基于环境卫星数据的黄河湿地植被生物量反演研究

回归模型拟合植被指数与生物量的定量关系是植被生物量反演的重要研究方法之一.研究在此基础上,基于环境卫星遥感数据和同步野外实地采样数据,以郑州黄河湿地自然保护区为试验区,比较MLRM(多元线性回归模型)与SCRM(一元曲线回归模型)反演植被生物量的能力,并估算研究区植被生物量,生成研究区生物量分布图.结果表明,文中所建立的MLRM在研究区具有较好的反演精度和预测能力.其模型显著性检验为极显著,相关系数为0.9791,模型拟合精度达到29.8 g/m2,其模型预测结果系统误差为49.9g/m2,均方根误差为67.2 g/m2,预测决定系数为0.8742,比传统的一元回归模型具有更高的精度和可靠性.估算研究区域2010年8月湿生植被生物量约为6.849199 t/hm2,相对误差为4.73％.

The study of vegetation biomass inversion based on the HJ satellite data in Yellow River wetland

Wetland vegetation is the important component of the wetland ecological system, and its biomass is one of the major indexes to measure the primary productivity. Remote sensing technology can be used effectively to extract information of wetland vegetation, which directly reflects their growth and development. Hence the research is very significant for the environment monitoring and protection of wetland ecosystems. This paper discusses the quantitative relationship between vegetation spectrum index and dry biomass of vegetation, based on the remote sensing data from China HJ-1A satellite and synchronous field sampling data, choosing the Zhengzhou Yellow River Wetland Nature Reserve as the study area. The methods based on vegetation indexes in this study are mainly used in vegetation biomass inversion in a large area, and have conducted quite successful algorithms and models. While a single vegetation index as input factor has a high accuracy and sensitivity in fitting the vegetation biomass model, its prediction result encounter larger errors and cannot accurately reflect true information of vegetation biomass. Authors in this paper puts variety of vegetation indexes together as the comprehensive input factors, and estimates vegetation biomass in wetland by using regression analysis method. Therefore, it improves largely the precision and reliability of vegetation biomass inversion based on multiple vegetation indexes. In this paper, the biomass of vegetation is abstracted as a phenomenon, while different kinds of vegetation indexes are considered as different influencing factors. The relationships between them are conceptualized into different mathematical models. Authors inspect the precision of inversion by comparing the estimate results from the two kinds of methods after their using the SCRM (single curve regression model) biomass method and MLRM (multiple linear regression model) biomass method respectively. Results showed that the MLRM have good precision and prediction ability, which can be well applied to estimate the wetland vegetation. The model test is significant (hitting a level of 0.000), while the model correlation coefficient is 0.9791, fitting accuracy reaches 29.8 g/m². The model prediction results has a system error of 49.9 g/m², the determine coefficient is 0.8742, the RMS error is 67.2 g/m², estimation of dry vegetation biomass total to 6.849199 t/hm² in study area, with the actual biomass estimated dry biomass has a difference of 0.323 749 t/hm², the result is 6.525 450 t/hm², and the relative error is 4.73%, which shows an accurate estimation on the biomass of wetland vegetation by using vegetation indexes extracted from HJ-1A remote sensing data, combined with field sampling data. Wetland is one of the three large ecological systems globally, which takes a key role on the plant biodiversity protection. And it has a great environmental function and lots ecological benefits as the most productive ecological system. Method of the MLRM has a better precision and forecasting ability by comparing that of SCRM in vegetation biomass inversion when used to estimating wetland vegetation biomass. As a new method of the vegetation detection and estimation based on remote sensing technology, it has a large application value to the ecological resource management and wetland ecological environment protection.