灌丛化草原物种多样性与生产力的权衡关系

针对全球变化背景下内蒙古典型草原小叶锦鸡儿灌丛密度增加对生态系统主要服务功能影响尚不清晰的问题,该文采用野外调查为主的方法,研究小叶锦鸡儿灌丛化对内蒙古典型草原物种多样性和草地生产力的影响,探讨灌丛化过程中二者间是否存在权衡关系。结果表明：随着灌丛高度、盖度、密度和生物量的增加,草地生产力显著增加,从8.55 g/m²增加到21.76 g/m²;草本群落的物种多样性指数显著降低,物种丰富度从22种/m²降到12种/m²;草地生产力与物种多样性间存在明显权衡关系。结构方程模型证明灌丛化对物种多样性和草地生产力都是直接影响,不存在通过影响土壤养分而间接影响物种多样性或草地生产力。本研究结果为灌丛草原利用与管理提供重要理论依据。     

联合星载高光谱影像和堆栈集成学习回归算法的红树林冠层叶绿素含量遥感反演

红树林是世界上生产力最高、价值最高的湿地生态系统之一。冠层叶绿素含量CCC（Canopy Chlorophyll Content）作为红树林重要的生物物理参量,是估算其生产力和评价其健康状况的重要指标。本文利用珠海一号高光谱卫星（OHS）影像与Sentinel-2A多光谱数据计算传统植被指数与组合植被指数并构建了高维数据集,综合利用正态分布检验、最大相关系数法与变量重要性评价进行数据降维和变量优选;分别基于单一线性回归算法、机器学习回归算法和堆栈集成学习回归算法构建了红树林CCC遥感反演模型,探明北部湾红树林CCC的最佳遥感反演模型,验证OHS高光谱影像与Sentinel-2A数据反演红树林CCC的精度差异,评估SNAP-SL2P算法反演红树林CCC的适用性。研究结果表明：（1）通过数据降维和变量选择处理,从高维度OHS数据集选取了8个特征变量,其中RSI_（12,17）、DSI_（12,18）和NDSI_（6,12）组合植被指数对红树林CCC反演精度的贡献率较高;（2）联合OHS数据和最优堆栈GBRT集成学习回归模型（Score=0.999,RMSE=0.963 μg/cm²）的训练精度优于最优RF机器学习回归模型（RMSE降低了7.531 μg/cm²）,明显优于最优Lasso线性回归模型（RMSE降低了19.383 μg/cm²）;（3）在最优堆栈集成学习回归模型下,OHS数据反演红树林CCC的精度（R²=0.761,RMSE=16.738 μg/cm²）高于Sentinel-2A影像（R²=0.615,RMSE=20.701 μg/cm²）;（4）联合OHS和Sentinel-2A数据的最优堆栈集成学习回归模型反演红树林CCC的精度都明显优于SNAP-SL2P算法（R²=0.356,RMSE=49.419 μg/cm²）。研究结果论证了正态分布检验、最大相关系数法和基于XGBoost的特征选择方法有效降低了高维数据集的维度,并得到了最优特征变量;OHS数据的最优堆栈GBRT集成学习回归模型训练精度最高,是估算红树林CCC的最优反演模型;OHS和Sentinel-2A数据都能有效反演红树林CCC（R²均大于0.61）,而OHS数据的估算精度更高（R²大于0.75）;SNAP-SL2P算法不能有效反演红树林CCC（R²小于0.4）,且对红树林CCC数值存在系统性低估。     

