Two diagrams "get the basic data" of the global farmland

“ Arable land is the lifeblood of grain production, and it is of great significance for the international community to ensure food security, regulate food market, cope with global change and promote green agricultural development. Wu Wenbin, chief scientist of the Intelligent Agricultural Innovation Team of the Institute of Agricultural Resources and Agricultural Zones of the Chinese Academy of Agricultural Sciences, told China Science Daily.
Recently, the team, in cooperation with the International Food Policy Research Institute, the International Institute for Applied Systems Analysis, the International Center for Corn and Wheat Improvement, and other organizations, has continuously updated and published the 2010 global high-precision arable land spatial distribution and crop spatial distribution of the drawing data products.
By the global academic peer and user review, compared with historical data products, the accuracy of this series of products has been greatly improved, coverage has been greatly expanded, availability has been greatly enhanced, providing an important basis for the "United Nations Sustainable Development Goals" assessment of data support.Wenbin, author of the paper, said that although many countries and scientific research institutions have developed many global or regional arable land mapping data products, their data products are mainly based on maps and satellite imagery data. However, the quality of the data is uneven, the consistency between them is poor, the degree of conformity with published official statistics is low, and there is a problem that remote sensing imagery "sees" is inconsistent with the "cultivated land data" found by statistical departments.
this brings a lot of uncertainty to the application of data products in government departments, international institutions and research universities.Lu Miao, the first author of the
paper and an associate researcher at the Institute, told China Science Daily that in order to solve the above problems and update the global arable land distribution data, the joint team led by the project systematically collected more than 10 sets of global and regional remote sensing arable land mapping data products: the global scale includes 5 data products, the regional scale mainly high-resolution products, covering North America, Europe and Oceania. In addition, they systematically collected global-scale statistics on arable land covering multiple levels at the national, provincial and municipal levels, filling gaps in statistics in developing countries such as Africa, Latin America and Asia.
, author of the paper and a senior fellow at the International Food Policy Institute, is responsible for collecting statistics on arable land and crops around the world. "There's a lot of surface data, but the statistics are very hard to come by." Mr. Yu told China Science Daily that statistics from some developing countries are not publicly available or are not available online and must be obtained exclusively from local public publications, sometimes even from local statistical authorities or management.
2018, when he set out to collect statistics on the distribution of global arable land, the most recent data available to some countries was in 2010. But in 2010, the African countries of South Sudan and North Sudan were the same country. "The borders have changed, the administrative divisions have changed, and the administration has changed. It is very difficult to obtain historical information. "With the International Food Policy Institute's global branch and office network, Mr. Yu asked local staff for help to collect these statistics.
based on the above-mentioned crowd source data, the joint research team put forward a new idea of "graded control check, bottom-up optimization, automatic distribution calibration" and constructed a new adaptive distribution model of statistical data.
"We evaluate the accuracy and consistency between existing remote sensing mapping products, and on this basis, compare statistical data with the quantity accuracy of remote sensing farmland mapping products, construct integration rules, and realize the integration and optimization of statistical data and remote sensing data products of cultivated land. Lu Miao said.
using this method, they developed the Global 2010 500-meter resolution arable land fusion drawing products (hereinafter referred to as arable land map), with an overall accuracy of 90.8%, higher than the existing five sets of global arable land products. At the same time, the consistency of cultivated land drawing products and statistical data is better than that of existing global arable land drawing products.
the above-mentioned research provides new ideas and methods for remote sensing mapping and related research on global scale arable land, and provides accurate and reliable basic information on the spatial distribution of cultivated land for global agricultural monitoring, food estimation and global change research." approach to global arable land maping is only the first step in our work, and more importantly, how to obtain information on the spatial distribution of crops around the world. However, due to data acquisition conditions and other factors, it is a major challenge to carry out research on the spatial distribution of crops on a global scale. Yu Qiangyi, the first author of the paper and an associate researcher at the Institute of Planning, told China Science that most of the similar crop distribution data products have remained in 2000, such as M3, MIRCA2000, SPAM2000, etc., which has been difficult to meet the current application needs in agriculture, environment and other fields.
