Cotton is an important global cash crops that serve as the primary source of natural fiber for textiles.A thorough understand-ing of the long-term variations in cotton cultivation is vital for optimizing cotton cultiv...Cotton is an important global cash crops that serve as the primary source of natural fiber for textiles.A thorough understand-ing of the long-term variations in cotton cultivation is vital for optimizing cotton cultivation management and promoting the sustainable development of the cotton industry.Xinjiang is the primary cotton-producing region in China.However,long-term data of cotton cultiv-ation areas with high spatial resolution are unavailable for Xinjiang,China.Therefore,this study aimed to identify and map an accurate 30-m cotton cultivation area dataset in Xinjiang from 2000 to 2020 by applying a Random Forest(RF)-based method that integrates Landsat and Moderate Resolution Imaging Spectroradiometer(MODIS)images,and validated the applicability and accuracy of dataset at a large spatial scale.Then,this study analyzed the spatiotemporal variations and influencing factors of cotton cultivation in the study period.The results showed that a high classification accuracy was achieved(overall accuracy>85%,F1>0.80),strongly agreeing with county-level agricultural statistical yearbook data(R2>0.72).Significant spatiotemporal variation in the cotton cultivation areas was found in Xinjiang,with a total increase of 1131.26 kha from 2000 to 2020.Notably,cotton cultivation area in southern Xinjiang expan-ded substantially,with that in Aksu increasing from 20.10%in 2000 to 28.17%in 2020,representing an expansion of 374.29 kha.In northern Xinjiang,the cotton areas in the Tacheng region also exhibited significant increased by almost ten percentage points in the same period.In contrast,cotton cultivation in eastern Xinjiang declined,decreasing from 2.22%in 2000 to merely 0.24%in 2020.Standard deviation ellipse analysis revealed a‘northeast-southwest’spatial distribution,with the centroid consistently located in Aksu and shifting 102.96 km over the 20-yr period.Pearson correlation analysis indicated that socioeconomic factors had a stronger influence on cotton cultivation than climatic factors,with effective irrigation area(r=0.963,P<0.05)and total agricultural machinery power(r=0.823)showing significant positive correlations,whereas climatic variables exhibiting weak associations(r<0.200).These results provide valuable scientific data for informed agricultural management,sustainable development,and policymaking.展开更多
天山北坡是西北地区的重要水源涵养区及草原畜牧业基地,其积雪融水对生态系统维持、农业灌溉及城市供水至关重要。为解决MODIS积雪产品易受云层干扰而导致的数据缺失问题,论文通过扩展MODIS数据输入,以已有积雪数据共同识别为积雪或非...天山北坡是西北地区的重要水源涵养区及草原畜牧业基地,其积雪融水对生态系统维持、农业灌溉及城市供水至关重要。为解决MODIS积雪产品易受云层干扰而导致的数据缺失问题,论文通过扩展MODIS数据输入,以已有积雪数据共同识别为积雪或非积雪的像元为“真值”,采用随机森林、支持向量机及BP神经网络等机器学习算法,确定积雪识别最佳方案。结合多种数据协同去云方法与隐马尔可夫随机场(hidden Markov random field,HMRF)算法,对去云效果进行对比分析,并使用高分辨率Landsat数据对实验结果的准确性进行验证。研究表明:(1)随机森林模型在积雪二分类任务中的表现最佳,准确率达90.15%,精确率达91.95%;(2)多种数据协同去云方法可以取得较好效果,Kappa系数为0.729,但结合HMRF方法的去云效果最佳,总体精度达82.84%,生产者精度为88.46%,Kappa系数为0.795;(3)年均积雪天数、积雪覆盖天数与海拔之间关系、月均积雪覆盖率与年均积雪覆盖面积变化趋势均与已有数据保持较高一致性。研究结果表明该方法能够有效提升积雪监测精度与时空连续性,为天山北坡及相似地区的积雪监测、冰雪水资源评估和生态环境管理提供了可靠的技术支撑。展开更多
MODIS植被指数数据是区域土地利用信息提取的重要数据源。为了对比MODIS两种主要植被指数(NDIV、EVI)在耕地信息提取中的应用,采用通过时间序列谐波分析法(Harmonic Analysis of Time Series,HANTS),对2006年全年MODIS 16天250m的NDVI和...MODIS植被指数数据是区域土地利用信息提取的重要数据源。为了对比MODIS两种主要植被指数(NDIV、EVI)在耕地信息提取中的应用,采用通过时间序列谐波分析法(Harmonic Analysis of Time Series,HANTS),对2006年全年MODIS 16天250m的NDVI和EVI时间谱数据进行了重构,从而进行了河西走廊绿洲中东部样区一系列耕地信息的提取实验,包括耕地、休耕地识别以及耕地复种指数、作物种类提取。在此基础上,对MODIS的NDVI与EVI数据的应用进行了对比分析。结果显示:(1)利用傅立叶谐波变换得到的EVI和NDVI时间谱曲线的谐波余项及谐波振幅对耕地进行识别,从识别精度来看,EVI要优于NDVI,识别精度分别为97.17%和95.99%,Kappa系数分别达到0.7938和0.6518;(2)通过计算时间序列曲线的波峰数能够提取耕地的复种指数,并且在EVI和NDVI曲线波峰阈值分别设为0.20和0.25时,休耕地能较为准确地被识别出来;(3)通过提取作物生长期内曲线的VI最大增长速率时间点以及峰值时间点等信息,作物种类能被初步识别,并且EVI较NDVI具有更强的识别能力。展开更多
