遥感与作物生长模型数据同化应用综述

2018-11-05 07:52黄健熙马鸿元高欣然刘峻明张晓东朱德海
农业工程学报 2018年21期
关键词:尺度作物观测

黄健熙,黄 海,马鸿元,卓 文,黄 然,高欣然,刘峻明,苏 伟,李 俐,张晓东,朱德海



遥感与作物生长模型数据同化应用综述

黄健熙1,2,黄 海1,马鸿元1,卓 文1,黄 然1,高欣然1,刘峻明1,2,苏 伟1,2,李 俐1,2,张晓东1,2,朱德海1,2

(1. 中国农业大学土地科学与技术学院,北京 100083;2. 农业农村部农业灾害遥感重点实验室,北京 100083)

遥感是获取大面积地表信息最有效的手段,在农业资源监测、作物产量预测中发挥着不可替代的重要作用;作物生长模型能够实现单点尺度上作物生长发育的动态模拟,可对作物长势以及产量变化提供内在机理解释。遥感信息和作物生长模型的数据同化有效结合二者优势,在大尺度农业监测与预报上具有巨大的应用潜力。该文系统综述了遥感与作物生长模型的同化研究,概述了遥感与作物生长模型数据同化系统的构建,在归纳国内外研究进展的基础上,总结了当前主流同化方法的特点以及在不同条件下的同化效果。进而具体分析影响同化精度的关键环节,明确了相关科学概念,并相应指出改善精度的策略或者方向。最后从多参数协同、多数据融合、动态预测、多模型耦合以及并行计算环境5个方面展望了遥感与作物生长模型数据同化的未来研究重点和发展趋势,同时结合农业应用现实需求,介绍一种数据同化与集合数值预报结合的应用框架,为大区域、高精度同化研究提供新的思路与借鉴。

作物;遥感;模型;作物生长模型;数据同化;农业监测;产量预报

0 引 言

遥感技术因其宏观性、直观性和大面积获取能力等特点在各个领域中应用广泛。随着遥感技术的长足发展,特别是时间、空间、光谱分辨率的不断提升,为全球陆地自然资源、大气和海洋监测等提供了重要的技术支撑。随着定量遥感反演算法和产品的日益完善,如叶面积指数(leaf area index,LAI)、蒸散发(evapotranspiration, ET)、土壤水分(soil moisture,SM)、吸收性光合有效辐射(fraction of absorbed photosynthetically active radiation,FAPAR)及地上生物量(aboveground biomass,AGB)等关键生物理化参数的定量产品,在区域尺度的农作物监测中发挥了重要作用。

作物生长模型是根据作物品种特性、气象条件、土壤条件以及作物管理措施,采用数学模型方法描述作物光合、呼吸、蒸腾、营养等机理过程,能够以特定时间步长动态模拟作物生长和发育期间的生理生化参数、结构参数以及作物产量,定量地描述光、温、水、肥等因子以及田间栽培和管理措施对作物生长和发育的影响[1-5]。在单点尺度上,基于作物光合、呼吸、蒸腾、营养等机理过程的作物生长模型依靠其内在的物理过程和动力学机制,可以准确模拟作物在单点尺度上生长发育的时间演进以及产量的形成动态过程。

利用数据同化技术把遥感反演参数信息融入到作物机理过程模型是当前改进区域作物生长模拟精度的重要途径[6-8]。当作物生长模型应用到区域尺度时,地表、近地表环境非均匀性决定了作物模型中的初始条件、土壤参数、作物参数、气象强迫因子空间分布的不确定性和获取资料的困难性。卫星遥感具有空间连续和时间动态变化的优势,能够有效解决作物模型中区域参数获取困难这一瓶颈[9-10]。然而,由于受卫星时空分辨率等因素的制约,遥感对地观测还不能真正揭示作物生长发育和产量形成的内在过程机理、个体生长发育状况及其与环境气象条件的关系,而这正是作物模型的优势所在。数据同化技术通过耦合遥感观测和作物模型,能实现两者的优势互补,提高区域作物生长过程模拟能力。将遥感信息引入作物生长模型,是促进大面积作物长势监测和产量预测向机理化和精确化方向发展的有效技术途径。

本文从现有文献中整理归纳,分析作物模型与遥感数据同化系统的基本构架,总结当前研究中模型、数据、方法等现状,明确同化系统中影响精度的关键环节和重点内容,讨论未来研究的技术要点和发展趋势。

1 遥感与作物生长模型数据同化系统的构建

数据同化的研究思想最早是由Charney等[11]在1969年提出,数据同化方法被逐渐应用于大气环流模式,例如数值天气预报、海洋预报模式、陆面模式等地学模拟系统。数据同化是集成观测和模式/模型这2种基本科学研究手段的重要方法,它能够将多源的、时间/空间不完整的观测整合到一个演进的作物生长过程模型中,从而更加准确一致地估计作物生长过程的各个状态变量。一般来说,一个数据同化系统均包含3个基本组成部分:动态模型、观测数据和同化算法。

图1为一个典型遥感与作物模型数据同化系统的流程图。作物生长模型选择WOFOST作物生长模型,在单点样本尺度(田间尺度)进行充分观测获得模型输入参数。在参数率定和模型本地化的基础上,WOFOST能够对作物的生长发育过程以及LAI、SM、生物量、单产进行较为准确的模拟。同时基于贝叶斯理论的马尔科夫链蒙特卡洛方法(Markov chain Monte Carlo, MCMC),获得这些参数的后验分布,实现对参数的估计,同时参数的后验分布能够定量表达在已有观测条件下模型参数的不确定性。光学和雷达遥感能定量反演出关键的农作物参数,例如生育期(development stage, DVS)、LAI、ET、SM、FAPAR、AGB等。因此,引入大区域的遥感参数、借助数据同化技术,在区域每个格网上对状态变量进行优化或者经过多次迭代优化出一套模型参数,实现对区域作物模型的优化,提高区域作物单产的模拟。特别通过耦合短临天气预报数据和作物生长模型,可以实现对未来时段农作物产量的预测。

注:MCMC、4Dvar、EnKF、SAR、DVS、LAI、ET、SM、FAPAR、AGB、CC分别表示马尔科夫链蒙特卡洛方法、四维变分、集合卡尔曼滤波、合成孔径雷达、生育期、叶面积指数、蒸散发、土壤水分、吸收性光合有效辐射、地上生物量、冠层覆盖度。下同。

2 遥感与作物生长模型数据同化研究进展

数据同化研究中,将动态模型(作物生长模型)与观测(遥感或地面试验)耦合的同化算法发挥着核心的作用,同化算法的性能直接影响着同化系统的运行效率和精度。基于代价函数的参数优化方法和基于估计理论的集合滤波方法是2类主要的现代数据同化方法。已有众多学者对遥感与作物模型的同化策略进行了系统性的综述[6,12-14],本文依据代表性文献的归纳整理,进一步分析2类同化方法的技术要点、研究进展以及内在差异等。

参数优化方法迭代调整作物模型中与生长发育和产量形成密切相关的、常规方式难以获得的参数或初始条件,最小化遥感观测值与模型模拟值之间的差异,以达到优化作物模型的目的(如图2所示)。参数优化算法精度主要取决于同化变量、优化算法以及目标函数(或遗传算法中的适应度函数)形式。用于作物模型同化的优化算法包括单纯型搜索算法、最大似然法、复合型混合演化算法(shuffled complex evolution method developed at the University of Arizona, SCE-UA)、Powell 共轭方向法、粒子群算法(particle swarm optimization, PSO)、遗传算法、模拟退火法等;代价函数的构建有均方根误差、最小二乘、三维变分(3DVar)、四维变分(4DVar)等形式。关于参数优化方法的代表性研究如表1所示。

图2 参数优化法同化原理示意图

顺序数据同化算法又称为滤波算法,其算法核心是在机理过程模型的动力框架内,融合来自于不同分辨率的遥感观测信息,让机理过程模型和各种观测算子集成为不断地依靠外部观测而自动调整模型轨迹,并且减小误差的预报系统(如图3所示)。顺序滤波使得模型模拟的状态变量不断更新为最优预报值,是一种时间连续且可应用于实时模拟的数据同化方法。常用到的顺序滤波算法有扩展卡尔曼滤波(extended Kalman filter, EKF)、集合卡尔曼滤波(ensemble Kalman filter, EnKF)和粒子滤波(particle filter, PF)等顺序同化算法。EnKF具有处理非线性观测算子的能力,并解决了同化过程中预报误差协方差求解困难的瓶颈,是顺序同化方法中的重要方法。EnKF假定观测和模型为高斯分布,而PF则可以假定观测和模型为非高斯分布,已经成为顺序同化方法中最具潜力的方法。关于顺序滤波同化的代表性研究如表2所示。

表1 参数优化法同化的主要研究

注:“A+B”形式的作物模型名称表示作物模型与辐射传输模型的耦合,其中A是作物模型名称,B是辐射传输模型名称; LNA、TSAVI、NDVI、EVI、、、CNA分别表示叶片氮积累量、转换型土壤调整指数、归一化植被指数、增强型植被指数、后向散射系数、波段反射率、冠层氮素累积量。下同。

Note: Crop growth model names in the form of “A+B” represent the coupling of crop growth model and the radiation transfer model, where A is the name of the crop model and B is the name of the radiation transfer model; LNA, TSAVI, NDVI, EVI,,and CNA respectively represent leaf nitrogen accumulation, transformed soil adjusted vegetation index, normalized vegetation index, enhanced vegetation index, backscatter coefficient, band reflectance, and canopy nitrogen accumulation. The same below.

