; Hasituya & Chen, 2017; Koc‐San, 2013; Lin, Jin, et al, 2021; Liu et al, 2019; Novelli et al, 2016; Ou et al, 2020, 2021; Wang & Lu, 2019). RF classifiers are popular among PCG mapping methods due to the accuracy (>90% for multispectral and hyperspectral data) of this method as well as its applicability on high dimensional data, such as hyperspectral imagery, and multisource data (Novelli et al, 2016).…”

“…Even though other classifiers (e.g., SVM or RF) can achieve better accuracy, the main advantage of using DT is that it provides clear decision rules with fixed threshold values that can be used in later works without any training phase (Nemmaoui et al, 2018) and the robustness of DT classifier is proven in PCG and PMF mapping in a number of studies Lu et al, 2014;Xiong et al, 2019). For example, Lu et al (2014) RF classifiers is another supervised classification algorithm, which is an ensemble classifier that produces multiple decision-trees from a randomly selected subset of training samples and variables (Belgiu & Dragut, 2016), that has been used widely for mapping PCGs and PMFs around the world (e.g., Acharki, 2022;Cui et al, 2022;Gonzalez-Yebra et al, 2018;Koc-San, 2013;Lin, Jin, et al, 2021;Liu et al, 2019;Novelli et al, 2016;Ou et al, 2020Ou et al, , 2021Wang & Lu, 2019). RF classifiers are popular among PCG mapping methods due to the accuracy (>90% for multispectral and hyperspectral data) of this method as well as its applicability on high dimensional data, such as hyperspectral imagery, and multisource data (Novelli et al, 2016).…”

“…Due to the increased demand for fruits and vegetables, the use of plastic‐mulched and plastic‐covered greenhouse agriculture (also known as plasticulture) has been increased since the mid‐twentieth century (Jensen & Malter, 1995, Espi et al, 2006) and this has caused changes in agricultural landscapes around the world (Levin et al, 2007) More than 50% of the world's plastic‐mulched agriculture is currently in Asia, where China tops in land areas occupied by plasticulture (Briassoulis et al, 2016). Total greenhouse farming areas in China in has increased from 810 km 2 in 2006 to 1565.5 km 2 in 2010 (Wu et al, 2016) and then to 13,200 km 2 in 2016 (Ou et al, 2021). In addition to China and Mediterranean countries, other areas, such as Vietnam and Morocco (Figure 1), have also intensified their plasticulture practices in recent decades.…”

Tracking Spatio-Temporal Dynamics of Greenhouse-Led Cultivated Land and its Drivers in Shandong Province, China

“…A sequential and fine-resolution GCL mapping dataset is the fundamental information for exploring annual spatiotemporal dynamics of GCL and its driving forces. In this study, we used a series of Landsat-derived maps of GCL in Shandong province, China during 1989-2018 with annual temporal resolution and 30 m spatial resolution (Ou et al, 2021). This dataset was developed based on 8,450 Landsat images on the Google Earth Engine (GEE) and an annual remote sensing mapping method of GCL oriented to the provincial area and long-term period, which was the first dataset with accurate and long-term GCL dynamic maps in China.…”