基于WOFOST模型与UAV数据的玉米生长后期地上生物量估算

地上生物量能够有效反映作物的生长状态,其信息的实时估算对产量预测和农田生产管理都有重要意义。作物生长模型因其详尽的生理生化基础和对生长过程数字化描述能力,成为生物量估算的理想模型。近年来,研究人员利用数据同化算法将时间序列遥感数据同化到作物生长模型中,实现了作物模型由基于气象站的点模拟到区域尺度面模拟的外推,使生物量模拟结果同时具备大范围和机理性两个方面的特点。这一模式下,时间序列的遥感数据质量将对生物量模拟精度产生直接影响,作物生长后期受到光谱饱和的影响,遥感数据的作物冠层信息获取能力会出现明显下降,因此有必要对该阶段遥感数据和作物模型的结合方式进行优化,提升生物量模拟精度。本文针对东北地区春玉米生物量遥感估算存在的问题,提出了利用WOFOST作物模型结合无人机（UAV）遥感数据实现作物生长后期生物量准确估算的新思路。新思路首先利用多光谱遥感数据获取WOFOST模型具备较高空间异质性的土壤速效养分参数以提升模型的空间信息模拟能力,使其能在一定程度上摆脱点尺度模拟的限制。同时,结合集合卡尔曼滤波算法将生长前期无人机（UAV）遥感数据同化到模型中,以缩短模型单独运行时间,减少模型运行过程中的参数误差累积,实现无遥感数据参与下的短期作物生长模拟,并输出生长后期相应的生物量模拟结果。最后,本文利用地面实测数据对新方法的生物量模拟精度进行了评价。结果表明,与全生育期数据同化相比,新方法的生物量估算精度有了明显的提升（全生育期同化：R² = 0.45,RMSE = 4254.30 kg/ha;新方法：R²= 0.86,RMSE = 2216.79 kg/ha）。     

用地基激光雷达提取单木结构参数——以白皮松为例

以白皮松(Pinus bungeana Zucc)为研究对象,针对地基激光雷达TLS扫描的3维点云数据在单株木垂直方向的分布特征,提出了一种基于体元化方法的树干覆盖度变化检测方法,获取单木枝下高;然后根据获取的枝下高引入2维凸包算法获取垂直方向分层树冠轮廓,并计算树冠体积和冠幅;同时获取的单木参数还有胸径与树高。结果表明:单木枝下高的估测精度较高,R2与RMSE分别为0.97 m和0.21 m;胸径估测结果的R2与RMSE分别为0.79 cm和1.07 cm;采用逐步线性回归方法建立单木树冠体积与其他单木参数的相关关系,模型变量包括冠幅、叶子填充树冠长度和胸径,样本数为20,模型的R2与RMSE分别是0.967 m3和2.64 m3。本文方法能较准确地估测枝下高,TLS数据具有对树冠结构3维建模的潜力。     

中国南方森林冠顶高度Lidar反演—以江西省为例

激光雷达(Lidar)与光学遥感的有效结合对中国南方区域森林冠顶高度反演意义重大,而国产卫星将为中国森林生态研究提供新的数据源。本文联合利用大脚印激光雷达GLA和国产MERSI数据,在实现GLAS波形数据处理和不同地形条件下森林冠顶高度反演算法基础上,建立了区域尺度不同森林类型林分冠顶高度GLAS+MERSI联合反演关系模型,进行了江西地区森林冠顶高度反演。总体上,GLAS激光雷达森林冠顶高度估算精度较高;且在与MERSI 250 m数据的联合反演模型中,针叶林模型精度较好(R²=0.7325);阔叶林次之(R²=0.6095);混交林较差(R²=0.4068)。分析发现,考虑了光学遥感生物物理参数的GLAS+MERSI联合关系模型在区域森林冠顶高度估算中有较高精度,且在空间分布上与土地覆盖数据分布特征非常一致。     

利用无人机数码影像估算马铃薯地上生物量

株高和地上生物量AGB（Above-Ground Biomass）是作物长势监测的重要指标,因此快速获取这些信息对指导田间管理具有重要意义。本研究通过无人机搭载高清数码相机分别获取马铃薯5个生育期的影像数据,地面实测株高H（heigh）和AGB以及地面控制点GCPs（Ground Control Points）的三维空间坐标。首先,利用试验区域的影像数据结合GCPs的位置信息从生成的数字表面模型DSM（Digital Surface Model）中提取出马铃薯的株高（Hdsm）。其次,选取26种植被指数和H、Hdsm组成新的数据集与AGB作相关性分析,筛选出相关性较高的前7个植被指数同Hdsm作为估算马铃薯AGB的输入参数。然后,使用MLR（Multiple Linear Regression）、SVM（Support Vector Machine）和ANN（Artificial Neural Network）方法分别基于植被指数、植被指数和Hdsm构建马铃薯多生育期AGB估算模型,对不同估算模型进行比较分析,从而选择出AGB估算的最佳模型。结果表明：基于DSM提取的Hdsm与实测株高H高度拟合（R²=0.86,RMSE=6.36 cm,NRMSE=13.42%）;各生育期基于3种回归技术均以植被指数融入Hdsm构建的模型精度最高,估算能力最强;各生育期利用MLR方法构建的AGB估算模型效果最佳,其次为SVM-AGB估算模型,而ANN-AGB估算模型效果最差。该研究可为马铃薯AGB快速、无损监测提供科学参考。     