图3 顺序滤波法同化原理示意

综上所述,在遥感与作物生长模型的数据同化研究中,大区域的同化应用的卫星遥感数据以MODIS为主,中等区域尺度上主要选择Landsat TM, ETM+及OLI等遥感数据。作物模型以WOFOST、CERES等使用最为广泛,作物对象主要选择小麦、玉米和水稻等粮食作物。LAI是遥感与作物模型同化中最常用的同化变量。此外,也有研究通过耦合作物模型与辐射传输模型,直接同化遥感观测和耦合模型模拟的反射率、植被指数或后向散射系数。总体上,产量的估算是最主要的应用目标。

表2 顺序滤波法同化主要研究

注:VTCI表示条件温度植被指数。

Note: VTCI represent vegetation temperature condition index.

方法上,参数优化方法依据遥感观测在同化单元上重新估计模型的一些参数或初始条件,从而实现模型在空间上的有效拓展。虽然代价函数形式和参数优化方法各异,但参数优化精度与遥感观测的频率和时间点密切相关[15,19,23]。而当作物模型与辐射传输模型耦合时,参数优化的效果很大程度依赖于辐射传输模型部分的参数精度,尤其在大区域尺度上,这些参数具有较大空间变异性[15]。4DVar是当前参数优化方法主流代表,其代价函数描述了模型初始参数的优化值与初始值的距离以及同化变量的遥感观测值和模拟值之间的距离。4DVar能够更好地同化异步观测数据,被广泛认为是一种有效而具有竞争力的同化方法[30,66-68];其主要缺点是在解析法来解时需要伴随模式的计算,对于作物生长模型这样的复杂系统,往往带来很大的计算难度和不确定性[30,32-33,66,69]。

以EnKF为代表的顺序滤波算法通过将冠层的连续观测信息纳入模型模拟,以减少被同化的状态变量的误差,从而提高模型模拟的准确性。EnKF的优点在于假设观测和模拟都存在误差,同时相比于其他同化算法,EnKF公式简洁,计算高效,更容易表达出作物生长模型非线性、高维度的特性[57,59,70]。总的来说,在具有高质量观测值进行模型标定的情况下,EnKF在相对较小的研究区域上能够取得较高的产量模拟精度[27,46];在中等研究区域,EnKF使用中低分辨率遥感数据(如MODIS LAI或者微波遥感反演的土壤水分等)进行同化估产也能够具有较好的表现[58,60]。但是当同化的状态变量(如LAI)与目标模拟产物(如产量、土壤水等)相关性较弱时,EnKF并不一定会提高同化的准确性[59,61]。尤其对于产量而言,其与气候、土壤、品种、管理等因素密切相关,仅依靠一两个状态变量的更新,并不能有效实现对模型模拟偏差的纠正[55,57]。此外,EnKF同化表现也取决于遥感观测质量、模型及观测不确定性的量化程度[71]。

参数优化方法与顺序滤波方法的区别在于,前者用整个同化窗口内的观测值来重新调整模型参数,而后者的观测值是顺序作用于模型,每一次后续的观测值只会影响从当前状态之后的模型变化性质。有研究表明,在短同化窗口内,EnKF可以取得更高的分析精度,而在长同化窗口,EnKF与4DVar的具有相似的准确性[70]。但对于不同的模型,同化结果可能存在差异,比如在WOFOST模型的同化研究中,EnKF方法同化LAI会引起“物候漂移”现象进而降低同化精度,而参数优化方法则能够取得较好结果[55]。

3 影响同化精度关键环节

3.1 同化单元大小

同化单元(最小分辨率)的大小选择常常取决于要解决的应用问题。例如,全球尺度的模拟一般需要10~50 km分辨率;国家尺度的模拟需要1~10 km分辨率;区域尺度模拟需要10 m~1 km分辨率。此外,数据同化单元大小也取决于作物模型输入参数(气象要素、作物和土壤以及田间管理)和遥感反演参数的时空分辨率。同化单元越小,空间差异性越显著。但同化单元的减小不会一直提高同化精度,而是存在一个最优同化单元,并与农田地块大小有紧密联系。通常,更细的同化单元会带来巨大的计算压力。因此,需要研发更高效的同化策略和适合于高性能计算的组织架构与模式。与格网同化单元不同的是,划分均质地块单元,基于单元进行同化,提高同化执行效率的策略,是未来遥感与作物生长模型数据同化值得探索的一个研究方向[22]。

3.2 遥感参数反演的不确定性

准确评估遥感反演参数的不确定性,并将其考虑到数据同化之中,对于提高数据同化系统的精度具有重要作用。随着定量遥感反演算法及产品不断完善,反映作物生长状态的关键生物理化参数,如LAI、ET、SM、FAPAR和AGB等,其获取途径越来越丰富,精度越来越高,能够不断满足作物生长模型同化的需求。参数反演精度取决于遥感数据源和反演方法,前者可分为单一遥感数据和多源遥感数据,后者通常包含经验模型、辐射传输模型、神经网络模型等。以LAI为例,MOD15数据集以MODIS为数据源,以三维辐射传输模型为主、经验模型为辅,生成覆盖全球的时间分辨率8 d、空间分辨率1 km的LAI产品[72],GEOV1数据集以VEGETATION为数据源,融合MOD15和CYCLOPES为训练数据,训练神经网络,生成覆盖全球的时间分辨率为10 d、空间分辨率为1/112°的LAI产品[73]。由于植被结构和生物物理特性的多样性、冠层和大气辐射传输过程的复杂性,植被参数和遥感观测间的转换仍存在较大的不确定性[74]。例如,有研究表明MODIS LAI产品对农作物LAI低估约33%~50%[75-76];MODIS ET产品在森林地区与实测值较为一致,农田地区的一致性较差[77],在流域或站点尺度上,其与实测值的相关系数大约在0.7左右[78-79]。因而遥感科学领域的真实性检验技术的发展,对遥感与作物模型数据同化系统中遥感反演参数的不确定性评估具有重要影响。

3.3 作物生长模型的不确定性

作物生长模型的不确定性主要来源于模型结构、模型参数以及气象驱动数据。模型结构的不确定性主要体现为模型对光合作用、水肥、营养和土壤水平衡等过程难以定量和准确描述,同时对诸如病虫害、极端灾害天气、渍灾等减产因素影响未在作物模型中考虑,也影响作物生长发育和结果输出的模拟效果。模型参数的不确定性主要反映在模型中部分初始田间管理条件和参数难以直接获取,传统的参数估计方法的目标是在一些特定的模型结构内找到一组最优参数[80]。例如,研究表明约一半的研究者采用“试错法”进行模型标定[81],即依据一定数量的实测值,通过调整几个特定参数,当模型模拟与实测的误差达到一定要求时,则以此时的参数作为标定参数,此方法具有较大主观性。一些学者通过构建反映模型模拟值与实际观测值差异的目标函数,通过最小化其差异,从而获得模型参数的估计值[82-83]。Liu[84]调整CERES-Maize模型的物候系数,直到模拟与观测的物候日期相一致,以实现参数估计。Thorp等[85]计算不同生长季内CERES-Maize模型模拟产量与实际产量的均方根误差(RMSE)作为目标函数,采用模拟退火算法(simulated annealing optimization algorithm)实现对模型参数的自动化估计。在复杂的过程模型中,不存在精确的反解,因此依靠观测结果构建的目标函数或者适应度函数,通过优化算法只得到唯一的参数估计是不可行的[86]。基于贝叶斯理论的马尔科夫链蒙特卡洛算法(MCMC)能够求得模型参数的后验分布,因而得到越来越多的应用[87-89]。气象数据是模型中驱动作物生长发育的重要数据,为了获得空间连续、时间连续的气象驱动数据集,往往需要使用插值方法得到区域范围气象数据。然而降水、风速等非连续宏观现象空间分布不均,使用插值方法的可靠性一直存在争议。另一方面,在进行作物产量预测预报时,气象预报数据的不确定性也将直接影响作物生长的模拟效果,是制约模型实际应用的瓶颈之一。目前的研究主要是采用历史天气数据、天气发生器或数值预报作为预报期驱动数据[90-91]。由于大气系统高度非线性、混沌的特征,气候和天气预报的不确定性是必然的[92-93],而依赖于气象驱动的作物生长模型也是如此。

3.4 数据同化策略及参数结合点

在数据同化策略方面,顺序滤波同化效率高,但也存在不足之处。首先顺序滤波同化直接修改状态变量(例如:LAI或SM),而时间序列LAI的变化容易导致作物生育期改变,因此,同化LAI通常会引起一定程度的“物候漂移”,导致同化精度不如参数优化方法[55]。其次诸如各种滤波算法如卡尔曼滤波等方法中对于模型和观测的误差隐含着预先假设,即模型和观测都存在随机误差而不存在系统偏差,而实际上区域化的模型运行难以满足这一条件。因此,滤波同化往往会迅速收敛至模型或观测,使得数据同化失去效果。更为普遍的是,作物模型和遥感观测存在尺度差异,这种差异已经不能够作出没有系统偏差的假设,因此在顺序滤波中观测信息只能选择站点尺度[94]、或者利用其他手段进行观测数据的修 正[45,51],这限制了顺序滤波的区域化应用。

通过构建各种形式的代价函数来优化参数的同化方法可以在一定程度上减小模型观测尺度不一致带来的系统偏差,而且没有顺序滤波引起物候漂移的缺点,因此,在作物模型同化领域应用中得到了广泛的应用。然而,优化方法也存在一些制约因素。首先,和顺序滤波方法相比,优化法的数据同化需要大量迭代计算搜索最优参数集合,而由于模型非线性的特征无法应用解析法,使得同化系统的运行效率偏低,并且随着参数的增加所需的搜索次数呈指数增长;算法的选择和代价函数的设计虽然能够缓解该问题,但随着区域尺度同化需求的增加,计算效率始终是优化方法的主要瓶颈。其次,优化算法和顺序滤波的最主要的差异是同化的时间周期,顺序滤波中同化的模拟是实时进行的,随着观测值的输入不断向前分析更新,当模拟结束时同化也随之结束,而优化方法是由算法在外部调用作物模型进行多次循环模拟,每一次迭代都需要完整运行整个时间周期,这使得优化法在进行实时模拟预报时不如顺序滤波灵活高效。

3.5 尺度效应及转换模型

由于农田表面存在复杂性,在某一尺度上观测到的地物性质、过程原理、形成规律,在另一尺度上可能一致、可能相似,也可能无效而需要进行修正,加之遥感具有多空间分辨率的数据特点,从定量遥感出发的地学描述必然存在多尺度问题,即遥感的尺度效应[95-98]。作物模型输入参数的空间分辨率格网大小对模型输出结果会有不同的精度,这是作物模型本身的尺度效应。遥感与作物模型数据同化系统中的尺度效应是指,遥感反演参数和作物模型模拟的参数(LAI、ET、SM以及AGB等)之间的差异以及不匹配。在遥感与作物模型数据同化方法的应用过程中:一方面,由于作物生长模型中不同参数的区域化方法不同,导致区域参数空间尺度不一致,解决不同空间尺度区域参数的空间匹配问题是实现局地模型空间尺度扩展和区域应用的前提;另一方面,由于地表空间异质性以及作物生长模型的非线性特点,导致尺度效应明显,遥感观测数据与作物生长模型变量之间的尺度不匹配仍是一个难题。因此,空间尺度转换是遥感与作物模型数据同化系统应用到区域尺度需解决的关键科学问题。

国内外学者针对空间尺度转换已开展了大量研究,思路主要有两个:一是模型整体的尺度转换,即对特定尺度下建立的物理定律、定理、模型以及概念进行全面修正[99-102];二是模型参数(或变量)的尺度转换,即对不同观测尺度(空间分辨率)下所获取的地表生物物理参数进行尺度差异校正[103-108]。尺度转换过程可分为向上尺度变换(将高空间分辨率信息转换成低分辨率的过程)和向下尺度变换(将低空间分辨率信息转换成高分辨率的过程)。然而,通过向下尺度变换方法对空间异质性进行建模是一个极其复杂的过程,需要严格的近似和先验知识,不易于程序操作和实际应用[109]。因此,面向遥感与作物模型数据同化,一种常用的尺度不匹配解决方案是模型参数的升尺度转换,即将物候信息与低空间分辨率的遥感数据结合起来,并通过中高分辨率影像反演而得的相对精确值,调整作物生长模型生成的同化参数轨迹,从而提高同化精度[27]。

4 遥感与作物生长模型数据同化研究趋势

4.1 单参数向多参数的转变

LAI是遥感与作物生长模型中最常用的同化变量,由于LAI是一个农作物光温水肥多要素交互作用后的综合指标,同化LAI难以定量描多要素对作物生长发育的影响。LAI能刻画冠层的生长和发育过程,决定了叶片光的截获和光合作用的大小。SM和ETa/ETp反映了土壤水的状况和作物胁迫的程度。因此,在雨养区域,联合同化LAI和SM,LAI和ETa/ETp能修正冠层生长发育和土壤水平衡过程。包姗宁等研究表明,在水分胁迫模式下, 同化ET和LAI双变量能取得比同化ET或LAI单变量更高的精度[28],张树誉等同化LAI和条件植被温度指数(vegetation temperature condition index, VTCI)提高了模型的估产精度[65]。值得一提是,在同化变量的选择方面,需要考虑不同变量之间的相关性,有些变量之间的相关性很强,选择其中一个变量即可,例如LAI与FAPAR, SM和ETa/ETp。

4.2 单一遥感数据源向多数据源的转变

光学数据易受云雨影响,常导致关键生育期监测数据缺失,在时间和信息量上已不能完全满足遥感与作物模型数据同化的需求。微波遥感具有全天候工作的特性,且具有一定的穿透性,能够提供多云多雨天气条件下的农田地表信息,同时弥补了光学数据在时效性和全天候的不足。研究表明,LAI、生物量等农学参数与SAR后向散射系数显著相关[110-113],这为SAR数据与作物生长模型的同化奠定了理论基础。未来,联合光学数据和微波遥感数据进行数据同化工作,将是遥感与作物模型数据同化的趋势。目前欧洲航天局哥白尼计划(GMES)中的地球观测卫星哨兵1号(Sentinel-1)和哨兵2号(Sentinel-2)为此提供了较为理想的数据来源。其中哨兵1号载有C波段合成孔径雷达,不受光照与气候条件限制,可提供全天时、全天候的连续影像;而哨兵2号是目前唯一具有三个红边波段的民用观测卫星,在大区域的农作物监测方面具有较大的应用潜力。两者的结合,能够提供高重访周期及分辨率的大区域监测数据,很大程度上满足遥感与作物模型数据同化系统的需求。

4.3 监测向预测的转变

现有的作物模型与遥感数据的同化研究绝大部分是在“重现”历史作物生长过程,即利用往年实测的整个作物生育期气象数据进行驱动模型。这种往往属于事后的产量监测(估测),因此,在作物生育期内已有的气象观测数据以及年型的分析上,通过引入未来时段的天气预报数据,使得作物生长模型具有预测能力,实现提前半个月到1个月的产量预测。

传统的作物模型运行过程只是输出数据到输出数据的单一映射,无法提供模型模拟的不确定性信息。作物模型本身以及输入参数、气象驱动存在着不确定性,如果以集合的思想将作物模型放入一个外部框架内,用不同集合预报数据驱动模拟的结果作为集合成员,可以使模拟结果的集合代表其概率分布,最终将单一数值的模拟输出转换为概率分布,实现作物产量的概率预测。

