【作者】 蔡福；

【导师】 周广胜；

【作者基本信息】 南京信息工程大学 ， 气象学， 2012， 博士

【摘要】 地表反照率(α)直接控制太阳辐射能在地表和大气之间的分配,是计算陆面与大气系统中能量和物质交换中重要的热力参数；空气动力学粗糙度(z0)和零平面位移(d)是影响陆一气通量交换的主要动力参数。地表热力与动力参数的准确表达直接影响陆—气通量的估算,进而影响各气象要素的模拟。现有陆面模型中的α、z0和d在同类下垫面大多使用固定值,与实际情况存在较大差距,从而导致陆面过程模拟的不确定性。玉米农田因其冠层高度(h)、叶面积指数(LAI)和植被覆盖度(FVEG)等下垫面结构和性质随生育期变化较大,使得α、z0和d等参数在一年中不断发生变化,导致辐射、水分、热量的分配和传输等一系列物理过程随之变化。本研究利用锦州玉米农田生态系统野外观测站2006-2008年的观测资料,分析了整个玉米生育期的α、z0,和d动态变化特征及其与相关影响因子的关系,建立了各参数的动态参数化方案,对BATSle模型原有静态赋值方案进行了改进,定量评价动力与热力参数变化对陆面过程模拟的影响。主要结论如下：(1)研究分析了BATSle模型对动态LAI、FVEG、Ze和α的敏感性。模型基本可以模拟出表层土壤温度(Tg)、净入射短波辐射(frs)及感热(Hs)的日动态及日际变化,但对表层土壤湿度(SWC)和潜热(λE)尤其在非降水日模拟能力较差。当LAI、FVEG较小时,各变量模拟误差较大,随着模型设定值与实际值逐渐接近,模拟误差不断减小,说明真实的参数设置对提高模拟精度非常必要。z0动态变化对Tg和Hs模拟有一定改进作用,动态LAI对改善Tg、Hs、A.E和SWC的模拟效果作用显著,FVEG将显著改善对Tg、frs、Hs和SWC的模拟。农田从裸土向有植被覆盖转变时各变量对参数动态敏感性更大。反照率的动态赋值可不同程度影响Tg、Hs和λE的模拟,其中Hs对其变化敏感性最大。(2)建立了基于太阳高度角(hθ)、SWC和LAI三因子的玉米农田α的动态参数化方案。非生长季,hθ和SWC与α分别呈对数和线性关系所建立的模型能较好地模拟出α的日动态特征,但在初春土壤刚化通时,对α模拟能力较差。生长季考虑h。、SWC和LAI与a为对数、线性和指数关系所建立的模型模拟精度较高,但在营养生长阶段很差,总体效果不理想。引入FVEG对裸土和植被分别赋权重所建立的α综合模型可反映α的季节动态变化,在整个生长季尤其营养生长阶段模拟精度改善明显。通过与双层模型和简化双层模型比较发现,综合模型除玉米生育后期模拟能力略小于简化双层模型外,其他大部分时段模拟能力都较强,对实现玉米农田α动态参数化更为理想。(3)利用α综合模型并引入动态LAI改进了BATSle模型,定量评价了模型改进对陆-气通量模拟的影响。改进模型实现了α的动态模拟,全年模拟误差明显减小,生长季内由于动态LAI的引入使α模拟结果更为真实；frs模拟精度改进明显,改进年总量占年总辐射的1.7%；净长波吸收辐射(frl)在5、6月FVEG快速变化时模拟精度提高显著；净辐射(nr)在生长季和全年白天模拟精度有所提高。Tg年平均改进量为0.62K,多数月份月平均改进量在1K以上。模型改进使热通量模拟过程更接近真实情况,而原模型模拟结果表面上看与实测值趋势较一致,但实际上是“虚假正确”。热分量中Hs模拟精度改善最明显,生长季好于非生长季,下垫面性质变化明显的6月和10月改善显著。λE改善程度小于Hs,生长季略好于非生长季,但误差仍较大,主要是模型对非降水日SWC明显低估所致,表明BATSle模型中对土壤水分的模拟需要改进。土壤热通量(G)模拟值对实测值的解释能力提高4%,模拟精度在非生长季高于生长季。(4)建立了基于最优方法估算的动力参数与影响因子关系的动态参数化方案。利用不同高度组合求得理查逊数(Ri)所计算的z0差异明显,z。随Ri增大而减小,2m和10m两层高度求取Ri更为合理。通过比较摩擦风速(u*)实测与模拟值的相关性选择最优方法计算z0和d。z0值在抽雄前小于0.2m,乳熟前后达最大,约为0.4m,d值在拔节后10天左右、h约为1.4m时开始出现,为0.8-1m,抽雄后为1-1.4m。从量级和变化趋势上看,本研究结果与相关研究基本一致。在d值出现前,z0与风速、LAI和h分别呈负指数和线性正相关关系,h和风速对z。作用的累加形式所建立的z0参数化模型模拟精度最高。d值出现后,风速与z0、d之和的关系明显大于与它们各自的关系,z0、d与LAI和h都呈指数正相关关系,LAI和h对z的影响大于d和z0+d,h对z0、d和z0+d影响大于LAI,d/h和z0/h分别为0.4-0.54和0.1-0.14；在h达到最大值前,d/h和z0/h分别随LAI减小和增大,LAI和风速与z0都呈指数关系建立的z0参数化模型模拟精度更高,h和风速与d都呈指数关系的d参数化模型模拟精度更高；当h不变时,LAI变化幅度较小且风速测量误差的增大使d参数化模拟模型精度降低,因z0日际变动很小使所建立的模型无法对其模拟。(5)评估了动力参数动态参数化方案对BATSle模型的改进作用。G改进最为明显,其次是Hs,而XE改进相对较少,生长季总改进量分别占总辐射的1.24%,0.36%和0.19%,其中G在7、8月改进最明显,改进量分别占月总辐射的2.35%和3.36%；Hs改进最明显的8月改进量占月总辐射1.68%；λE在7、8月改进最明显,改进量分别占月总辐射的0.67%和2.29%。当h小于1.6m时,模型对d的敏感性较弱,d可以忽略。(6)评估了动力与热力参数动态参数化方案对BATSle模型的改进作用。利用所建立的α、z0和d动态参数化方案同时对模型改进后,α、nr和Tg模拟精度有所提高；热通量各分量中Hs改进幅度最大,其次是G,λE改进最小。nr和Hs因α动态参数化而改进的幅度大于z0和d动态参数化,Tg、λE和G模拟精度则因z0和d改进而提高明显。在α、z0和d得到改善的情况下,一些时段λE模拟精度却有所下降,主要是原模型对SWC模拟不合理所引起,这一情况表明,陆面过程模拟中,一个或几个参量的改进可能会使一些变量的模拟得到改善,但因陆面过程的复杂性某些要素的模拟精度反而下降,一个很重要原因就是原模型对某些要素模拟过程中出现误差相抵的问题,使模拟结果出现“虚假正确”,而某些参量的改进恰恰打破了原有的“平衡”,导致所谓的误差增大,而实际上模型是向着更为合理的过程改进,改进后模型模拟误差的增大并不完全代表改进过程的无效。因此,陆面过程模型的一些过程仍不完善,需要不断加以改进。更多还原

【Abstract】 Surface albedo (a) controls directly distribution of solar radiation energy between the earth surface and atmosphere, and it is a very important thermal parameter used to calculate exchanges of energy and materials between terrestrial ecosystems and atmosphere. Aerodynamics roughness (z0) and zero plane displacement (d) are significant dynamic parameters influencing flux exchanges bwteeen terrestrial ecosystems and atmosphere. Accurate description of these parameters can improve simulation accuracy of the exchanges of energy and materials between terrestrial ecosystems and atmosphere as well as meteorological elements. Usually, a,zo and d are expressed with fixed value during the same kind of underlying surface in existing land surface models, and do not consider the change with time. Rainfed maize agroecosystem is a typical and representative underlying surface type in northeast China because its extreme changes in surface construction and properties including canopy height (h), leaf area index (LAI) and vegetation coverage (FVEG) with growth of maize cause variations of a, zo and d and then lead to changes in a series of physical process such as distributing and transferring processes of radiation, water and heat. Based on continuous observation data of land-atmosphere flux exchanges, meteorological and biological elements during2006-2008from Jinzhou agricultural ecosystem research station, dynamic characteristics and relationships with relevant influence factors of a,zo and d in rainfed maize whole growth period are analyzed and their dynamic parameterization schemes are set up and used to improve BATSle. At the last, the effect of improved model on simulating land surface process is investigated. The main conclusions are listed as follows.