A comparison of cover calculation techniques for relating point-intercept vegetation sampling to remote sensing imagery

Accurate and timely spatial predictions of vegetation cover from remote imagery are an important data source for natural resource management. High-quality in situ data are needed to develop and validate these products. Point-intercept sampling techniques are a common method for obtaining quantitative information on vegetation cover that have been widely implemented in a number of local and national monitoring programs. The use of point-intercept data in remote sensing projects, however, is complicated due to differences in how vegetation cover indicators can be calculated. Decisions on whether to use plant intercepts from any canopy layer (i.e., any-hit cover) or only the first plant intercept at each point (i.e., top-hit cover) can result in discrepancies in cover estimates which are used to train remotely-sensed imagery. Our objective in this paper was to explore the theory of point-intercept sampling relative to training and testing remotely-sensed imagery, and to test the strength of relationships between top-hit and any-hit methods of calculating vegetation cover and high-resolution satellite imagery in two study areas managed by the Bureau of Land Management in northwestern Colorado and northeastern California. We modeled top-hit and any-hit percent cover for six vegetation indicators from 5m-resolution RapidEye imagery using beta regression. Model performance was judged using normalized root mean-squared error (RMSE) from a 5-fold cross validation. Any-hit cover estimates were significantly higher (α < 0.05) than top-hit cover estimates for forbs and grasses in the White River study area, but only marginally higher in Northern California. Pseudo-R² values for beta regression models of vegetation cover from RapidEye image information varied from 0.1525 to 0.7732 in White River and 0.2455 to 0.6085 in Northern California, with little pattern to whether any-hit or top-hit indicators produced better model fit. However, normalized RMSE was lower for any-hit cover (indicating better model performance) or minimally higher than top-hit cover for all indicators in each study area. Our results do not support the idea that top-hit cover estimates from point-intercept sampling are the most appropriate for remote sensing applications in arid and semi-arid shrub-steppe environments. In fact, having two sets of different indicators calculated from the same data may cause additional confusion in a situation where there is already considerable debate on how vegetation cover should be measured and used. Ultimately, selection of indicators to use for developing remote sensing classification or predictive models should be based first on the meaning or interpretation of the indicator in the ecosystem of interest, and second on how well the indicator performs in modeling applications.     

The cover of living and dead corals from airborne remote sensing

Trends in coral cover are widely used to indicate the health of coral reefs but are costly to obtain from field survey over large areas. In situ studies of reflected spectra at the coral surface show that living and recently dead colonies can be distinguished. Here, we investigate whether such spectral differences can be detected using an airborne remote sensing instrument. The Compact Airborne Spectrographic Imager (Itres Research Ltd, Canada) was flown in two configurations: 10 spectral bands with 1-m² pixels and 6 spectral bands with 0.25-m² pixels. First, we show that an instrument with 10 spectral bands possesses adequate spectral resolution to distinguish living Porites, living Pocillopora spp., partially dead Porites, recently dead Porites (total colony mortality within 6 months), old dead (>6 months) Porites, Halimeda spp., and coralline red algae when there is no water column to confuse spectra. All substrata were distinguished using fourth-order spectral derivatives around 538 nm and 562 nm. Then, at a shallow site (Tivaru) at Rangiroa Atoll, Tuamotu Archipelago (French Polynesia), we show that live and dead coral can be distinguished from the air to a depth of at least 4 m using first- and fourth-order spectral derivatives between 562–580 nm. However, partially dead and recently dead Porites colonies could not be distinguished from an airborne platform. Spectral differences among substrata are then exploited to predict the cover of reef substrata in ten 25-m² plots at nearby Motu Nuhi (max depth 8 m). The actual cover in these plots was determined in situ using quadrats with a 0.01-m² grid. Considerable disparity occurred between field and image-based measures of substrate cover within individual 25-m² quadrats. At this small scale, disparity, measured as the absolute difference in cover between field and remote-sensing methods, reached 25% in some substrata but was always less than 10% for living coral (99% of which consisted of Porites spp.). At the scale of the reef (all ten 25-m² quadrats), however, disparities in percent cover between imagery and field data were less than 10% for all substrata and extremely low for some classes (e.g. <3% for living Porites, recently dead Porites and Halimeda). The least accurately estimated substrata were sand and coralline red algae, which were overestimated by absolute values 7.9% and 6.6%, respectively. The precision of sampling was similar for field and remote-sensing methods: field methods required 19 plots to detect a 10% difference in coral cover among three reefs with a statistical power of 95%. Remote-sensing methods required 21 plots. However, it took 1 h to acquire imagery over 92,500 m² of reef, which represents 3,700 plots of 25 m² each, compared with 3 days to survey 10 such plots underwater. There were no significant differences in accuracy between 1-m² and 0.25-m² image resolutions, suggesting that the advantage of using smaller pixels is offset by reduced spectral information and an increase in noise (noise was observed to be 1.6–1.8 times greater in 0.25-m² pixels). We show that airborne remote sensing can be used to monitor coral and algal cover over large areas, providing that water is shallow and clear, and that brown fleshy macroalgae are scarce, that depth is known independently (e.g. from sonar survey).     