作物产量的同化预估也是农业应用的现实需求,本文基于已有研究[27,114-115],介绍一种数据同化与集合数值预报结合的应用框架。以WOFOST模型为例,通过与PROSAIL模型耦合直接实现与中高分辨率遥感影像(例如Sentinel-2影像)进行同化。为降低高分辨率遥感同化引起的海量计算压力,引入一种“分块聚类”策略,即:1)在作物生长季节将遥感影像按时间序列叠加在一起,并将其裁剪为数个网格单元(比如10 km´10 km);2)对分块后的全生长季时间序列反射率进行聚类分析,每个网格单元被聚类为指定数目类别(比如40个),然后为每个聚类类别分配此类别中所有像素的平均反射率值作为该类别的反射率值;3)对网格单元中的所有聚类运行数据同化算法,从而为网格单元中每个聚类类别获得同化的产量;4)使用同化后每个聚类的产量和聚类分析图恢复出空间上的产量分布图,从而获得了原高空间分辨率下的产量空间分布。气象数据由播种日至预报时刻的气象数据观测、TIGGE集合预报[116-117]、与集合预报成员最相似历史年份气象资料3部分拼接而成,在模型标定的基础上,驱动整个生育期的模型运行,并实现由集合预报数据得到作物产量的概率预报。

4.4 单一作物模型向多作物模型耦合的转变

目前主流的作物生长模型有CERES、WOFOST、APSIM、AquaCrop、STICS等,由于作物实际生长的机理过程的复杂性,单一的作物生长模型往往根据不同应用需求目的,对作物生长多过程的模拟各有侧重。例如CERES模型以光能因子为主要驱动,通过不同模块集成对不同作物进行模拟[3];WOFOST模型以CO2同化作为关键模拟,侧重于产量的形成和定量描述[2];在WOFOST基础上开发的SWAP模型则更细致对土壤水分进行模拟,在土壤水平衡过程模拟上具有优势[118];APSIM模型以土壤水、盐作为主要驱动,在模拟土壤要素对作物生长的影响上得到了广泛验证[119];AquaCrop是典型的土壤水分驱动模型,强调于水分胁迫对作物产量形成的影响[4];STICS模型[120]则较好地考虑了管理因素对作物生长的影响。因此,集成与整合多个作物模型,实现不同作物模型在多过程模拟中的优势互补,能更为准确模拟作物生长与土壤、气象、水肥以及管理措施等的交互影响。国际上,多作物模型集合研究已经取得了一些进展[121-122]。特别是将集合预报的技术引入到多作物模型中,有望能进一步提高作物生长模型在不确定性条件下的预测模拟精度。

4.5 单机同化模式向高性能并行计算模式的转变

传统的遥感与作物模型数据同化系统中所有的输入参数,是以文件系统的形式进行存储和管理。对输入输出的读写效率较低,对于大区域的海量计算具有局限性。GPU在浮点运算和并行计算方面相对于CPU有着巨大的优势,采用GPU模式可以大幅提高计算速度[123]。GPU加速能提高计算速度,却对于复杂逻辑计算效率不高,且代码编写调试复杂。基于Hadoop MapReduce实现的分布式算法能够在批量处理数据的任务中相对于单机实现的算法有极大的性能优势[124]。然而,在许多迭代计算的任务中,每一次迭代过程的中间数据集都需要从硬盘中加载,如此频繁的IO操作会消耗大量时间,限制了基于MapReduce在多次迭代算法上的速度性能优势。将作物生长模型输入参数以栅格数据的形式存储到大数据环境下(例如HBase),在同化过程中动态生成作物模型的输入参数,同时采用SPARK的内存计算模式实现同化过程中需要的多次迭代过程,能极大提高计算效率[125],满足全国尺度遥感与作物模型数据同化系统的快速计算要求。值得一提的是,基于谷歌地球引擎(Google Earth Engine, GEE)强大的计算资源和丰富的遥感数据[126],在GEE平台下实现遥感与作物模型数据同化系统,能极大提升在10~30 m中等分辨率尺度的同化模型计算效率,为大区域范围的地块尺度的遥感同化产量预测提供了有效的技术途径,也将是遥感与作物模型同化应用的重要发展方向。

5 结 论

当前遥感与作物生长模型的同化研究在模型、数据和算法上已经相对成熟,但随着近年来对地观测技术不断发展,以及考虑到农业应用在时间、空间上的现实需求,同化更丰富类型、更高分辨率的遥感数据对现有策略存在较大挑战。因而未来在完善和创新同化系统架构的同时,也需要进一步提高性能与效率。

1)农业系统的复杂性决定了遥感与作物生长模型同化的较大不确定性,需要通过模型参数优化、遥感定量反演、数据同化算法等关键技术的共同发展,才能进一步提高同化系统精度。

2)为更精确描述大区域作物生长状态、满足服务农业生产的实际需求,多参数协同、多数据融合、动态预测以及多模型耦合成为遥感与作物生长模型数据同化系统的必然发展趋势。

3)密集的计算是限制遥感与作物生长模型数据同化大区域应用精度的主要因素之一,基于大数据环境的同化系统构建为提高同化效率提供了可行方案。此外, GEE平台的出现与发展将进一步满足数据同化系统的并行计算要求,将是未来同化研究的重点方向。

[1] Williams J R. The EPIC crop growth model[J]. Transactions of the ASAE, 1989, 32(2): 497-511.

[2] Diepen C A, Wolf J, Keulen H, et al. WOFOST: A simulation model of crop production[J]. Soil Use & Management, 2010, 5(1): 16-24.

[3] Jones J W, Hoogenboom G, Porter C H, et al. The DSSAT cropping system model[J]. European Journal of Agronomy, 2003, 18(3): 235-265.

[4] Theodorec H, Lee H, Pasquale S, et al. AquaCrop: The FAO crop model to simulate yield response to water[J]. Agronomy Journal, 2009, 101(3): 426-459.

[5] Moulin S, Bondeau A, Delecolle R. Combining agricultural crop models and satellite observations: From field to regional scales[J]. International Journal of Remote Sensing, 1998, 19(6): 1021-1036.

[6] Dorigo W A, Zurita-Milla R, de Wit A J W, et al. A review on reflective remote sensing and data assimilation techniques for enhanced agroecosystem modeling[J]. International Journal of Applied Earth Observation and Geoinformation, 2007, 9(2): 165-193.

[7] 赵艳霞,周秀骥,梁顺林. 遥感信息与作物生长模式的结合方法和应用:研究进展[J]. 自然灾害学报,2005,14(1):103-109. Zhao Yanxia, Zhou Xiuji, Liang Shunlin. Methods and application of coupling remote sensing data and crop growthmodels: Advance in research[J]. Journal of Natural Disaster, 2005, 14(1): 103-109.(in Chinese with English abstract)

[8] 邢雅娟,刘东升,王鹏新. 遥感信息与作物生长模型的耦合应用研究进展[J]. 地球科学进展,2009,24(4):444-451. Xing Yajuan, Liu Dongsheng, Wang Pengxin. Advances of the coupling application of remote sensing information and crop growth model[J]. Advanced Earth Sciences, 2009, 24(4): 444-451. (in Chinese with English abstract)

[9] 吴蕾,柏军华,肖青,等. 作物生长模型与定量遥感参数结合研究进展与展望[J]. 农业工程学报,2017,33(9):155-166. Wu Lei, Bai Junhua, Xiao Qing, et al. Research progress and prospect on combining crop growth models with parameters derived from quantitative remote sensing[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(9): 155-166. (in Chinese with English abstract)

[10] Plummer S E. Perspectives on combining ecological process models and remotely sensed data[J]. Ecological Modelling, 2000, 129(2/3): 169-186.

[11] Charney J G, Halem M, Jastrow R. Use of incomplete historical data to infer the present state of the atmosphere (Incomplete historical data to infer state of atmosphere based on global circulation model, noting tradeoff of temperature for wind and time for space)[J]. Journal of the Atmospheric Sciences, 1969, 26: 1160-1163.

[12] Liang S. Quantitative Remote Sensing of Land Surfaces[M]. New York: John Wiley & Sons, 2004.

[13] Lewis P, Gómez-Dans J, Kaminski T, et al. An earth observation land data assimilation system (EO-LDAS)[J]. Remote Sensing of Environment, 2012, 120(120): 219-235.

[14] Jin X, Kumar L, Li Z, et al. A review of data assimilation of remote sensing and crop models[J]. European Journal of Agronomy, 2018, 92: 141‒152.

[15] Guérif M, Duke C L. Adjustment procedures of a crop model to the site specific characteristics of soil and crop using remote sensing data assimilation[J]. Agriculture Ecosystems & Environment, 2000, 81(1): 57-69.

[16] Dong T, Liu J, Qian B, et al. Estimating winter wheat biomass by assimilating leaf area index derived from fusion of Landsat-8 and MODIS data[J]. International Journal of Applied Earth Observations & Geoinformation, 2016, 49: 63-74.

[17] Jin M, Liu X, Wu L, et al. An improved assimilation method with stress factors incorporated in the WOFOST model for the efficient assessment of heavy metal stress levels in rice[J]. International Journal of Applied Earth Observations & Geoinformation, 2015, 41: 118-129.

[18] Zhou G, Liu X, Zhao S, et al. Estimating FAPAR of rice growth period using radiation transfer model coupled with the WOFOST model for analyzing heavy metal stress[J]. Remote Sensing, 2017, 9(5): 424-438.

[19] Jégo G, Pattey E, Liu J. Using leaf area index, retrieved from optical imagery, in the STICS crop model for predicting yield and biomass of field crops[J]. Field Crops Research, 2012, 131(2): 63-74.

[20] Dente L, Satalino G, Mattia F, et al. Assimilation of leaf area index derived from ASAR and MERIS data into CERES-Wheat model to map wheat yield[J]. Remote Sensing of Environment, 2008, 112(4): 1395-1407.

[21] 王航,朱艳,马孟莉,等. 基于更新和同化策略相结合的遥感信息与水稻生长模型耦合技术的研究[J]. 生态学报,2012,32(14):4505-4515. Wang Hang, Zhu Yan, Ma Mengli, et al. Coupling remotely sensed information with a rice growth model by combining updating and assimilation strategies[J]. Acta Ecologica Sinica, 2012, 32(14): 4505-4515. (in Chinese with English abstract)

[22] Launay M, Guerif M. Assimilating remote sensing data into a crop model to improve predictive performance for spatial applications[J]. Agriculture Ecosystems & Environment, 2005, 111(1/2/3/4): 321-339.

[23] Guérif M, Duke C L. Calibration of the SUCROS emergence and early growth module for sugar beet using optical remote sensing data assimilation[J]. European Journal of Agronomy, 1998, 9(2/3): 127-136.

[24] 陈劲松,黄健熙,林珲,等. 基于遥感信息和作物生长模型同化的水稻估产方法研究[J]. 中国科学:信息科学,2010(S1):173-183. Chen Jinsong, Huang Jianxi, Lin Hui, et al. Rice yield estimation by assimilation remote sensing into crop growth model[J]. Science China Information Sciences, 2010(S1):173-183. (in Chinese with English abstract)

[25] Ma G, Huang J, Wu W, et al. Assimilation of MODIS-LAI into the WOFOST model for forecasting regional winter wheat yield[J]. Mathematical & Computer Modelling, 2013, 58(3/4): 634-643.

[26] Shen S H, Yang S B, Li B B, et al. A scheme for regional rice yield estimation using ENVISAT ASAR data[J]. Science in China, 2009, 52(8): 1183-1194.

[27] Huang J, Tian L, Liang S, et al. Improving winter wheat yield estimation by assimilation of the leaf area index from Landsat TM and MODIS data into the WOFOST model[J]. Agricultural and Forest Meteorology, 2015, 204: 106-121.

[28] 包姗宁,曹春香,黄健熙,等. 同化叶面积指数和蒸散发双变量的冬小麦产量估测方法[J]. 地球信息科学学报,2015,17(7):871-882. Bao Shanning, Cao Chunxiang, Huang Jianxi, et al. Research on winter wheat yield estimation based on assimilation of leaf area index and evapotranspiration data[J]. Journal of Geo-Information Science, 2015, 17(7): 871-882.(in Chinese with English abstract)

[29] Zhou H, Wu J, Li X, et al. Improving soil moisture estimation by assimilating remotely sensed data into crop growth model for agricultural drought monitoring[C]// Geoscience and Remote Sensing Symposium. Piscataway: IEEE, 2016: 4229‒4232.

[30] He B, Li X, Quan X, et al. Estimating the aboveground dry biomass of grass by assimilation of retrieved LAI into a crop growth model[J]. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2015, 8(2): 550-561.

[31] Fang H, Liang S, Hoogenboom G. Integration of MODIS LAI and vegetation index products with the CSM-CERES- Maize model for corn yield estimation[J]. International Journal of Remote Sensing, 2011, 32(4): 1039-1065.

[32] Jin H, Li A, Wang J, et al. Improvement of spatially and temporally continuous crop leaf area index by integration of CERES-Maize model and MODIS data[J]. European Journal of Agronomy, 2016, 78: 1-12.

[33] Dong Y, Wang J, Li C, et al. Comparison and analysis of data assimilation algorithms for predicting the leaf area index of crop canopies[J]. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2013, 6(1): 188-201.

[34] Huang J, Ma H, Su W, et al. Jointly Assimilating MODIS LAI and ET products into the SWAP model for winter wheat yield estimation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, 8(8): 4060-4071.

[35] 朱元励,朱艳,黄彦,等. 应用粒子群算法的遥感信息与水稻生长模型同化技术[J]. 遥感学报,2010,14(6):1226-1240. Zhu Yuanli, Zhu Yan, Huang Yan, et al. Assimilation technique of remote sensing information and rice growth model based on particle swarm optimization[J]. Journal of Remote Sensing, 2010, 14(6): 1226-1240. (in Chinese with English abstract)

[36] Fang H, Liang S, Hoogenboom G, et al. Corn-yield estimation through assimilation of remotely sensed data into the CSM-CERES-Maize model[J]. International Journal of Remote Sensing, 2008, 29(10): 3011-3032.

[37] Liu F, Liu X, Ding C, et al. The dynamic simulation of rice growth parameters under cadmium stress with the assimilation of multi-period spectral indices and crop model[J]. Field Crops Research, 2015, 183: 225-234.

[38] Jin X, Kumar L, Li Z, et al. Estimation of winter wheat biomass and yield by combining the AquaCrop model and field hyperspectral data[J]. Remote Sensing, 2016, 8(12): 1-15.

[39] Jin X, Li Z, Yang G, et al. Winter wheat yield estimation based on multi-source medium resolution optical and radar imaging data and the AquaCrop model using the particle swarm optimization algorithm[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 126: 24-37.

[40] Silvestro P, Pignatti S, Pascucci S, et al. Estimating wheat yield in China at the field and district scale from the assimilation of satellite data into the Aquacrop and simple algorithm for yield (SAFY) models[J]. Remote Sensing, 2017, 9(6): 509-532.

[41] Li Z, Wang J, Xu X, et al. Assimilation of two variables derived from hyperspectral data into the DSSAT-CERES model for grain yield and quality estimation[J]. Remote Sensing, 2015, 7(9): 12400-12418.

[42] Ines A V M, Honda K, Gupta A D, et al. Combining remote sensing-simulation modeling and genetic algorithm optimization to explore water management options in irrigated agriculture[J]. Agricultural Water Management, 2006, 83(3): 221-232.

[43] Vazifedoust M, van Dam J C, Bastiaanssen W, et al. Assimilation of satellite data into agrohydrological models to improve crop yield forecasts[J]. International Journal of Remote Sensing, 2009, 30(10): 2523-2545.

[44] Wu S, J. Huang, X. Liu, et al. Assimilating MODIS-LAI into crop growth model with EnKF to predict regional crop yield[C]//International Conference on Computer and Computing Technologies in Agriculture. Springer, Berlin, Heidelberg, 2011:410‒418.

[45] 黄健熙,武思杰,刘兴权,等. 基于遥感信息与作物模型集合卡尔曼滤波同化的区域冬小麦产量预测[J]. 农业工程学报,2012,28(4):142-148. Huang Jianxi, Wu Sijie, Liu Xingquan, et al. Regional winter wheat yield forecasting based on assimilation of remote sensing data and crop growth model with Ensemble Kalman method[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2012, 28(4): 142-148. (in Chinese with English abstract)

[46] Zhao Y, Chen S, Shen S. Assimilating remote sensing information with crop model using Ensemble Kalman Filter for improving LAI monitoring and yield estimation[J]. Ecological Modelling, 2013, 270(2): 30-42.