Through cooperation with international research institutions, they collected agricultural production data from 186 countries around the world, and with the support of spatial data such as arable land mapping, completed the global spatial distribution of 42 major crops, such as rice, wheat and maize, and updated the global crop spatial distribution data products to 2010 for the first time, greatly enhancing the continuity of crop spatial distribution data products.
" set of data covered crops from the previous 26 to 42 species, basically completely covering the world's major crop types, improve the availability of crop spatial distribution data products. Yu Qiangyi said that the data set not only includes the global major crop area, total yield, yield and other spatial distribution of the overall situation, but also from the perspective of actual production inputs, each crop is divided into four agricultural production systems - irrigation-type, high-input rain-fed system, low-input rain-fed system, self-sufficient system type, improve the objectivity of crop spatial distribution data products.
this work reflects the strengths of the global network of agricultural research collaborations. Yu Liangzhi said that the Chinese Academy of Agricultural Sciences, under the support of the International Agricultural Science Program, in cooperation with the global agricultural scientific research advantage units, in data collation, processing, cleaning, production and other aspects of strict quality control, and for different crops, different regions, different production systems to carry out strict accuracy verification, greatly ensuring the reliability of data products.Yang Peng, author of the
paper and a researcher at the Institute of Planning, compared the spatial distribution of crops in different years and found that in recent years, rice cultivation in Eastern Europe, Africa, northeast China, northwest India, South Australia and other regions increased significantly, while rice cultivation in Central Asia and South America decreased significantly; "The biggest significance of our work is to fill the gaps in this kind of data products around the world. Wu Wenbin believes that data is the basis of scientific research. Even if the future will rely heavily on artificial intelligence for analytical work, without a large amount of basic data, AI will not be able to learn and gain analytical capabilities.
more than a decade ago, Yu Liangzhi has been working on the construction of the Spatial Production Design Model, SPAM. SPAM, in turn, is MAPS. Yu said the work is aimed at providing scientists and policymakers around the world with high-precision data on the spatial distribution of arable land and crops. The result of these two collaborations is the latest issue of SPAM 2010.
Said that high-precision data can be used to estimate global crop yields, analyze changes in crop planting area, and then provide basic data for agricultural structural adjustment, agricultural emission reduction and other policy development.
using this set of data products, they found that arable land in South American countries is on the rise, but its crop yields are not growing much in real terms, while countries such as China and the United States have maintained increasing yields despite declining arable land.
", this not only shows that the latter's arable land output rate is significantly higher than the former, but also shows that China's agricultural science and technology has helped to increase grain production capacity, to stabilize grain production no longer need to reclaim large amounts of cultivated land. This provides an important guarantee for the construction of ecological civilization in China and the realization of the goal of 'carbon-medium'. Yu Qiangyi said.
said the work would be updated over a five- to 10-year time period and hoped to "update to satellite remote sensing data as an alternative to statistics."
well known, satellite remote sensing data can reach tens of meters or even sub-meters, but because the technical bottleneck problem has not been solved, remote sensing data alone is difficult to obtain high-precision arable land and crop distribution data products.
explained that due to the complexity of farmland cultivation, the data algorithms applicable in small areas are difficult to apply in large areas, especially in china, where such plots are scattered.
In addition, the accuracy of satellite remote sensing data is supported by a large amount of ground data, for example, what crops are planted in the land, what seasons they are grown, and how many crops are grown, which require manual on-site research and verification.
same time, there are not many mature global data products, that is to say, there is not much basic data for AI to learn, AI has not yet learned to identify crops only from satellite imagery. This also reduces the accuracy of products that use remote sensing data only to generate high-precision arable land and crop distribution data.
to solve the above technical problems, it may take about 5 to 10 years. The next step, Wu said, will be to create a "crop map" of China, showing crops grown at different times in each farm in a data product.
relevant paper information: