The accurate and timely estimates of crop physiological growth stages are essential for efficient crop management and precise modeling of agricultural systems. Satellite remote sensing has been widely used to retrieve vegetation phenology metrics at local to global scales. However, most of these phenology metrics (e.g., green-up) are different from crop growth stages (e.g., emergence) used in crop management and modeling. As such, an integrated framework referred to as PhenoCrop was developed to: 1) establish a connection between remote sensing-derived phenology metrics and key crop growth stages based on Wang and Engle plant phenology model and 2) use fused MODIS-Landsat 30 m 8-day reflectance data generated using Kalman Filter-based data fusion technique to produce onset dates of key growth stages of corn (Zea mays L.) and soybeans (Glycine max L.) at 30 m spatial resolution. In this paper, we described the PhenoCrop framework, and tested its performance for the State of Nebraska for 2012–2016 by comparison to observations of estimated key growth stages at four experimental sites, and state-level statistical data from Crop Progress Reports (CPRs) published by the United States Department of Agriculture’s (USDA) National Agricultural Statistical Services (NASS). In addition, to evaluate the suitability of using coarse or high spatial resolution satellite imagery, fused MODIS-Landsat-based estimates were compared with those produced using EOS MODIS 250 m (MOD9Q1) reflectance data.

The PhenoCrop estimates captured the typical spatial trends of gradual delay in the progression of the growing season from southeast to northwest Nebraska. Also inter-annual differences due to factors such as weather fluctuations and change in management strategies (e.g., early season in 2012) were evident in the estimates. Validation results revealed that average root mean square error (RMSE) of the state-level estimates of corn and soybean growth stages ranged from 1.10 to 4.20 days and from 3.81 to 7.89 days, respectively, while pixel level estimates had a RMSE ranging from 3.72 to 8.51 days for corn and 4.76–9.51 days for soybean growth stages. Although MODIS 250 m based estimates showed similar general spatial patterns observed in the fused MODIS-Landsat based estimates, the accuracy and ability to capture field scale variations was improved with fused MODIS-Landsat data. Overall, results showed the ability of PhenoCrop framework to provide reliable estimates of crop growth stages that can be highly useful in crop modeling and crop management during the growing season.

准确，及时地估计作物生理生长期对有效的作物管理和农业系统的精确建模至关重要。卫星遥感已广泛用于从本地到全球范围检索植被物候指标。但是，大多数这些物候指标（例如绿色）与作物管理和建模中使用的作物生长阶段（例如出苗）不同。因此，开发了一个称为PhenoCrop的集成框架，以：1）基于Wang和Engle植物物候模型建立遥感衍生的物候指标与关键作物生长阶段之间的联系，以及2）使用融合的MODIS-Landsat 30 m 8使用基于卡尔曼滤波器的数据融合技术生成的一天反射率数据，以生成玉米关键生长阶段的开始日期（Zea mays L. ）和大豆（Glycine max L.）的空间分辨率为30 m。在本文中，我们描述了PhenoCrop框架，并通过与四个实验点的估计关键生长阶段的观察结果以及作物进展报告（CPR）的州级统计数据进行比较，测试了其在内布拉斯加州的2012-2016年绩效由美国农业部（USDA）国家农业统计服务（NASS）出版。此外，为了评估使用粗略或高分辨率的卫星图像的适用性，将基于融合MODIS-Landsat的估计值与使用EOS MODIS 250 m（MOD9Q1）反射率数据产生的估计值进行了比较。并通过与四个实验点的估计关键生长阶段的观察结果以及美国农业部（USDA）发布的《作物进展报告》（CPR）的州级统计数据进行比较，测试了其在2012-2016年内布拉斯加州的绩效）国家农业统计局（NASS）。此外，为了评估使用粗略或高分辨率的卫星图像的适用性，将基于融合MODIS-Landsat的估计值与使用EOS MODIS 250 m（MOD9Q1）反射率数据产生的估计值进行了比较。并通过与四个实验点的估计关键生长阶段的观察结果以及美国农业部（USDA）发布的《作物进展报告》（CPR）的州级统计数据进行比较，测试了其在2012-2016年内布拉斯加州的绩效）国家农业统计局（NASS）。此外，为了评估使用粗略或高分辨率的卫星图像的适用性，将基于融合MODIS-Landsat的估计值与使用EOS MODIS 250 m（MOD9Q1）反射率数据产生的估计值进行了比较。

PhenoCrop估计值记录了从内布拉斯加州东南部到西北部生长季节进程中逐渐延迟的典型空间趋势。估算中还明显显示出由于天气波动和管理策略变化（例如，2012年初）等因素造成的年度间差异。验证结果表明，国家和地区对玉米和大豆生长阶段的估计的平均均方根误差（RMSE）分别为1.10至4.20天和3.81至7.89天，而像素水平的估计均方根误差为3.72至玉米生育期为8.51天，大豆生育期为4.76-9.51天。尽管基于MODIS 250 m的估计显示了在融合的基于MODIS-Landsat的估计中观察到的相似的一般空间模式，融合的MODIS-Landsat数据提高了捕获场尺度变化的准确性和能力。总体而言，结果表明，PhenoCrop框架能够提供可靠的作物生长阶段估计值，这在生长期的作物建模和作物管理中非常有用。