SAR与光学遥感影像的玉米秸秆覆盖度估算

秸秆是农田生态系统的重要组成部分。秸秆覆盖度（CRC）的遥感估算可以大范围、快速地获取地面秸秆覆盖信息，对保护性耕作的推广具有十分重要的意义。基于Sentinel-1 SAR影像和Sentinel-2光学影像分别构建了雷达指数与光学遥感指数，结合吉林省梨树县春秋两期实地采样数据，探究遥感指数与玉米秸秆覆盖度的相关性。为进一步提升玉米秸秆覆盖度的估算精度，结合雷达指数与光学遥感指数，采用最优子集回归的方法建立玉米秸秆覆盖度的估算模型，完成研究区的玉米秸秆覆盖度估算制图。结果表明：土壤质地分区建模可有效解决土壤异质性问题，提升反演精度。各遥感指数在秋季高覆盖时期的表现均优于春季低覆盖时期。STI和NDTI指数在光学遥感指数中表现最好，R²分别为0.701和0.697，而在雷达指数中，基于余弦矫正法的γVH0指数与实测CRC的相关性最高，R²为0.564。结合雷达指数与光学遥感指数能够有效地提高秸秆覆盖度估算精度，在最优子集回归法下基于结合指数构建的回归模型最优，R²为0.799，RMSE为13.67%，达到了较高的精度。研究结果为秸秆覆盖度估算的精度提升提供了一种新思路。     

利用Sentinel-2影像超分辨率重建的红树林冠层氮含量反演

氮素是植被整个生命周期的必要元素,红树林冠层氮素含量（CNC）遥感估算对红树林健康监测具有重要意义。以广东湛江高桥红树林保护区为研究区,本文旨在基于Sentinel-2影像超分辨率重建技术进行红树林CNC估算和空间制图。研究首先基于三次卷积重采样、Sen2Res和SupReMe算法实现Sentinel-2影像从20 m分辨率到10 m的重建;然后以重建后的影像和原始20 m影像为数据源构建40个相关植被指数,采用递归特征消除法（SVM-RFE）确定CNC估算的最优变量组合,进而构建CNC反演的核岭回归（KRR）模型;最后选取最优模型实现CNC制图。研究结果表明：基于Sen2Res和SupReMe超分辨率算法的重建影像不仅与原始影像具有很高的光谱一致性,且明显提高了影像的清晰度和空间细节。红树林CNC反演波段主要集中在红（B4）、红边（B5）、近红外波段（B8a）以及短波红外波段（B11和B12）,与“红边波段”相关的植被指数（RSSI和TCARIre1/OSAVI）也是红树林CNC反演的有效变量。基于3种方法重建后10 m的影像构建的模型反演精度（R²val>0.579）均优于原始20 m的影像（R²val=0.504）;基于Sen2Res算法重建影像构建的反演模型拟合精度（R²val=0.630,RMSE_val=5.133,RE_val=0.179）与基于三次卷积重采样重建影像的模型拟合精度（R²val=0.640,RMSE_val=5.064,RE_val=0.179）基本相当,前者模型验证精度（R²cv=0.497,RMSE_cv=5.985,RE_cv=0.214）较高且模型变量选择数量最为合理。综合重建影像光谱细节及模型精度,基于Sen2Res算法重建的Sentinel-2影像在红树林CNC估算中具有良好的应用潜力,能为区域尺度红树林冠层健康状况的精细监测提供有效的方法借鉴和数据支撑。     

光学与微波植被指数协同反演农作物叶面积指数的可行性分析

光学遥感是目前反演植被叶面积指数LAI（Leaf Area Index）的主要手段,但是当叶面积指数较大时存在光学遥感信息饱和、反演精度显著降低的问题。叶面积指数和平均叶倾角对光学、微波波段范围内反射和散射特性都有重要影响,主要表现在植被结构参数的变化可以引起冠层孔隙率和消光截面大小的改变。本文以典型农作物玉米为例,通过构建统一的PROSAIL和MIMICS模型输入参数,生成一套玉米全生长期光学二向反射率和全极化微波后向散射系数模拟库和冠层参数库。通过对模拟数据与LAI敏感性和相关性分析得出：（1）光学植被指数MNDVI_{(800 nm,2000 nm)},在LAI为0—3时敏感,基于MNDVI与LAI的回归模型可以估算LAI变化 0.4的情况,RMSE是0.33,R²是0.958。（2）微波植被指数SAR_{SRVI(1.4 GHz HH,9.6 GHz HV)},在LAI为3—6时敏感,基于SAR_SRVI与LAI的回归模型可以估算LAI变化1的情况,RMSE为0.22,R²是0.9839。研究表明,采用分段敏感的植被指数,协同光学和微波遥感反演玉米全生长期叶面积指数是可行的。     