[47] Zhu X, Zhao Y, Feng X. A methodology for estimating leaf area index by assimilating remote sensing data into crop model based on temporal and spatial knowledge[J]. Chinese Geographical Science, 2013, 23(5): 550-561.

[48] Liu J, Huang J, Tian L, et al. Regional winter wheat yield prediction by integrating MODIS LAI into the WOFOST model with sequential assimilation technique[J]. Journal of Food, Agriculture & Environment, 2014, 12(1): 180-187.

[49] 陈思宁,赵艳霞,申双和,等. 基于PyWOFOST作物模型的东北玉米估产及精度评估[J]. 中国农业科学,2013,46(14):2880-2893. Chen Sining, Zhao Yanxia, Shen Shuanghe, et al. Study on maize yield estimation and accuracy assessment based on PyWOFOST crop model in northeast China[J]. Scientia Agricultura Sinica, 2013, 46(14): 2880-2893.(in Chinese with English abstract)

[50] Li Y, Zhou Q, Zhou J, et al. Assimilating remote sensing information into a coupled hydrology-crop growth model to estimate regional maize yield in arid regions[J]. Ecological Modelling, 2014, 291: 15-27.

[51] Huang J, Sedano F, Huang Y, et al. Assimilating a synthetic Kalman filter leaf area index series into the WOFOST model to improve regional winter wheat yield estimation[J]. Agricultural and Forest Meteorology, 2016, 216: 188-202.

[52] Ma H, Huang J, Zhu D, et al. Estimating regional winter wheat yield by assimilation of time series of HJ-1 CCD NDVI into WOFOST–ACRM model with Ensemble Kalman Filter[J]. Mathematical and Computer Modelling, 2013, 58(3/4): 759-770.

[53] Cheng Z, Meng J, Qiao Y, et al. Preliminary study of soil available nutrient simulation using a modified WOFOST model and time-series remote sensing observations[J]. Remote Sensing, 2018, 10(1): 64-84.

[54] Wang J, Li X, Lu L, et al. Estimating near future regional corn yields by integrating multi-source observations into a crop growth model[J]. European Journal of Agronomy, 2013, 49(49): 126-140.

[55] Curnel Y, de Wit A J W, Duveiller G, et al. Potential performances of remotely sensed LAI assimilation in WOFOST model based on an OSS Experiment[J]. Agricultural and Forest Meteorology, 2011, 151(12): 1843-1855.

[56] 解毅,王鹏新,刘峻明,等. 基于四维变分和集合卡尔曼滤波同化方法的冬小麦单产估测[J]. 农业工程学报,2015,31(1):187-195. Xie Yi, Wang Pengxin, Liu Junming, et al. Winter wheat yield estimation based on assimilation method combined with 4DVAR and EnKF[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(1): 187-195. (in Chinese with English abstract)

[57] de Wit A J W, van Diepen C A. Crop model data assimilation with the Ensemble Kalman filter for improving regional crop yield forecasts[J]. Agricultural And Forest Meteorology, 2007, 146(1/2): 38-56.

[58] Chakrabarti S, Bongiovanni T E, Judge J, et al. Assimilation of downscaled SMOS soil moisture for quantifying drought impacts on crop yield in agricultural regions in Brazil[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(9): 3867-3879.

[59] Pauwels V, Verhoest N E, De Lannoy G J, et al. Optimization of a coupled hydrology–crop growth model through the assimilation of observed soil moisture and leaf area index values using an ensemble Kalman filter[J]. Water Resources Research, 2007, 43(4): 244-247.

[60] Ines A V M, Das N N, Hansen J W, et al. Assimilation of remotely sensed soil moisture and vegetation with a crop simulation model for maize yield prediction[J]. Remote Sensing of Environment, 2013, 138: 149-164.

[61] Nearing G S, Crow W T, Thorp K R, et al. Assimilating remote sensing observations of leaf area index and soil moisture for wheat yield estimates: An observing system simulation experiment[J]. Water Resources Research, 2012, 48(5): 213-223.

[62] Jiang Z, Chen Z, Chen J, et al. Application of crop model data assimilation with a particle filter for estimating regional winter wheat yields[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(11): 4422-4431.

[63] Liu J, Huang J, Tian L, et al. Particle filter-based assimilation algorithm for improving regional winter wheat yield estimation[J]. Sensor Letters, 2014, 12(3): 763-769.

[64] Machwitz M, Giustarini L, Bossung C, et al. Enhanced biomass prediction by assimilating satellite data into a crop growth model[J]. Environmental Modelling & Software, 2014, 62: 437-453.

[65] 张树誉,孙辉涛,王鹏新,等. 基于同化叶面积指数和条件植被温度指数的冬小麦单产估测[J]. 干旱地区农业研究,2017,35(6):266-271. Zhang Shuyu, Sun Huitao, Wang Pengxin, et al. Winter wheat yield estimation based on the assimilated leaf area index and vegetation temperature condition index[J]. Agricultural Research in the Arid Areas, 2017, 35(6): 266-271. (in Chinese with English abstract)

[66] Gustafsson N. Discussion on ‘4D‐Var or EnKF?’[J]. Tellus Series A-dynamic Meteorology & Oceanography, 2007, 59(5): 774-777.

[67] Huang X, Xiao Q, Barker D M, et al. Four-dimensional variational data assimilation for WRF: Formulation and preliminary results[J]. Monthly Weather Review, 2009, 137(1): 299-314.

[68] Gauthier P, Tanguay M, Laroche S, et al. Extension of 3DVAR to 4DVAR: Implementation of 4DVAR at the meteorological service of canada[J]. Monthly Weather Review, 2007, 135(6): 2339-2354.

[69] Qin J, Liang S, Liu R, et al. A weak-constraint-based data assimilation scheme for estimating surface turbulent fluxes[J]. IEEE Geoscience and Remote Sensing Letters, 2007, 4(4): 649-653.

[70] Kalnay E, Hong L I, Miyoshi T, et al. 4-D-Var or ensemble Kalman filter?[J]. Tellus Series A-dynamic Meteorology & Oceanography, 2007, 59(5): 758-773.

[71] Casa R, Varella H, Buis S, et al. Forcing a wheat crop model with LAI data to access agronomic variables: Evaluation of the impact of model and LAI uncertainties and comparison with an empirical approach[J]. European Journal of Agronomy, 2012, 37(1): 1-10.

[72] Myneni R B, Hoffman S, Knyazikhin Y, et al. Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data[J]. Remote Sensing of Environment, 2002, 83(1/2): 214-231.

[73] Baret F, Weiss M, Lacaze R, et al. GEOV1: LAI and FAPAR essential climate variables and FCOVER global time series capitalizing over existing products. Part1: Principles of development and production[J]. Remote Sensing of Environment, 2013, 137: 299-309.

[74] 刘洋,刘荣高,陈镜明,等. 叶面积指数遥感反演研究进展与展望[J]. 地球信息科学学报,2013,15(5): 734-743. Liu Yang, Liu Ronggao, Chen Jingming, et al. Current status and perspectives of leaf area index retrieval from optical remote sensing data[J]. Journal of Geo-Information Science, 2013, 15(5): 734-743.(in Chinese with English abstract)

[75] 杨飞,孙九林,张柏,等. 基于PROSAIL模型及TM与实测数据的MODIS LAI精度评价[J]. 农业工程学报,2010,26(4):192-197. Yang Fei, Sun Jiulin, Zhang Bai, et al. Assessment of MODIS LAI product accuracy based on the PROSAIL model, TM and field measurements[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(4): 192-197. (in Chinese with English abstract)

[76] 付立哲,屈永华,王锦地. MODIS LAI产品真实性检验与偏差分析[J]. 遥感学报,2017,21(2):206-217.

Fu Lizhe, Qu Yonghua, Wang Jindi. Bias analysis and validation method of the MODIS LAI product[J]. Journal of Remote Sensing, 2017, 21(2): 206-217. (in Chinese with English abstract)

[77] Kim H W. Validation of MODIS 16 global terrestrial evapotranspiration products in various climates and land cover types in Asia[J]. Ksce Journal of Civil Engineering, 2012, 16(2): 229-238.

