Cotton cultivated area detection and yield monitoring combining remote sensing with field data in lower Indus River basin, Pakistan
Cotton cultivated area detection and yield monitoring combining remote sensing with field data in lower Indus River basin, Pakistan
复制标题
DOI：
10.1007/s10661-023-11004-3
复制
发表时间：
2023-02
影响因子：
3
通讯作者：
M. Naveed;Hong S. He;Shengwei Zong;Haibo Du;Zulqarnain Satti;Hang Sun;Shuai Chang
中科院分区：
文献类型：
--
作者：
M. Naveed;Hong S. He;Shengwei Zong;Haibo Du;Zulqarnain Satti;Hang Sun;Shuai Chang
随着时间的推移，地球人口不断增加，人类对农业原料的需求也在增加。它对于保持关于世界农业系统动态和特性的最新和准确信息至关重要。棉花作为经济作物，其空间分布信息的更新是监测区域作物面积和生长变化的必要条件。采用8 d增强型植被指数（EVI）时间序列检测棉花作物面积，采用二项概率法获得棉花作物发生的概率分布。利用作物报告数据，利用高斯克里格法推导出检测棉花种植区内的棉花产量。我们还使用农民的田间数据来验证棉花产量结果。在所有研究年份（2004-2019）中，modis衍生的棉花种植面积与tehsil水平的统计数据具有很强的相关性（R2= 0.84）。棉花作物面积检测的总准确率为84.6%，产量预测的总准确率为92.1%。我们的研究提出了新的方法来绘制棉花面积和产量，通过机器学习可以适用于其他地区。
As the Earth’s population continuously increase with the passage of time, the demand for agricultural raw material for human need increases. It is critical to maintaining updated and accurate information about the dynamics and properties of the world agricultural systems. As cash crop, the updated information of the spatial distribution of cotton field is necessary to monitor the crop area and growth changes at regional level. We used 8-day enhanced vegetation index (EVI) time series to detect cotton crop area and binomial probabilistic approach to obtain the probability distribution of cotton crop occurrence. We used Gaussian kriging to derive cotton yield inside the detected cotton crop areas through crop reporting data. We also used field data from farmers to validate the cotton yield results. A strong correlation between theMODIS-derived cotton cultivated area and statistical data at the tehsil level were achieved (R2= 0.84) for all study years (2004–2019). The total accuracy for the cotton crop area detection was 84.6% and yield prediction was 92.1%. Our study presents new approaches to map cotton area and yield, which are applicable to other regions through machine learning.