基于遥感的区域尺度森林地上生物量估算研究

森林是陆地生态系统最大的碳库,精确估算森林生物量是陆地碳循环研究的关键。首先从机载LiDAR数据中提取高度和密度统计量,采用逐步回归模型进行典型样区生物量估算;然后利用机载LiDAR数据估算的生物量作为样本数据,与多光谱遥感数据Landsat8 OLI的波段反射率及植被指数建立回归模型,实现区域尺度森林地上生物量估算。实验结果显示,机载LiDAR数据估算的鼎湖山样区生物量与地面实测生物量的相关性R2达0.81,生物量RMSE为40.85 t/ha,说明机载LiDAR点云数据的高度和密度统计量与生物量存在较高的相关性。以机载LiDAR数据估算的生物量为样本数据,结合多光谱遥感数据Landsat8 OLI估算粤西北地区的森林地上生物量,精度验证结果为:R2为0.58,RMSE为36.9 t/ha;针叶林、阔叶林和针阔叶混交林等3种不同森林类型生物量的估算结果为:R2分别为0.51(n=251)、0.58(n=235)和0.56(n=241),生物量RMSE分别为24.1 t/ha、31.3 t/ha和29.9 t/ha,估算精度相差不大。总体上看,利用遥感数据可以开展区域尺度的森林地上生物量估算,为森林固碳监测提供有力的参考数据。     

Estimating residual biomass of olive tree crops using terrestrial laser scanning

Agricultural residues have gained increasing interest as a source of renewable energy. The development of methods and techniques that allow to inventory residual biomass needs to be explored further. In this study, the residual biomass of olive trees was estimated based on parameters derived from using a Terrestrial Laser Scanning System (TLS). To this end, 32 olive trees in 2 orchards in the municipality of Viver, Central Eastern Spain, were selected and measured using a TLS system. The residual biomass of these trees was pruned and weighed. Several algorithms were applied to the TLS data to compute the main parameters of the trees: total height, crown height, crown diameter and crown volume. Regarding the last parameter, 4 methods were tested: the global convex hull volume, the convex hull by slice volume, the section volume, and the volume measured by voxels. In addition, several statistics were computed from the crown points for each tree. Regression models were calculated to predict residual biomass using 3 sets of potential explicative variables: firstly, the height statistics retrieved from 3D cloud data for each crown tree, secondly, the parameters of the trees derived from TLS data and finally, the combination of both sets of variables. Strong relationships between residual biomass and TLS parameters (crown volume parameters) were found (R² = 0.86, RMSE = 2.78 kg). The pruning biomass prediction fraction was improved by 6%, in terms of R², when the variance of the crown-point elevations was selected (R² = 0.92, RMSE = 2.01 kg). The study offers some important insights into the quantification of residual biomass, which is essential information for the production of biofuel.     

Aboveground total and green biomass of dryland shrub derived from terrestrial laser scanning

Sagebrush (Artemisia tridentata), a dominant shrub species in the sagebrush-steppe ecosystem of the western US, is declining from its historical distribution due to feedbacks between climate and land use change, fire, and invasive species. Quantifying aboveground biomass of sagebrush is important for assessing carbon storage and monitoring the presence and distribution of this rapidly changing dryland ecosystem. Models of shrub canopy volume, derived from terrestrial laser scanning (TLS) point clouds, were used to accurately estimate aboveground sagebrush biomass. Ninety-one sagebrush plants were scanned and sampled across three study sites in the Great Basin, USA. Half of the plants were scanned and destructively sampled in the spring (n = 46), while the other half were scanned again in the fall before destructive sampling (n = 45). The latter set of sagebrush plants was scanned during both spring and fall to further test the ability of the TLS to quantify seasonal changes in green biomass. Sagebrush biomass was estimated using both a voxel and a 3-D convex hull approach applied to TLS point cloud data. The 3-D convex hull model estimated total and green biomass more accurately (R² = 0.92 and R² = 0.83, respectively) than the voxel-based method (R² = 0.86 and R² = 0.73, respectively). Seasonal differences in TLS-predicted green biomass were detected at two of the sites (p < 0.001 and p = 0.029), elucidating the amount of ephemeral leaf loss in the face of summer drought. The methods presented herein are directly transferable to other dryland shrubs, and implementation of the convex hull model with similar sagebrush species is straightforward.     

