Scientists from the Innovation Team of Smart Agriculture of the Institute of Agricultural Resources and Regional Planning (IARRP) of the Chinese Academy of Agricultural Sciences (CAAS), Central China Normal University, Boston University, Kent State University, and the International Food Policy Research Institute presented a new methodology for mapping crop type distributions via integration of remote sensing data and agricultural statistics.

The research results were published in the top international remote sensing academic journal Remote Sensing of Environment (IF 9.09).

Reliable crop type maps are vital for agricultural monitoring, agricultural resources allocation and utilization, and crop planting structure optimization. Medium and low spatial resolution remote sensing data such as MODIS are widely used for crop type mapping and planting structure monitoring due to the wide observation range, multiple spectrum bands and high time-frequency, and are advantageous for detecting the seasonal dynamics of different crop types. However, the inherently low spatial resolution often causes mixed pixels, in addition to many uncertainties in training samples due to atmospheric interference, image preprocessing, and machine learning algorithms, which may significantly affect the accuracy of crop spatial distribution mapping.

Agricultural statistics are also often used in crop spatial distribution mapping. Although it is difficult to provide detailed crop spatial distribution information, it has unique advantages in crop type and quantity mapping and time continuity. Previous studies have often used agricultural statistics as reference data to evaluate the accuracy of satellite-derived crop type maps but have rarely utilized them to improve crop type distribution mapping. The utility of integrating agricultural statistics with satellite images to produce high-accuracy crop type maps has rarely been explored.

This study presents a methodology for mapping sub-pixel crop type distributions via the integration of MODIS time series and agricultural statistics. First, an optimized random forest regression (RF-r) model was used, and a backward feature elimination strategy was implemented to select the best spectrum and time series and improve the accuracy of crop abundance estimation from remote sensing. Second, the study developed an Iterative Area Gap Spatial Allocation (IAGSA) method to reconcile the discrepancies between the crop acreage estimated from MODIS-based maps and the agricultural statistics.

The study tested its approach in Heilongjiang province, which has the highest agricultural production in China. It took the main crops (rice, corn, and soybeans) as the research objects and verified the reliability and stability of the method. The study found that the MODIS-derived crop fractions agreed with those derived from the high-resolution images, with R2 > 0.75 for all crop types. The sub-pixel crop type maps adjusted by IAGSA were not only consistent with the agricultural statistics for crop acreage, but also retained the spatial distribution patterns of the original MODIS-derived crop fraction. IAGSA reconciles the discrepancy between a MODIS-based crop area and statistical data. That is, the smaller the spatial scale of the statistical data, the greater the spatial heterogeneity of the optimized remote sensing results.

The results suggest the advantages of integrating remote sensing data and agricultural statistics, which improves the accuracy of crop spatial distribution mapping based on low and medium-resolution remote sensing data, provides the potential to map crop types across regions on a large scale, and develops the technical methods of integration of remote sensing data sources and non-remote sensing data sources. Overall, the integration provides a new reference for the collaborative use of multi-source data.

This research was funded by the Innovation Group Program of the National Natural Science Foundation of China, the National Key Research and Development Program of China, and the International Agricultural Science Project Program. Hu Qiong, a doctoral dissertation candidate of IAPPR, CAAS is the first author, and research fellow Wu Wenbin is the corresponding author.

https://www.sciencedirect.com/science/article/pii/S0034425721000833

Qiong Hu, He Yin, Mark A. Friedl, Liangzhi You, Zhaoliang Li, Huajun Tang and Wenbin Wu*. Integrating coarse-resolution images and agricultural statistics to generate sub-pixel crop type maps and reconciled area estimates. Remote Sensing of Environment, 2021, 258, 112365.

Figure 1. Crop sub-pixel mapping via the integration of remote sensing data and agricultural statistics

Figure 2. The crop distribution map of Heilongjiang Province via the integration of MODIS data and agricultural statistics, where “a” is the crop abundance map optimized by the IAGSA method and “b” is the crop abundance variation and variation statistics before and after optimization