(1) Revealing the sensitivity of BATSle to dynamic LAI, FVEG, a and z0. BATSle is able to simulate reasonable daily pattern and interdiurnal change of surface soil temperature(Tg), net absorbed solar energy flux(Frs) and sensible heat flux(Hs) as well as undesirable surface soil water content(SWC) and latent heat flux(λE) especially on no-precipitation day. The simulation errors is greater and decreasing when LAI and FVEG are smaller and approaching ground truth, indicating that it is very necessary for increase simulation precision to use more real parameter settings in the model. Dynamic assignment of zo, LAI and FVEG plays an important role in improving simulation precision respectively to Tg, Hs and Tg, Hs, λE, SWC and Tg, frs, Hs, SWC. On the whole, every variable is sensitive to parameter dynamic when rainfed maize ecosystem surface change from bare soil to vegetation. In addition, a assigned with dynamic value affects to varying degrees to simulation of Tg, λE and Hs, especially the latter.(2) Developing a dynamic parameterization scheme of a based on solar altitude(ho), SWC and LAI. The bare soil a scheme founded considering respectively logarithm and linear relationship between a and he and SWC is better than those considering other relationships and is able to simulate diurnal pattern of a with smaller error in most of the non-growing season except early spring. In the growing season, the simulation precision of a scheme founded with statistical regression method considering respectively logarithm, linear and exponential relationship between a and he, SWC and LAI those play an important role to a is higher than those considering other relationships. For the limitation of data, the scheme underestimates evidently a in most of the study periods especially in vegetative growth phase of maize. As FVEG is introduced and used to bestow weighing to soil and vegetation, the synthesis model whose simulation error decreases significantly in whole growing season especially in vegetative growth phase is able to reflect seasonal variation of a and has dynamic simulation ability, which change an untrue hypothesis that vegetation a is only fixed value in many land surface model and makes the model universal-adapted to simulate dynamic a in different phases of rainfed maize ecosystem. Compared with the double-layer and simplified double-layer model of a, simulation ability of the synthesis model is stronger than that of another two in most of time expect in later growing period when is weaker than that of simplified double-layer model.(3) Evaluating the simulation of improving BATSle model through introducing the synthesis model and dynamic LAI. The results show that the improving model realizes dynamic simulation of a whose annual simulation error decreases obviously and the simulation value is more accurate for dynamic LAI introduced in growing season, which improves the simulation precision of radial component such as frs whose annual improving quantity(IQ) accounts for1.7percent of annual global radiation, net absorb long wave radiation(frl) in May and June when FVEG changes quickly and nr in growing season and daytime,respectively. IQ of yearly and monthly Tg is0.62K and above1K, respectively. Concidering difference of canopy heat flux between bare soil and vegetation, their simulation results analyzed respectively show that simulation process of heat flux by the improve model is more close to the fact including Hs whose improvement is the most obvious, especially in the growing season in June and August when underlying surface characteristics change evidently than in non-growing season, and λE whose improvement is less than the former and circumstance is consistent mainly with Hs, but simulation error is large because of notable underestimation of SWC on no precipitation day, which demonstrates that expression of soil water content of BATSle must be improved, as well as soil heat flux(G) whose simulation precision is higher in the non-growing season than in the growing season and the explaining ability of simulation to observation increases4%. Yet simulation process of the primary model is consistent with the fact in trend seen from the outside but is an indisguise in facts.