Estimating aboveground and organ biomass of plant canopies across the entire season of rice growth with terrestrial laser scanning

Non-destructive and accurate estimation of crop biomass is crucial for the quantitative diagnosis of growth status and timely prediction of grain yield. As an active remote sensing technique, terrestrial laser scanning (TLS) has become increasingly available in crop monitoring for its advantages in recording structural properties. Some researchers have attempted to use TLS data in the estimation of crop aboveground biomass, but only for part of the growing season. Previous studies rarely investigated the estimation of biomass for individual organs, such as the panicles in rice canopies, which led to the poor understanding of TLS technology in monitoring biomass partitioning among organs. The objective of this study was to investigate the potential of TLS in estimating the biomass for individual organs and aboveground biomass of rice and to examine the feasibility of developing universal models for the entire growing season. The field plots experiments were conducted in 2017 and 2018 and involved different nitrogen (N) rates, planting techniques and rice varieties. Three regression approaches, stepwise multiple linear regression (SMLR), random forest regression (RF) and linear mixed-effects (LME) modeling, were evaluated in estimating biomass with extensive TLS and biomass data collected at multiple phenological stages of rice growth across the entire season. The models were calibrated with the 2017 dataset and validated independently with the 2018 dataset.The results demonstrated that growth stage in LME modeling was selected as the most significant random effect on rice growth among the three candidates, which were rice variety, growth stage and planting technique. The LME models grouped by growth stage exhibited higher validation accuracies for all biomass variables over the entire season to varying degrees than SMLR models and RF models. The most pronounced improvement with a LME model was obtained for panicle biomass, with an increase of 0.74 in R² (LME: R² = 0.90, SMLR: R² = 0.16) and a decrease of 1.15 t/ha in RMSE (LME: RMSE =0.79 t/ha, SMLR: RMSE =2.94 t/ha). Compared to SMLR and RF, LME modeling yielded similar estimation accuracies of aboveground biomass for pre-heading stages, but significantly higher accuracies for post-heading stages (LME: R² = 0.63, RMSE =2.27 t/ha; SMLR: R² = 0.42, RMSE =2.42 t/ha; RF: R² = 0.57, RMSE =2.80 t/ha). These findings implied that SMLR was only suitable for the estimation of biomass at pre-heading stages and LME modeling performed remarkably well across all growth stages, especially for post-heading. The results suggest coupling TLS with LME modeling is a promising approach to monitoring rice biomass at post-heading stages at high accuracy and to overcoming the saturation of canopy reflectance signals encountered in optical remote sensing. It also has great potential in the monitoring of other crops in cloud-cover conditions and the instantaneous prediction of grain yield any time before harvest.     

Terrestrial laser scanning to estimate plot-level forest canopy fuel properties

This paper evaluates the potential of a terrestrial laser scanner (TLS) to characterize forest canopy fuel characteristics at plot level. Several canopy properties, namely canopy height, canopy cover, canopy base height and fuel strata gap were estimated. Different approaches were tested to avoid the effect of canopy shadowing on canopy height estimation caused by deployment of the TLS below the canopy. Estimation of canopy height using a grid approach provided a coefficient of determination of R² = 0.81 and an RMSE of 2.47 m. A similar RMSE was obtained using the 99th percentile of the height distribution of the highest points, representing the 1% of the data, although the coefficient of determination was lower (R² = 0.70). Canopy cover (CC) was estimated as a function of the occupied cells of a grid superimposed upon the TLS point clouds. It was found that CC estimates were dependent on the cell size selected, with 3 cm being the optimum resolution for this study. The effect of the zenith view angle on CC estimates was also analyzed. A simple method was developed to estimate canopy base height from the vegetation vertical profiles derived from an occupied/non-occupied voxels approach. Canopy base height was estimated with an RMSE of 3.09 m and an R² = 0.86. Terrestrial laser scanning also provides a unique opportunity to estimate the fuel strata gap (FSG), which has not been previously derived from remotely sensed data. The FSG was also derived from the vegetation vertical profile with an RMSE of 1.53 m and an R² = 0.87.     