[78] Velpuri N M, Senay G B, Singh R K, et al. A comprehensive evaluation of two MODIS evapotranspiration products over the conterminous United States: Using point and gridded FLUXNET and water balance ET[J]. Remote Sensing of Environment, 2013, 139: 35-49.

[79] Xue B, Wang L, Li X, et al. Evaluation of evapotranspiration estimates for two river basins on the Tibetan Plateau by a water balance method[J]. Journal of Hydrology, 2013, 492: 290-297.

[80] Mertens J, Madsen H, Feyen L, et al. Including prior information in the estimation of effective soil parameters in unsaturated zone modelling[J]. Journal of Hydrology, 2004, 294(4): 251-269.

[81] Seidel S J, Palosuo T, Thorburn P, et al. Towards improved calibration of crop models: Where are we now and where should we go?[J]. European Journal of Agronomy, 2018, 94: 25-35.

[82] Irmak A, Jones J W, Batchelor W D. Estimating spatially variable soil properties for application of crop models in precision agriculture[J]. Transactions of the ASAE, 2001, 44(5): 1343-1353.

[83] Gauch H G, Hwang J T G, Fick G W. Model evaluation by comparison of model-based predictions and measured values[J]. Agronomy Journal, 2003, 95(6): 1442-1446.

[84] Liu W T H. Application of CERES-Maize Model to yield prediction of a Brazilian maize hybrid[J]. Agricultural & Forest Meteorology, 1989, 45(3): 299-312.

[85] Thorp K R, Batchelor W D, Paz J O, et al. Using cross-validation to evaluate CERES-Maize yield simulationswithin a decision support system for precision agriculture[J]. Transactions of the ASABE, 2007, 50(4): 1467-1479.

[86] Romanowicz R J, Beven K J. Comments on generalised likelihood uncertainty estimation[J]. Reliability Engineering & System Safety, 2006, 91(10): 1315-1321.

[87] Toshichika I, Masayuki Y, Motoki N. Parameter estimation and uncertainty analysis of a large-scale crop model for paddy rice: Application of a Bayesian approach[J]. Agricultural & Forest Meteorology, 2009, 149(2): 333-348.

[88] Marin F, Jones J W, Boote K J. A stochastic method for crop models: Including uncertainty in a sugarcane model[J]. Agronomy Journal, 2017, 109(2) : 483-495.

[89] 何亮,侯英雨,赵刚,等. 基于全局敏感性分析和贝叶斯方法的WOFOST作物模型参数优化[J]. 农业工程学报,2016,32(2):169-179. He Liang, Hou Yingyu, Zhao Gang, et al. Parameters optimization of WOFOST model by integration of global sensitivity analysis and Bayesian calibration method[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2016, 32(2): 169-179. (in Chinese with English abstract)

[90] Baigorria G A, Jones J W, O Brien J J. Potential predictability

of crop yield using an ensemble climate forecast by a regional circulation model[J]. Agricultural and Forest Meteorology, 2008, 148(8/9): 1353-1361.

[91] Hoogenboom G. Contribution of agrometeorology to the simulation of crop production and its applications[J]. Agricultural & Forest Meteorology, 2000, 103(1/2): 137-157.

[92] Lorenz E N, Hilborn R C. The essence of chaos[J]. Physics Today, 1993, 63(48):862-863.

[93] 丑纪范. 大气科学中非线性与复杂性研究的进展[J]. 中国科学院院刊,1997,12(5):325-329. Chou Jifan. Progress in nonlinearity and complexity research in atmospheric sciences[J]. Proceedings of the Chinese Academy of Sciences, 1997,12(5):325-329.(in Chinese with English abstract)

[94] 姜志伟. 区域冬小麦估产的遥感数据同化技术研究[D]. 北京:中国农业科学院,2012. Jiang Zhiwei. Study of the Remote Sensing Data Assimilation Technology for Regional Winter Wheat Yield Estimation[D]. Beijing: Chinese Academy of Agricultural Sciences, 2012. (in Chinese with English abstract)

[95] 李小文. 地球表面时空多变要素的定量遥感项目综述[J].地球科学进展,2006,21(8):771-780. Li Xiaowen. Review of the project of quantitative remote sensing of major factors for spatial-temporal heterogeneity on the land surface[J]. Advances in Earth Science, 2006, 21(8):771-780. (in Chinese with English abstract)

[96] 李小文,曹春香,张颢. 尺度问题研究进展[J]. 遥感学报,2009,13(s1):12-20. Li Xiaowen, Cao Chunxiang, Zhang Hao. Development of research in scale effects[J]. Journal of Remote Sensing, 2009, 13(s1): 12-20. (in Chinese with English abstract)

[97] 李小文,王祎婷. 定量遥感尺度效应刍议[J]. 地理学报,2013,68(9):1163-1169.Li Xiaowen, Wang Yiting. Prospects on future developments of quantitative remote sensing[J]. Acta Geographica Sinica, 2013, 68(9):1163-1169. (in Chinese with English abstract)

[98] 李小文,赵红蕊,张颢,等. 全球变化与地表参数的定量遥感[J]. 地学前缘,2002,9(2):365-370. Li Xiaowen, Zhao Hongrui, Zhang Hao, et al. Global change study and quant itat ive remote sensing for land surface parameters[J]. Earth Science Frontiers, 2002, 9(2): 365-370. (in Chinese with English abstract)

[99] Raffy M, Gregoire C. Semi-empirical models and scaling: A least square method for remote sensing experiments[J]. International Journal of Remote Sensing, 1998, 19(13): 2527-2541.

[100] 李小文,王锦地,Strahler A H. 尺度效应及几何光学模型用于尺度纠正[J]. 中国科学:技术科学,2000,30(s1):12-17. Li Xiaowen, Wang Jindi, Strahler A H. Scale effect and geometric optics model for scale correction[J]. Science in China: Series E, 2000, 30(s1): 12-17. (in Chinese with English abstract)

[101] Li X, Wang J, Strahler A H. Scale effect of Planck’s law over nonisothermal blackbody surface[J]. Science in China, 1999, 42(6): 652-656.

[102] Gao Q, Yu M, Yang X, et al. Scaling simulation models for spatially heterogeneous ecosystems with diffusive transportation[J]. Landscape Ecology, 2001, 16(4): 289-300.

[103] Martínez B, García-Haro F J, Coca C D. Derivation of high-resolution leaf area index maps in support of validation activities: Application to the cropland Barrax site[J]. Agricultural & Forest Meteorology, 2009, 149(1): 130-145.

[104] Miller J R, Chen J M, Rubinstein I G. Evaluating image-based estimates of leaf area index in boreal conifer stands over a range of scales using high-resolution CASI imagery[J]. Remote Sensing of Environment, 2004, 89(2): 200-216.

[105] Garrigues S, Allard D, Baret F, et al. Influence of landscape spatial heterogeneity on the non-linear estimation of leaf area index from moderate spatial resolution remote sensing data[J]. Remote Sensing of Environment, 2006, 105(4): 286-298.

[106] 徐希孺,范闻捷,陶欣. 遥感反演连续植被叶面积指数的空间尺度效应[J]. 中国科学:地球科学,2009,39(1):79-87. Xu Xiru, Fan Wenjie, Tao Xin. Spatial scale effects of remote sensing inversion of continuous vegetation leaf area index[J]. Scientia Sinica Terrae, 2009, 39(1): 79-87. (in Chinese with English abstract)

[107] Raffy M. Change of scale in models of remote sensing: A general method for spatialization of models[J]. Remote Sensing of Environment, 1992, 40(2): 101-112.

[108] Chen J M. Spatial scaling of a remotely sensed surface parameter by contexture-three land-atmospheric modeling experiments[J]. Remote Sensing of Environment, 1999, 69(1): 30-42.

[109] Duveiller G, Baret F, Defourny P. Crop specific green area index retrieval from MODIS data at regional scale by controlling pixel-target adequacy[J]. Remote Sensing of Environment, 2011, 115(10): 2686-2701.

[110] Inoue Y, Kurosu T, Maeno H, et al. Season-long daily measurements of multifrequency (Ka, Ku, X, C, and L) and full-polarization backscatter signatures over paddy rice field and their relationship with biological variables[J]. Remote Sensing of Environment, 2002, 81(2/3): 194-204.