(4) Establishing dynamic parameterization schemes of zo and d by the optimization methods. It is found that zo from bulk Rickardson number (Ri) considering different heights combination is discrepant, decrease with increasing Ri and is reasonable simulated with the height combination of2m and10m. z0is smaller than0.2m before tasseling stage and comes to the maximum about0.4m before and after milk stage. d value begins to appear10days after jointing when h is1.4m height and at this time its height is0.8to1m then1to1.4m after tasseling stage. For magnitude and change trend, zo and d in this study are consistent with related research results. Before d appears, negative exponent and positive linear relationships between zo and wind speed, LAI, h are found. Simulation precision of zo parameterization scheme considering accumulation form of the effect of h and wind speed on zo is highest. After d appears, relationship between wind speed and zo+d is more notable than those between wind speed and zo or d. At the same time, positive exponent relationships between zo or d and LAI or h are found. LAI and h have more influence to zo than d and zo+d for the former greater than the latter. In addition, d/h and zo/h is0.4to0.54and0.1to0.14respectively. Before h comes to the maximum, d/h and zo/h are decreasing and increasing with LAI respectively. Simulation precisions of z0and d parameterization model considering multiplicative form of exponent relationships between zo and LAI, wind speed as well as d and h, wind speed respectively are highest, respectively. When h is invariable, simulation precisions of d and zo parameterization model is decreasing and can’t be simulated respectively as a result of small variation of LAI and greater measure error of wind speed for the former and smaller diurnal variation for the latter.(5) Evaluating the simulation of improving BATSle model through introducing dynamic aerodynamic parameter scheme. The simulation precision of each component of land surface heat flux is in different degree improved with the order of G, Hs and λE whose growing season IQ account for1.24,0.36and0.19percent of global radiation respectively when original BATSle model is modified with newly-built dynamic zo and d parameterization scheme. IQ of G, Hs and λE are larger in July and August, in August and September, in July and August accounting for2.35and3.36,1.68and0.4as well as0.67and2.29percent of monthly global radiation respectively than in other months. Furthermore, we come to a conclusion that d is able to be ignored when that is smaller than1.6m because of slow response from BATSle model.(6) Evaluating the simulation of improving BATSle model through introducing dynamic aerodynamic and thermodynamic parameter scheme. Simulation precision of a, nr and Tg are improved when original BATSle model is simultaneously modified with newly-built dynamic a, zo and d parameterization scheme. As a result, simulation of each component of land surface heat flux is improved with the order of Hs, G, and λE. Considering contribution of each parameter, a dynamic parameterization contributes more than that of z0and d to nr and Hs, on the contrary, z0and d dynamic parameterization contribute more than that of a to simulation of Tg, λE and G. Simulation precision of λE decreases in some periods though that parameterization of a, zo and d are improved, which is owed to SWC unreasonably simulated by original BATSle model. Those show that simulation of some variables may be improved but those of another deteriorative when one or several parameters are mended because the fictitious balance situation canceling out simulation errors between different variables and making simulation result seem to be true is destroyed by improving of some variables, which leads error to amplify. But, the model is being more reasonably improved. As a result, enlargement of simulation error does n’t absolutely represent that the improvement of model is invalid. Altogether, many processes in land surface model are still not perfect and in need to be improved更多还原