作物缺水指数法的简化及在干旱遥感监测中的应用

从能量平衡原理出发,对潜在蒸散的计算进行了简化,从而对作物缺水指数法干旱遥感监测模型进行了简化、简化后的模型涉及因子减少,计算量明显降低,更接近于实际应用．利用该方法及NOAA/AVHRR卫星遥感资料和有关气象要素资料,对陕西省关中地区春季干旱进行了监测．结果表明,简化后的作物缺水指数法仍然充分考虑了下垫面的植被覆盖状况和地面风速、水汽压等气象要素,对该区春季干旱的监测效果好于使用植被供水指数法的监测效果．从简化后模型的计算量、时效响应和实际对比结果来看,该简化模型可以进行实际应用．另外,在计算过程中局部出现了CWSI>1的情况,这在实际中是不可能的,文中对此也进行了初步讨论．     

Environmental controls of NDVI and sheep production in the Tierra del Fuego steppe of Argentina

Abstract. We analysed vegetation dynamics in Tierra del Fuego steppes using Normalized Difference Vegetation Index (NDVI) data provided by advanced very high‐resolution radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration (NOAA) polar satellite. Our objective, at a regional scale, was to analyse the spatial variability of NDVI dynamics in relation to parent material and geographic location, representing the fertility and climate gradients respectively; at a local scale, it was to analyse the inter‐annual variability associated with climate and its relation with sheep production indices. The general pattern of NDVI dynamics was analysed with Principal Component Analysis. We found that the geographic location was more important than landscape type in explaining NDVI dynamics despite the fact that the variation in landscape type reflects a fertility gradient strongly associated with floristic composition and secondary productivity. Discriminant Analysis was performed to identify the variables that better distinguish geographic units. The Northern region (with the lowest precipitation and the highest temperatures) had lower NDVI values over the year. In the Central region, NDVI reached the highest value of the season, surpassing both other regions. The Southern region (the coldest and moistest) had its growth pattern displaced towards the summer. For the Central region we analysed 10 years of monthly NDVI data with PCA. We found that precipitation from August to December and winter temperature are the most important determinants of overall NDVI values. Lamb production was correlated with spring and early summer NDVI values. Sheep mortality is affected by low NDVI values in late summer and high annual amplitude. Satellite information allowed us to characterize the vegetation dynamics of three ecological areas across the Fuegian steppe.     

大尺度不透水面遥感估算方法比较——以京津唐为例

城市不透水面既是常用的城市化程度指标,也是衡量环境质量的重要指标。采用遥感技术准确提取城市不透水面并分析其空间扩张过程,对生态城市建设具有重要意义。基于Landsat 5 TM影像,采用NDVI二元法和线性光谱分解法,分别提取北京、天津和唐山3个城市不透水面信息,并将不透水面估算结果与近同期的ALOS影像提取结果对比验证。结果表明,线性光谱分解法获取的不透水面结果较好,RMSE为20.6%,能有效提取大范围的不透水面信息。     

Interannual variability of NDVI and its relationship to climate for North American shrublands and grasslands