基金Under the auspices of the National Natural Science Foundation of China(No.42101342,U2243205)the Third Comprehensive Scientific Expedition to Xinjiang(No.2021XJKK1403)。
文摘Cotton is an important global cash crops that serve as the primary source of natural fiber for textiles.A thorough understand-ing of the long-term variations in cotton cultivation is vital for optimizing cotton cultivation management and promoting the sustainable development of the cotton industry.Xinjiang is the primary cotton-producing region in China.However,long-term data of cotton cultiv-ation areas with high spatial resolution are unavailable for Xinjiang,China.Therefore,this study aimed to identify and map an accurate 30-m cotton cultivation area dataset in Xinjiang from 2000 to 2020 by applying a Random Forest(RF)-based method that integrates Landsat and Moderate Resolution Imaging Spectroradiometer(MODIS)images,and validated the applicability and accuracy of dataset at a large spatial scale.Then,this study analyzed the spatiotemporal variations and influencing factors of cotton cultivation in the study period.The results showed that a high classification accuracy was achieved(overall accuracy>85%,F1>0.80),strongly agreeing with county-level agricultural statistical yearbook data(R2>0.72).Significant spatiotemporal variation in the cotton cultivation areas was found in Xinjiang,with a total increase of 1131.26 kha from 2000 to 2020.Notably,cotton cultivation area in southern Xinjiang expan-ded substantially,with that in Aksu increasing from 20.10%in 2000 to 28.17%in 2020,representing an expansion of 374.29 kha.In northern Xinjiang,the cotton areas in the Tacheng region also exhibited significant increased by almost ten percentage points in the same period.In contrast,cotton cultivation in eastern Xinjiang declined,decreasing from 2.22%in 2000 to merely 0.24%in 2020.Standard deviation ellipse analysis revealed a‘northeast-southwest’spatial distribution,with the centroid consistently located in Aksu and shifting 102.96 km over the 20-yr period.Pearson correlation analysis indicated that socioeconomic factors had a stronger influence on cotton cultivation than climatic factors,with effective irrigation area(r=0.963,P<0.05)and total agricultural machinery power(r=0.823)showing significant positive correlations,whereas climatic variables exhibiting weak associations(r<0.200).These results provide valuable scientific data for informed agricultural management,sustainable development,and policymaking.
文摘天山北坡是西北地区的重要水源涵养区及草原畜牧业基地,其积雪融水对生态系统维持、农业灌溉及城市供水至关重要。为解决MODIS积雪产品易受云层干扰而导致的数据缺失问题,论文通过扩展MODIS数据输入,以已有积雪数据共同识别为积雪或非积雪的像元为“真值”,采用随机森林、支持向量机及BP神经网络等机器学习算法,确定积雪识别最佳方案。结合多种数据协同去云方法与隐马尔可夫随机场(hidden Markov random field,HMRF)算法,对去云效果进行对比分析,并使用高分辨率Landsat数据对实验结果的准确性进行验证。研究表明:(1)随机森林模型在积雪二分类任务中的表现最佳,准确率达90.15%,精确率达91.95%;(2)多种数据协同去云方法可以取得较好效果,Kappa系数为0.729,但结合HMRF方法的去云效果最佳,总体精度达82.84%,生产者精度为88.46%,Kappa系数为0.795;(3)年均积雪天数、积雪覆盖天数与海拔之间关系、月均积雪覆盖率与年均积雪覆盖面积变化趋势均与已有数据保持较高一致性。研究结果表明该方法能够有效提升积雪监测精度与时空连续性,为天山北坡及相似地区的积雪监测、冰雪水资源评估和生态环境管理提供了可靠的技术支撑。
文摘MODIS植被指数数据是区域土地利用信息提取的重要数据源。为了对比MODIS两种主要植被指数(NDIV、EVI)在耕地信息提取中的应用,采用通过时间序列谐波分析法(Harmonic Analysis of Time Series,HANTS),对2006年全年MODIS 16天250m的NDVI和EVI时间谱数据进行了重构,从而进行了河西走廊绿洲中东部样区一系列耕地信息的提取实验,包括耕地、休耕地识别以及耕地复种指数、作物种类提取。在此基础上,对MODIS的NDVI与EVI数据的应用进行了对比分析。结果显示:(1)利用傅立叶谐波变换得到的EVI和NDVI时间谱曲线的谐波余项及谐波振幅对耕地进行识别,从识别精度来看,EVI要优于NDVI,识别精度分别为97.17%和95.99%,Kappa系数分别达到0.7938和0.6518;(2)通过计算时间序列曲线的波峰数能够提取耕地的复种指数,并且在EVI和NDVI曲线波峰阈值分别设为0.20和0.25时,休耕地能较为准确地被识别出来;(3)通过提取作物生长期内曲线的VI最大增长速率时间点以及峰值时间点等信息,作物种类能被初步识别,并且EVI较NDVI具有更强的识别能力。
