Journal of Geo-information Science >
Progress and Prospect on Mapping Cropping Systems Using Time Series Images
Received date: 2021-10-05
Revised date: 2021-10-26
Online published: 2022-03-25
Supported by
National Natural Science Foundation of China(42171325)
National Natural Science Foundation of China(4771468)
Science Bureau of Fujian Province(2020N5002)
Copyright
Updated spatiotemporal explicit data on cropping system is vital for ensuring the implementation of the national food security strategy and reasonable cropping structures. Time series analysis techniques are playing a more important role in agricultural remote sensing along with the continuously improved quality of remote sensing time series images. This paper analyzes main progresses and challenges in the field of cropping systems mapping using time series images from three aspects: mapping framework, remote sensing feature parameters, and data products. We find that: (1) The current cropping system mapping framework which mainly includes cropping intensity and agricultural planting structures, needs to cope with the problems of pre-requirements of cropland distribution data with high-quality. However, the existing land use/cover data could not fully fulfil this requirement due to the complex spectral characteristics of cropland introduced by multiple cropping systems over large regions. It is difficult to accurately derive information on cropping intensity using traditional time series vegetation indices datasets. Specifically, cropland fallow/abandonment in humid regions might be misclassified as single crop due to its corresponding high values of vegetation indices. Cropland abandonment and fallow are not negligible in recent decades and need further investigations, especially in China; (2) Novel multi-dimensional spectral indices based on red-edge and short-wave near-infrared bands are efficient in revealing the crop growth processes. Great progresses have been made in crop mapping in recent years. However, crop mapping at large scale is challenged by the minor differences among different crops as well as distinct heterogeneity within the same crop across different regions and multiple years; (3) There are increasing available remote sensing products of cropping intensity from national to global scale, however, the timeliness and spatiotemporal continuity need to be further improved; (4) Except for a few countries in North America and Europe, crop distribution maps at national scale are not fully available or limited to several staple crops with coarse resolution. There is a deficiency of finer datasets on cropping systems at large scale, especially in the complex multi-cropped regions. Fortunately, new technologies (i.e., cloud computing platform and deep learning algorithms) and emerging multi-sources remote sensing data with higher spatial, spectral, and temporal resolution provide great opportunities for spatiotemporally continuously detecting changes in cropping system at large scale. Future research should be focused on the following directions. First, we could improve the research strategy by developing an integrated mapping framework for directly deriving information on cropland and cropping patterns without relying on existing cropland distribution data. Second, we need to enrich the phenological features through exploring multiple-dimensional and less exploited spectral indices, such as the pigment indices, soil indices, nitrogen indices, and dry matter indices. Finally, we can develop spatiotemporal continuous change detection techniques for automatically tracking changes in cropping systems at multiple years and large scale.