Abstract. Our objective was to analyse the interannual variability of different characteristics of the seasonal dynamics of NDVI and their relationships with climatic variables for grassland and shrubland sites of North America. We selected twenty-five sites located in relatively undisturbed areas. We analysed the variability of seven traits derived from the annual dynamics of the NDVI at each site: the annual integral, the difference between maximum and minimum NDVI, the dates of the inflection points of a double logistic model fitted to the NDVI curve, the difference between these dates, the date of maximum NDVI, and the coefficient of determination of the double logistic model. The temporal variability of traits that integrated aspects of primary productivity over the year was lower than those related to seasonality. This suggests that from year to year, grassland and shrubland ecosystems would differ more in the timing of production and senescence than in the total amount of carbon fixed. The integral of NDVI showed less temporal variability than annual precipitation. The coefficient of variation of both precipitation and the NDVI integral were positively related. The slope of the relationship was significantly lower than 1, indicating that the variability of ecosystem function is a lower proportion of the variability of annual precipitation in areas with a high relative variability of this climatic variable than in areas of low variability. The variability of most of the NDVI traits analysed showed a negative and, in general, non-linear relationship with annual precipitation. The same kind of relationship has been reported elsewhere for annual precipitation and its coefficient of variation. Mean annual precipitation has been reported as the main control of above-ground net primary production in grassland and shrubland ecosystems. Our results suggest that this climatic variable is also associated with the interannual variability of carbon gains, such as the primary production and its seasonality.     

两代AVHRR GIMMS NDVI数据集的对比分析——以新疆地区为例

最新发布的1981—2012年的AVHRR GIMMS NDVI3g数据为了解区域植被的近期变化状况提供了数据基础。深入理解该版本与老版本GIMMS NDVIg(1981—2006年)之间的关系,对于使用新数据时充分利用已有老版本的研究结果具有重要意义。以我国西北干旱区的典型区域——新疆为例,研究了两个数据集在反映生长季、春季、夏季和秋季植被现状,植被变化趋势及其对气候变化响应方面的异同。研究结果表明:两个数据集在描述植被活动空间分布、变化趋势及其与气候的相关性方面大体相似,但在数值、动态变化率及其对气候变化响应强度等方面存在的差异也不容忽略。NDVI3g数据生长季和各季节NDVI数值多大于NDVIg,尤其是在夏季和在植被覆盖较好的区域。区域尺度,NDVI3g所反映的植被变化趋势更为平稳,尤其是在夏季和较长的时段,这可能与像元尺度NDVI3g显著增加范围小于NDVIg,而显著减少范围多于NDVIg有关。两个数据集对气温、降水量、潜在蒸散发和湿润指数的响应具有大体一致的空间格局,但对气候因子变化的敏感性存在差异,哪一个数据集更为灵敏依赖于不同的气候因子和时段。一般规律是NDVI3g与热量因子显著正相关的区域小于NDVIg,而与水分因子显著正相关的区域则大于NDVIg。利用长期的生态数据集,尽快理清两个数据集在表征植被变化之间的异同并建立两者的转换关系,对于合理开展植被变化、碳平衡、生态系统服务功能评估等广泛利用NDVI数据的相关研究十分重要。     

Long-term continuous monitoring of dissolved oxygen in cell culture medium for perfused bioreactors using optical oxygen sensors

For long-term growth of mammalian cells in perfused bioreactors, it is essential to monitor the concentration of dissolved oxygen (DO) present in the culture medium to ascertain the health of the cells. An optical oxygen sensor based on dynamic fluorescent quenching was developed for long-term continuous measurement of DO for NASA-designed rotating perfused bioreactors. Tris(4,7-diphenyl-1,10-phenanthroline) ruthenium(II) chloride is employed as the fluorescent dye indicator. A pulsed, blue LED was chosen as the excitation light source. The sensor can be sterilized using an autoclave. The sensors were tested in a perfused rotating bioreactor supporting a BHK-21 (baby hamster kidney) cell culture over one 28-day, one 43-day, and one 180-day cell runs. The sensors were initially calibrated in sterile phosphate-buffered saline (PBS) against a blood-gas analyzer (BGA), and then used continuously during the entire cell culture without recalibration. In the 180-day cell run, two oxygen sensors were employed; one interfaced at the outlet of the bioreactor and the other at the inlet of the bioreactor. The DO concentrations determined by both sensors were compared with those sampled and measured regularly with the BGA reference. The sensor outputs were found to correlate well with the BGA data throughout the experiment using a single calibration, where the DO of the culture medium varied between 25 and 60 mm Hg at the bioreactor outlet and 80-116 mm Hg at the bioreactor inlet. During all 180 days of culture, the precision and the bias were +/-5.1 mm Hg and -3.8 mm Hg at the bioreactor outlet, and +/- 19 mm Hg and -18 mm Hg at inlet. The sensor dynamic range is between 0 and 200 mm Hg and the response time is less than 1 minute. The resolution of the sensor is 0.1 mm Hg at 50 mm Hg, and 0.25 mm Hg at 130 mm Hg.