Hyperspectral Reflectance Data and Agronomic Traits Can Predict Biomass Yield in Winter Rye Hybrids
Hyperspectral Reflectance Data and Agronomic Traits Can Predict Biomass Yield in Winter Rye Hybrids
复制标题
DOI：
10.1007/s12155-019-10080-z
复制
发表时间：
2020-03-01
影响因子：
3.6
通讯作者：
Miedaner, Thomas
中科院分区：
文献类型：
--
作者：
Galan, Rodrigo Jose;Bernal-Vasquez, Angela-Maria;Miedaner, Thomas
冬黑麦是一种潜在的替代沼气基质，传统上是为了提高粮食产量而种植的。因此，我们的目标是评估结合从可见光到近红外光谱数据得出的植被指数作为农艺性状替代指标预测干物质产量的可能性。共有404个精品黑麦杂交种进行了谷物产量测试，其中274个杂交种的子集也在德国的4个地点进行了2年的干物质产量测试，使环境总数达到8个。光谱数据是由无人机在两个日期的太阳正午前后通过无人机收集的，测量的波长范围为410到993 nm。供试杂交种间籽粒产量差异在3.64~10.53 mg ha(-1)之间，干物质产量差异在8.44~14.66 mg ha(-1)之间。23个植被指数和株高、籽粒产量、干物质产量等农艺性状的遗传力大多表现为中高遗传力(h(2)和0.5)，它们的遗传方差显著(P<；0.001)不为零。在高干物质产量的间接选择中，株高比籽粒产量更具信息量。通过多元回归方法建立的高光谱数据和农艺数据相结合的指数显示，交叉验证的预测能力为0.75，比仅包含农艺性状的模型改进了约6%。在早期选择阶段，所开发的指数可以作为在粮食产量试验的基础上经济有效地选择生物量试验的优良候选者的合适工具。
Winter rye (Secale cereale L.), a potential alternative biogas substrate, is traditionally bred for grain yield. Thus, our objective was to evaluate the possibility to predict dry matter yield combining vegetation indices derived from visible to NIR spectral data as a proxy for agronomic traits. A total of 404 elite rye hybrids were tested for grain yield, and a subset of this comprising 274 hybrids were also tested for dry matter yield over 2 years and at 4 locations in Germany bringing the total number of environments to 8. Spectral data were collected around solar noontime on mostly clear sky by an unmanned aerial vehicle (UAV) on two dates and measured in the wavelength range between 410 and 993 nm. Observed variation among tested hybrids ranged between 3.64-10.53 Mg ha(-1) for grain yield and 8.44-14.66 Mg ha(-1) for dry matter yield across different sites. The 23 vegetation indices and the agronomic traits, such as plant height, grain yield, and dry matter yield, showed mostly moderate to high heritability estimates (h(2) > 0.50), and their genetic variances were significantly (P < 0.001) different from zero. Plant height was more informative than grain yield for indirect selection of high dry matter yield. An index combining hyperspectral and agronomic data developed by a multiple regression procedure showed a cross-validated prediction ability of 0.75, representing an improvement of about 6% to a model incorporating only agronomic traits. During earlier selection stages, the developed index could be a suitable tool for the cost-effective selection of superior candidates for biomass trials based on grain yield trials.