QIU Bingwen , YAN Chao , HUANG Wenqing . Progress and Prospect on Mapping Cropping Systems Using Time Series Images[J]. Journal of Geo-information Science, 2022 , 24(1) : 176 -188 . DOI: 10.12082/dqxxkx.2022.210604
表1 国家或全球尺度耕地复种指数部分代表性数据产品Tab. 1 Some representative data products of cropping intensity at national or global scales |
范围 | 数据源 | 分辨率 | 时间/年 | 监测方法 | 方法校验 | 代表文献 |
---|---|---|---|---|---|---|
全球 | 国际或次国家尺度作物播种和收获统计面积 | 30 弧分 | 1998—2002 | 作物生长季节的重叠性 | 与多季种植统计数据等对比分析 | Waha等[24] |
全球 | 15 d GIMMS NDVI | 8 km | 1982—2012 | 峰值法 | 未验证 | Chen等[26] |
全球 | 15 d GIMMS NDVI | 8 km | 2010 | 峰值法 | 未验证 | Wu等[2] |
全球 | Landsat、Sentinel-2 MSI | 30 m | 2016—2018 | 穿过1/2变幅的点位数 | 近4000个参考点位(OA>93%) | Zhang等[17] |
亚洲 | 8 d MODIS EVI | 500 m | 2009—2012 | 改进的峰值法 | 191个调研点位(OA≥11%) | Gray等[1] |
中国 | 8 d MODIS EVI | 500 m | 2001—2015 | 改进的峰值法 | 起始年份各自200个目视解译样点(OA=96%) | 闫慧敏等[14] |
中国 | 旬SPOT NDVI | 1 km | 1999—2013 | 峰值法 | 1515个目视解译样点(OA=92%) | 丁明军等[27] |
中国 | 7 d VIP EVI2 8 d MODIS EVI2 | 500 m/5 km(20世纪) | 1982—2018 | 小波特征图谱法 | 近1000个调研点位(OA>91%) | 邱炳文等[18,28] |
注:OA为总体精度。 |
表2 大尺度农作物分布遥感制图部分代表性数据产品Tab. 2 Some representative crop spatial distribution data products at large scale using remote sensing images. |
范围 | 数据源 | 分辨率 | 时间/年 | 制图方法 | 作物 | 验证数据及精度 | 代表文献 |
---|---|---|---|---|---|---|---|
美国 | 多源遥感数据 | 30 m | 逐年更新 | 决策树监督分类 | 多种农作物 | 基于调研点位(OA=95%) | Boryan等[74] |
加拿大 | 多源遥感数据 | 30 m | 逐年更新 | 决策树监督分类 | 多种农作物 | 基于调研点位(OA>85%) | Davidson等[76] |
中国和印度 | 8 d MODIS EVI2、LSWI | 500 m | 2000—2015 | 物候指标(水体指数与植被指数之差) | 水稻 | 基于统计数据(中国:R2=0.77—0.92;印度R2=0.79—0.81) | Zhang等[62] |
中国 | 8 d GLASS LAI | 1 km | 2000—2015 | 植被指数曲线拐点和阈值法 | 水稻、小麦和玉米 | 基于统计数据(R2>0.8) | Luo等[77] |
中国 | 8 d MODIS EVI2、NMDI | 500 m | 2005—2017 | 物候指标(生长盛期NMDI增减比值指数) | 玉米 | 基于2020个调研点位(OA=91%);基于统计数据(R2=0.89) | Qiu等[63] |
中国东南10省市 | 8 d MODIS EVI2、LSWI | 500 m | 2001—2013 | 物候指标(植被水体变化比值指数) | 水稻 | 基于763个调研点位(OA=95%) | Qiu等[71, 78] |
中国冬小麦主产区10省市 | 8 d MODIS EVI2、物候站点数据 | 500 m | 2001—2012 | 物候指标(基于生长期植被指数增量) | 冬小麦 | 基于653调研点位(OA=92%);基于统计数据(R2=0.84) | Qiu等[66] |
东北三省 | Sentinel-2 MSI | 10 m | 2017—2019 | 随机森林 | 水稻、玉米和大豆等 | 每年上万个调研点位(OA=81—86%);基于统计数据(R2>0.84) | You等[79] |
注:OA为总体精度。表中基于物候指标的分类方法,均基于简单决策判别规则。 |
表3 目前常规与未来新型基于时序遥感数据的农作物种植制度研究对比分析Tab. 3 Comparative analysis of traditional and novel research framework for cropping systems |
对比参数 | 常规研究 | 发展趋势 |
---|---|---|
遥感影像 | MODIS、Landsat等 | Sentinel、Landsat、国产GF、无人机等多源高时空谱遥感数据 |
光谱指数 | 常规植被指数 | 新型多维度遥感指数 |
特征参数 | 基于常规植被指数的农作物物候特征 | 植被色素、氮含量、干物质、含水量等多维度属性在不同生长阶段变化特性及其相互时序关系特征等 |
研究框架 | 耕地掩膜-剔除撂荒区-熟制识别-农作物制图 | 一体化遥感监测(耕作区-农作物种植模式) |
变化检测 | 逐年分类制图-分类后检测变化 | 时序遥感变化检测、时空连续性好 |
研究方法 | 监督分类、非监督分类算法 | 深度学习与基于物候的专家知识方法相结合 |
支撑平台 | ENVI等遥感处理平台 | GEE等云计算平台 |
研究尺度 | 中小尺度、中低分辨率 | 大尺度、中高分辨率 |
[1] |
|
[2] |
|
[3] |
杨婷, 赵文利, 王哲怡, 等. 基于遥感影像 NDVI 数据的中国种植制度分布变化[J]. 中国农业科学, 2015, 48(10):1915-1925.
[
|
[4] |
闫慧敏, 刘纪远, 曹明奎. 近20年中国耕地复种指数的时空变化[J]. 地理学报, 2005,(4):559-566.
[
|
[5] |
唐华俊, 吴文斌, 余强毅, 等. 农业土地系统研究及其关键科学问题[J]. 中国农业科学, 2015, 48(5):900-910.