Landscape Controls over Major Nutrients and Primary Productivity of Arctic Lakes

Increasing surface temperatures in the Arctic are affecting the dynamics between lakes and their landscapes. In this paper, we use landscape metrics for land cover and statistical analysis to explore the interactions between such measures as shape and patch density indices for land cover and lake primary productivity. The objective was to identify metrics that could be used to predict lake primary productivity, as measured by chlorophyll a, total nitrogen and total phosphorus estimates. Land cover and landscape metrics for the Toolik region of Alaska were derived using satellite imagery and Fragstats software. The metrics, treated as independent variables in a stepwise regression, were derived for two levels of land cover. The first comprised the entire watershed studied; the second was derived using buffers created around water channels within each watershed. A statistically significant model for each dependent variable was obtained. Results suggest that, of the metrics tested; those related to broad leaf vegetation complexes were most useful in predicting lake primary productivity. The Landscape Shape Index for riparian patches and the Patch Density for heath complex were the two most important metrics for predicting variation in chlorophyll a (p<0.001, r² = 0.52). For total nitrogen estimates, the most significant metrics were Percentage of riparian complex and Patch Density for fen complex (p<0.001, r² = 0.48). Total phosphorus estimates were most influenced by the Patch Density for shrub complex, the Mean Shape Index for moist acidic tundra complex, and the Patch Density for aquatic vegetation (p<0.001, r² = 0.52).     

Automatic Tree Identification and Diameter Estimation Using Single Scan Terrestrial Laser Scanner Data in Central Indian Forests

Forest inventory parameters, primarily tree diameter and height, are required for several management and planning activities. Currently, Terrestrial Laser Scanning (TLS) is a promising technology in automated measurements of tree parameters using dense 3D point clouds. In comparison with conventional manual field inventory methods, TLS systems would supplement field data with detailed and relatively higher degree of accurate measurements and increased measurement frequency. Although, multiple scans from TLS captures more area, they are resource and time consuming to ensure proper co-registration between the scans. On the other hand, Single scans provide a fast and recording of the data but are often affected by occlusions between the trees. The current study evaluates potential of single scan TLS data to (1) develop an automatic method for tree stem identification and diameter estimation (diameter at breast height—DBH) using random sample consensus (RANSAC) based circle fitting algorithm, (2) validate using field based measurements to derive accuracy estimates and (3) assess the influence of distance to scanner on detection and measurement accuracies. Tree detection and diameter measurements were validated for 5 circular plots of 20 m radius using single scans in dry deciduous forests of Betul, Madhya Pradesh. An overall tree detection accuracy of 85 and 70% was observed in the scanner range of 15 and 20 m respectively. The tree detection accuracies decreased with increased distance to the scanner due to the decrease in visible area. Also, estimated stem diameter using TLS was found to be in agreement with the field measured diameter (R² = 0.97). The RMSE of estimated DBH was found to be 3.5 cm (relative RMSE ~20%) over 202 trees detected over 5 plots. Results suggest that single scan approach suffices the cause of accuracy, reducing uncertainty and adds to increased sampling frequency in forest inventory and also implies that TLS has a seemingly high potential in forest management.     

Comparing the spectral settings of the new generation broad and narrow band sensors in estimating biomass of native grasses grown under different management practices

The challenge of assessing and monitoring the influence of rangeland management practices on grassland productivity has been hampered in southern Africa, due to the lack of cheap earth observation facilities. This study, therefore, sought to evaluate the capability of the newly launched Sentinel 2 multispectral imager (MSI) data, in relation to Hyperspectral infrared imager (HyspIRI) data in estimating grass biomass subjected to different management practices, namely, burning, mowing and fertilizer application. Using sparse partial least squares regression (SPLSR), results showed that HyspIRI data exhibited slightly higher grass biomass estimation accuracies (RMSE = 6.65 g/m², R² = 0.69) than Sentinel 2 MSI (RMSE = 6.79 g/m², R² = 0.58) across all rangeland management practices. Student t-test results then showed that Sentinel 2 MSI exhibited a comparable performance to HyspIRI in estimating the biomass of grasslands under burning, mowing and fertilizer application. In comparing the RMSEs derived using wave bands and vegetation indices of HyspIRI and Sentinel, no statistically significant differences were exhibited (α = 0.05). Sentinel (Bands 5, 6 and 7) and HyspIRI (Bands 730 nm, 740 nm, 750 nm, 710 nm), as well as their derived vegetation indices, yielded the highest predictive accuracies. These findings illustrate that the accuracy of Sentinel 2 MSI data in estimating grass biomass is acceptable when compared with HyspIRI. The findings of this work provide an insight into the prospects of large-scale grass biomass modeling and prediction, using cheap and readily available multispectral data.     

