Search

目的/意义人工智能（Artificial Intelligence，AI）技术已在学术和工程应用领域掀起了研究高潮，在地球物理参数和农业气象遥感参数反演方面也表现出了强大的应用潜力。目前大部分AI技术在地学和农学的应用还是“黑箱”，没有物理意义或缺乏可解释性及通用性。为了促进AI在地学和农学的应用和培养交叉学科的人才，本研究提出基于AI耦合物理和统计方法的地球物理参数反演范式理论。方法首先基于物理能量平衡方程进行物理逻辑推理，从理论上构造反演方程组，然后基于物理推导构建泛化的统计方法。通过物理模型模拟获得物理方法的代表性解以及利用多源数据获得统计方法代表性的解作为深度学习的训练和测试数据库，最后利用深度学习进行优化求解。［结果和讨论］判定形成具有通用性和物理可解释的范式条件包括：（1）输入与输出变量（参数）之间必须存在因果关系；（2）输入和输出变量（参数）之间理论上可以构建闭合的方程组（未知数个数少于或等于方程组个数），也就是说输出参数可以被输入参数唯一确定。如果输入参数（变量）和输出参数（变量）之间存在很强的因果关系，则可以直接使用深度学习进行反演。如果输入参数和输出参数之间存在弱相关性，则需要添加先验知识来提高输出参数的反演精度。此外，本研究以农业气象遥感中的关键参数地表温度、发射率、近地表空气温度和大气水汽含量联合反演作为案例对理论进行了证明，分析结果表明本理论是可行的，并且可以辅助优化设计卫星传感器波段组合。结论本理论和判定条件的提出在地球物理参数反演史上具有里程碑意义。

Land surface temperature (LST) is a fundamental parameter in global climate, environmental, and geophysical studies. Remote sensing is an essential approach for obtaining large-scale and frequently updated LST data. However, due to the wide field of view of remote sensing sensors, the observed LST with diverse view geometries suffers from inconsistency caused by the thermal radiation directionality (TRD) effect, which results in LST products being incomparable, especially during daytime. To address this issue and correct current off-nadir LSTs to nadir LSTs, a semi-physical time-evolved kernel-driven model (TEKDM) is proposed, which depicts multitemporal TRD patterns during the daytime. In addition, we employ a Bayesian optimization method to calibrate seven unknown parameters in the TEKDM. Validation results using the U.S. Climate Reference Network (USCRN) sites show that the RMSE (MBE) for GOES-16 and MODIS off-nadir LST products is reduced from 3.29 K (−2.0 K) to 2.34 K (−0.02 K), with an RMSE reduction of 0.95 K (29%) and a significant reduction in systematic bias. Moreover, the proposed method successfully eliminates the angular and temporal dependence of the LST difference between the satellite off-nadir LST and in situ nadir LST. In summary, this study presents a feasible approach for estimating the high-accuracy nadir LST, which can enhance the applicability of LST products in various domains. [ABSTRACT FROM AUTHOR]

Accurate predictions of sea surface temperature (SST) are crucial due to the significant impact of SST on the global ocean-atmospheric system and its potential to trigger extreme weather events. Many existing machine-learning-based SST predictions adapt the traditional iterative point-wise prediction mechanism, whose predicting horizons and accuracy are limited owing to the high sensitivity to cumulative errors during iterative predictions. Therefore, this paper proposes a novel granulation-based long short-term memory (LSTM)-random forest (RF) combination model that can fully capture the feature dependencies involved in the fluctuation of SST sequences, reduce the cumulative error in the iteration process, and extend the prediction horizons, which includes two sub-models (adaptive granulation model and hybrid prediction model). They can restack the one-dimensional SST time-series into multidimensional feature variables, and achieve a strong forecasting ability. The analysis shows that the proposed model can achieve more accurate prediction-hours in nearly all prediction ranges from 1 to 125 h. The average prediction error of the proposed model in 25–125 h is 0.07 K, similar to that (0.067 K) in the first 24 h, which exhibits a high generalization performance and robustness and isthus a promising platform for the medium- and long-term forecasting of hourly SSTs. [ABSTRACT FROM AUTHOR]

The ear is the top layer of mature wheat and has very different geometric and thermal characteristics from that of leaves. Compared to the directional brightness temperature (DBT) of wheat canopy without ears, the DBT at the ear stage has specific features, and the ear effects could not be explained by previous models. This paper proposes a hybrid geometric optical and radiative transfer model to reveal the combined influences of the geometric structure of ears and leaf; the temperature distribution of ear, leaf, and soil; and the sun-target-sensor geometry on the canopy DBT. The soil, leaf, and ear layers are taken into account in the model so it is named as the Soil Leaf Ear Combined (SLEC) DBT model. We compare the model prediction with the field measurement data. The results show that the new SLEC DBT model can simulate the DBT of wheat at the ear stage with an accuracy of 0.78 K. Index Terms--Directional brightness temperature (DBT), ear, hybrid model, wheat.