[
|
[6] |
李丹, 韩留生, 黄思宇, 等. 基于 HJ-1 数据的冬小麦种植面积,长势与收割进度遥感监测[J]. 中国农学通报, 2016, 32(12):20-28.
[
|
[7] |
葛中曦, 黄静, 赖佩玉, 等. 耕地复种指数遥感监测研究进展[J]. 地球信息科学学报, 2021:1-16.
[
|
[8] |
吴文斌, 余强毅, 陆苗, 等. 耕地复种指数研究的关键科学问题[J]. 中国农业科学, 2018, 51(9):1681-1694.
[
|
[9] |
|
[10] |
|
[11] |
[12] |
董金玮, 吴文斌, 黄健熙, 等. 农业土地利用遥感信息提取的研究进展与展望[J]. 地球信息科学学报, 2020, 22(4):772-783.
[
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
|
[19] |
|
[20] |
|
[21] |
|
[22] |
|
[23] |
李升发, 李秀彬. 耕地撂荒研究进展与展望[J]. 地理学报, 2016, 71(3):370-389.
[
|
[24] |
|
[25] |
|
[26] |
|
[27] |
丁明军, 陈倩, 辛良杰, 等. 1999-2013年中国耕地复种指数的时空演变格局[J]. 地理学报, 2015, 70(7):1080-1090.
[
|
[28] |
李宇, 邱炳文, 何玉花, 等. 基于MODIS数据的2001—2018年中国耕地复种指数反演研究[J]. 地理科学进展, 2020, 39(11):1874-1883.
[
|
[29] |
宫鹏, 张伟, 俞乐, 等. 全球地表覆盖制图研究新范式[J]. 遥感学报, 2016, 20(5):1002-1016.
[
|
[30] |
黄思宇, 陈水森, 李丹, 等 . 基于 NDVI 物候特征的华南地区冬种马铃薯遥感提取方法[J]. 热带地理, 2016, 36(6):976-984.
[
|
[31] |
吴文斌, 胡琼, 陆苗, 等. 农业土地系统遥感制图[M]. 北京: 科学出版社, 2020.
[
|
[32] |
|
[33] |
|
[34] |
|
[35] |
|
[36] |
|
[37] |
|
[38] |
|
[39] |
|
[40] |
|
[41] |
|
[42] |
|
[43] |
|
[44] |
|
[45] |
|
[46] |
|
[47] |
|
[48] |
|
[49] |
|
[50] |
|
[51] |
|
[52] |
|
[53] |
|
[54] |
吴炳方, 李强子. 基于两个独立抽样框架的农作物种植面积遥感估算方法[J]. 遥感学报, 2004, 8(6):551-568.
[
|
[55] |
|
[56] |
郝鹏宇, 唐华俊, 陈仲新, 等. 基于历史增强型植被指数时序的农作物类型早期识别[J]. 农业工程学报, 2018, 34(13):179-186.
[
|
[57] |
|
[58] |
|
[59] |
|
[60] |
曾志康, 覃泽林, 黄启厅, 等. 基于国产高时空分辨率卫星影像的作物种植信息提取研究[J]. 福建农业学报, 2017, 32(5):560-567.
[
|
[61] |
|
[62] |
|
[63] |
|
[64] |
许青云, 杨贵军, 龙慧灵, 等. 基于 MODIS NDVI 多年时序数据的农作物种植识别[J]. 农业工程学报, 2014, 30(11):134-144.
[
|
[65] |
段春华, 张汛, 尹凡. 基于国产高分卫星影像的水稻种植区提取——以江苏苏中地区为例[J]. 卫星应用, 2021.
[
|
[66] |
|
[67] |
|
[68] |
|
[69] |
|
[70] |
|
[71] |
|
[72] |
|
[73] |
|
[74] |
|
[75] |
|
[76] |
|
[77] |
|
[78] |
|
[79] |
|
[80] |
吴炳方. 全国农情监测与估产的运行化遥感方法[J]. 地理学报, 2000, 55(1):23-35.
[
|
[81] |
|
[82] |
|
[83] |
|
[84] |
|
[85] |
|
[86] |
梁继, 郑镇炜, 夏诗婷, 等. 高分六号红边特征的农作物识别与评估[J]. 遥感学报, 2020, 24(10):1168-1179.
[
|
[87] |
赵忠明, 孟瑜, 岳安志, 等. 遥感时间序列影像变化检测研究进展[J]. 遥感学报, 2016, 20(5):1110-1125.
[
|
[88] |
|
[89] |
|
[90] |
|
[91] |
|
/
〈 | 〉 |