[111] Toan T L, Laur H, Mougin E, et al. Multitemporal and dual-polarization observations of agricultural vegetation covers by X-band SAR images[J]. European Journal of Nutrition, 2017, 56(3): 1339-1346.

[112] Shao Y, Fan X, Liu H, et al. Rice monitoring and production estimation using multitemporal RADARSAT[J]. Remote Sensing of Environment, 2001, 76(3): 310-325.

[113] Molijn R A, Iannini L, Mousivand A, et al. Analyzing C-band SAR polarimetric information for LAI and crop yield estimations[J]. Proceedings of SPIE-The International Society for Optical Engineering, 2014, 9239: 4-15.

[114] 马鸿元. 遥感与作物模型数据同化的冬小麦产量集合预报方法研究[D]. 北京:中国农业大学,2018. Ma Hongyuan. Research on Winter Wheat Yield Ensemble Forecasting Based on Data Assimilation of Remote Sensing and Crop Model[D]. Beijing: China Agricultural University, 2018.(in Chinese with English abstract)

[115] 陈上,窦子荷,蒋腾聪,等. 基于聚类法筛选历史相似气象数据的玉米产量DSSAT-CERES-Maize预测[J]. 农业工程学报,2017,33(19):147-155. Chen Shang, Dou Zihe, Jiang Tengcong, et al. Maize yield forecast with DSSAT-CERES-Maize model driven by historical meteorological data of analogue years by clustering algorithm[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(19): 147-155. (in Chinese with English abstract)

[116] Park Y Y, Buizza R, Leutbecher M. TIGGE: Preliminary results on comparing and combining ensembles[J]. Quarterly Journal of the Royal Meteorological Society, 2008, 134(637):

2029-2050.

[117] Swinbank R, Kyouda M, Buchanan P, et al. The TIGGE project and its achievements[J]. Bulletin of the American Meteorological Society, 2016, 97(1): 49-67.

[118] Eitzinger J, Trnka M, Hösch J, et al. Comparison of CERES, WOFOST and SWAP models in simulating soil water content during growing season under different soil conditions[J]. Ecological Modelling, 2004, 171(3): 223-246.

[119] Keating B A, Carberry P S, Hammer G L, et al. An overview of APSIM, a model designed for farming systems simulation[J]. European Journal of Agronomy, 2003, 18(3): 267-288.

[120] Brisson N, Gary C, Justes E, et al. An overview of the crop model STICS[J]. European Journal of Agronomy, 2011, 18(3): 309-332.

[121] Martre P, Wallach D, Asseng S, et al. Multimodel

ensembles of wheat growth: many models are better than one[J]. Global Change Biology, 2015, 21(2): 911-925.

[122] Ruane A C, Hudson N I, Asseng S, et al. Multi-wheat- model ensemble responses to interannual climate variability[J].Environmental Modelling & Software, 2016, 81: 86-101.

[123] Nickolls J, Dally W J. The GPU Computing Era[J].IEEE Micro, 2010, 30(2): 56-69.

[124] Lee K H, Lee Y J, Choi H, et al. Parallel data processing with MapReduce: A survey[J]. Acm Sigmod Record, 2012, 40(4): 11-20.

[125] Zaharia M, Chowdhury M, Franklin M J, et al. Spark: cluster computing with working sets[C]//Usenix Conference on Hot Topics in Cloud Computing. Berkeley: USENIX Association, 2010: 95-101.

[126] Gorelick N, Hancher M, Dixon M, et al. Google Earth Engine: Planetary-scale geospatial analysis for everyone[J]. Remote Sensing of Environment, 2017, 202: 18-27.

Review on data assimilation of remote sensing and crop growth models

Huang Jianxi1,2, Huang Hai1, Ma Hongyuan1, Zhuo Wen1, Huang Ran1, Gao Xinran1, Liu Junming1,2, Su Wei1,2, Li Li1,2, Zhang Xiaodong1,2, Zhu Dehai1,2

(1.100083,2.100083,)

Data assimilation technology, which can combine the advantages of remote sensing and crop growth models, has great potential in large-scale application of agricultural monitoring and yield forecasting. This review included 5 parts. And the first part introduced the framework of data assimilation system of crop growth model and remote sensing. The data assimilation system contained 3 basic components: dynamic model, observation data and assimilation algorithm. Taking the WOFOST model as an example, a schematic representation of assimilating remotely sensed data into a crop model was shown. The second part summarized the progress of data assimilation of crop growth model and remote sensing. The parameter optimization methods based on cost function and the sequential filtering methods based on estimation theory were two major groups of modern data assimilation strategies. The main difference between the two groups was that each subsequent observation for sequential filtering assimilation would only influence the change nature of the model from the current state; in contrast, parameter optimization methods adjusting the estimation using all of the available observations throughout the assimilation window. In general, MODIS data was the most commonly used remotely sensed data for large regional assimilation research, and data of Landsat TM, ETM+ and OLI were the major remotely sensed data used at regional scale. General models, like WOFOST, CERES, etc. were most widely used in agricultural data assimilation researches. The main object of these researches was food crops such as wheat, corn and rice. LAI (leaf area index) was most commonly used as the assimilation variable linking remote sensing and crop models. In addition, a number of studies found that time series of reflectance, vegetation index or backscattering coefficient could be directly assimilated into a coupled crop growth–radiative-transfer model to avoid the process of regional LAI retrieval. In general, yield estimation and forecast was the most important application. The third part discussed some key aspects affecting the assimilation accuracy, including 5 parts: 1) The pixel size for assimilation, which depended mainly on the specific application. However, heterogeneous, smallholder farming environments presented significant challenges for remotely sensed data assimilation for crop yield forecasting, as field size within these highly fragmented landscapes was often smaller than the pixel size of remote sensing products that were freely available. 2) The uncertainty of remote sensed parameter inversion, which needed to be quantitatively evaluated to ensure the accuracy of data assimilation. 3) The uncertainty of crop growth models, which caused by model structure, model parameters and weather driven data. 4) Data assimilation strategies and linking parameters. Two main data assimilation strategies, parameter optimization and sequential filtering methods, both had pros and cons. Therefore, more effective assimilation algorithms still needed to be developed. 5) The scale effect. Due to the variability of land cover and the complexity of the crop planting pattern in agricultural landscapes, the scale mismatch between the remotely sensed observations and the state variables of crop growth models remained a difficult challenge. The fourth part summarized the research trend of data assimilation for crop growth model and remote sensing. It included 5 directions: 1) from single assimilated parameter to multiple ones. 2) from single remotely sensed data to multiple ones, especially the combination of optical remote sensing and SAR(synthetic aperture radar). 3) from monitoring to forecasting. Based on former researches, an application framework combining data assimilation and numerical prediction was proposed. 4) from single crop growth model to multi-crop model coupling. 5) from single machine system to the high-performance parallel computing system, especially considering the recent advances in Google Earth Engine.

crops; remote sensing; models; crop growth models; data assimilation; agricultural monitoring; yield forecasting

10.11975/j.issn.1002-6819.2018.21.018

TP79; S31

A

1002-6819(2018)-21-0144-13

2018-06-14

2018-08-19

国家自然科学基金资助(41671418,41471342)

黄健煕,副教授,博士,博士生导师,主要从事农业定量遥感等研究。Email:jxhuang@cau.edu.cn

黄健熙,黄 海,马鸿元,卓 文,黄 然,高欣然,刘峻明,苏 伟,李 俐,张晓东,朱德海.遥感与作物生长模型数据同化应用综述[J]. 农业工程学报,2018,34(21):144-156. doi:10.11975/j.issn.1002-6819.2018.21.018 http://www.tcsae.org

Huang Jianxi, Huang Hai, Ma Hongyuan, Zhuo Wen, Huang Ran, Gao Xinran, Liu Junming, Su Wei, Li Li, Zhang Xiaodong, Zhu Dehai. Review on data assimilation of remote sensing and crop growth models[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(21): 144-156. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2018.21.018 http://www.tcsae.org

猜你喜欢
尺度作物观测
作物遭受霜冻该如何补救
四种作物 北方种植有前景
财产的五大尺度和五重应对
内生微生物和其在作物管理中的潜在应用
作物遭受药害的补救措施
天文动手做——观测活动(21) 软件模拟观测星空
2018年18个值得观测的营销趋势
宇宙的尺度
可观测宇宙
高分辨率对地观测系统