【作者】 娜日苏；

【导师】 苏和；

【作者基本信息】 中国农业科学院 ， 草地资源利用与保护， 2010， 博士

【摘要】 为了探索一种适合于大范围、迅速、连续监测和评价草地健康的方法,将传统的草地资源地面调查技术与EOS-MODIS遥感技术和地面高光谱技术进行有机组合,在锡林郭勒盟西乌珠穆沁旗选择退化特征较为明显的丘陵草甸草原作为研究样地开展了相关研究。采用2007年6月份~8月份调查的地面数据,根据“天然草地退化、沙化、盐渍化的分级指标,GB19377-2003”中的草地退化等级评定标准,对研究区草地进行了退化等级划分。结果表明,根据6月份和7月份的地面数据评定草地退化等级,容易产生劣化评价误差;而利用8月份的地面数据可以将研究区的草地划分为未退化(对照)、轻度退化、中度退化和重度退化4个等级。通过对比分析不同退化梯度样地内的绿色植被盖度等13个植被指标和土壤紧实度等42个土壤指标随草地退化程度加深而变化的规律,从中筛选对草地退化程度指示性强,且与退化梯度变化具有同步变化规律的绿色植被盖度、裸地百分比、凋落物盖度、总生物量、凋落物重量、优势种优势度、退化指示植物优势度、草层高度等8个植被指标和土壤容重、土壤含水量、0cm~5cm土壤有机质含量、0cm~5cm土壤全磷含量、0cm~5cm土壤速效钾含量等5个土壤指标。并进一步采用主成分分析法从13个植被和土壤指标中,提取了草层高度等5个可以表征草地退化的主成分,并运用模糊数学方法构建了草地健康评价综合指数(RHI);确定了利用草地健康评价综合指数RHI评价丘陵草甸草原时,可参考的健康阈值为:未退化1.00~0.82,轻度退化0.65~0.52、中度退化0.58~0.48、重度退化0.46~0.39。最后利用不同退化梯度样地MODIS-NDVI指数和高光谱反射率数据,通过对植被指数与RHI,单波段光谱与RHI的相关性分析,构建了可估测草地RHI值的遥感模型和光谱模型。经模型的可靠性和精度检验最终确定了最适估测模型。其中: MODIS-NDVI指数遥感估测模型为: y = 3.518x - 0.692 R~2 = 0.7046精度90.66% y = 33.67x~2 - 20.546x + 3.5414 R~2 = 0.7764精度90.39% y = 1.2257Ln(x) + 1.8351 R~2 = 0.682精度90.38% y = 4.4797x~2.0502 R~2 = 0.738精度90.51% y = 0.0658e5.8664x R~2 = 0.7578精度90.82%光谱植被指数估测模型为: NDVI: y = 0.2553e1.8456x R~2 = 0.6694精度77.85% RVI: y = 0.2947x0.6795 R~2 = 0.6328精度82.23% MSAVI: y = 1.6218x~2 - 0.7669x + 0.4611 R~2 = 0.6358精度80.55%本研究综合草地生态系统“植物-土壤-凋落物”的变化,总结出了一套基于遥感和光谱技术的草地健康快速评价新方法,可为相关研究人员提供参考。更多还原

【Abstract】 Hilly meadow steppes which degraded obviously were selected as sample plots to carry out the research in West Ujimqin Banner of Xilin Gol League in order to search for a method to evaluate grassland healthy extensively, quickly and successively which combined the traditional steppe resource ground survey technique with EOS-MODIS remote sensing technique and ground hyper- spectral technique.the division of degradation degrees on sample plots was carried out based on classification standard of natural steppe degradation, desertification and salinity (GB19377-2003), using ground data from June to August in 2007. The result showed that it was easy to cause deterioration evaluation errors if based on the data of June and July; the sample plots could be divided into four grades: non-degraded(CK), light, moderate and heavy degraded based on the data of August.the regularity that 13 vegetation indexes and 42 soil indexes varied with that degradation degree deepened was compared and analyzed. Eight vegetation indexes(forage cover degree, bare ground cover degree, litter cover degree, total biomass, litter weight, dominant species dominance index, degradation indicator plant dominance, grass height etc.) and five soil indexes (soil density, soil water content, soil organic carbon content from 0 to 5cm, soil total P content from 0 to 5cm, soil K content from 0 to 5cm etc.) were selected which have strong indicative to grassland degradation degree and have same regulation with degradation gradient.5 principal components which could character grassland degradation were selected using principal component analysis method. Health Assessment Index of grassland (RHI) was constructed using vague mathematics. The referred health thresholds were determined: non-degraded from 1.00 to 0.82, light degraded from 0.65 to 0.52, moderate-degraded from 0.58 to 0.48, heavy-degraded from 0.46 to 0.39 when hilly meadow grasslands were evaluated using RHI.correlation analysis was carried between vegetation index and RHI, single band spectrum and RHI using MODIS-NDVI index and data of high spectrum reflectance. Remote sensing model and spectrum model were constructed to evaluate RHI of the grasslands. The optimal estimation model was determined by the test of reliability and accuracy.Estimation model of MODIS-NDVI index was as follows: y = 3.518x - 0.692 R~2 = 0.7046 Accuracy 90.66%; y = 33.67x~2 - 20.546x + 3.5414 R~2 = 0.7764 Accuracy 90.39%; y = 1.2257Ln(x) + 1.8351 R~2 = 0.682 Accuracy 90.38%; y = 4.4797x~2.0502 R~2 = 0.738 Accuracy 90.51%; y = 0.0658e5.8664x R~2 = 0.7578 Accuracy 90.82%。Estimation model of spectral vegetation index was as follows: NDVI: y = 0.2553e1.8456x R~2 = 0.6694 Accuracy 77.85% RVI: y = 0.2947x0.6795 R~2 = 0.6328 Accuracy 82.23% MSAVI: y = 1.6218x~2 - 0.7669x + 0.4611 R~2 = 0.6358 Accuracy 80.55%。A set of new methods of grass health and evaluation was summaried based on remote sensing and spectroscopy technique integrated with the changes of grassland ecosystem of "plant - soil - litter", which could provide reference for related researchers.更多还原