Using multi-polarization C- and L-band synthetic aperture radar to estimate biomass and soil moisture of wheat fields

Biomass and soil moisture are two important parameters for agricultural crop monitoring and yield estimation. In this study, the Water Cloud Model (WCM) was coupled with the Ulaby soil moisture model to estimate both biomass and soil moisture for spring wheat fields in a test site in western Canada. This study exploited both C-band (RADARSAT-2) and L-band (UAVSAR) Synthetic Aperture Radars (SARs) for this purpose. The WCM-Ulaby model was calibrated for three polarizations (HH, VV and HV). Subsequently two of these three polarizations were used as inputs to an inversion procedure, to retrieve either soil moisture or biomass without the need for any ancillary data. The model was calibrated for total canopy biomass, the biomass of only the wheat heads, as well as for different wheat growth stages. This resulted in a calibrated WCM-Ulaby model for each sensor-polarization-phenology-biomass combination. Validation of model retrievals led to promising results. RADARSAT-2 (HH-HV) estimated total wheat biomass with root mean square (RMSE) and mean average (MAE) errors of 78.834 g/m² and 58.438 g/m²; soil moisture with errors of 0.078 m³/m³ (RMSE) and 0.065 m³/m³ (MAE) are reported. During the period of crop ripening, L-band estimates of soil moisture had accuracies of 0.064 m³/m³ (RMSE) and 0.057 m³/m³ (MAE). RADARSAT-2 (VV-HV) produced interesting results for retrieval of the biomass of the wheat heads. In this particular case, the biomass of the heads was estimated with accuracies of 38.757 g/m² (RSME) and 33.152 g/m² (MAE). For wider implementation this model will require additional data to strengthen the model accuracy and confirm estimation performance. Nevertheless this study encourages further research given the importance of wheat as a global commodity, the challenge of cloud cover in optical monitoring and the potential of direct estimation of the weight of heads where wheat production lies.     

Evaluation of machine learning models for rice dry biomass estimation and mapping using quad-source optical imagery

ABSTRACT

Several machine learning regression models have been advanced for the estimation of crop biophysical parameters with optical satellite imagery. However, literature on the comparative performances of such models is still limited in range and scope, especially under multiple data sources, despite the potential of multi-source imagery to improving crop monitoring in cloudy areas. To fill in this knowledge gap, this study explored the synergistic use of Landsat-8, Sentinel-2A, China’s environment and disaster monitoring and forecasting satellites (HJ-1 A and B) and Gaofen-1 (GF-1) data to evaluate four machine learning regression models that include Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbor (k-NN), and Gradient Boosting Decision Tree (GBDT), for rice dry biomass estimation and mapping. Taking a major rice cultivation area in southeast China as case study during the 2016 and 2017 growing seasons, a cross-calibrated time series of the Enhanced Vegetation Index (EVI) was obtained from the quad-source optical imagery and on which the aforementioned models were applied, respectively. Results indicate that in the before rice heading scenario, the most accurate dry biomass estimates were obtained by the GBDT model (R² of 0.82 and RMSE of 191.8 g/m²) followed by the RF model (R² of 0.79 and RMSE of 197.8 g/m²). After heading, the k-NN model performed best (R² of 0.43 and RMSE of 452.1 g/m²) followed by the RF model (R² of 0.42 and RMSE of 464.7 g/m²). Whist the k-NN model performed least in the before heading scenario, SVM performed least in the after heading scenario. These findings may suggest that machine learning regression models based on an ensemble of decision trees (RF and GBDT) are more suitable for the estimation of rice dry biomass, at least with optical satellite imagery. Studies that would extend the evaluation of these machine learning models, to other parameters like leaf area index, and to microwave imagery, are hereby recommended.