Testing parameter sensitivities and uncertainty analysis of Biome-BGC model in simulating carbon and water fluxes in broadleaved-Korean pine forests

生态过程模型的发展为研究者在长时间序列和区域尺度的研究提供了便利, 但模型模拟的准确性受到模型自身结构、模型参数估计合理性的影响。敏感性分析能够定量或定性筛选出对模型模拟结果影响较大的敏感参数, 是模型参数校准过程中的重要工具, 也是建模和应用的先决条件。该文以阔叶红松林为研究对象, 采用全局敏感性分析方法——傅里叶幅度灵敏度检验扩展法(EFAST)对Biome-BGC模型的生理生态参数进行了敏感性分析, 分别分析了红松(Pinus koraiensis)和阔叶树的净初级生产力(NPP)、蒸散(ET)对参数变化的敏感性。结果表明: (1)模拟红松NPP的不确定性高于阔叶树, 但二者的模拟ET的不确定性均较小。阔叶树的NPP和ET对生理生态参数的敏感性总体上都小于红松。(2)无论是红松、阔叶或其他植被类型, 模拟NPP均表现出对叶片碳氮比、细根碳氮比、比叶面积(SLA)和冠层截留系数的敏感性, 这4个参数的高敏感性主要是由模型自身结构所决定的, 与植被类型和研究地区的关系较小。对模拟ET而言, 细根与叶片碳分配比、新茎与新叶碳分配比和SLA均是影响红松和阔叶树ET的敏感参数, 但红松ET主要受参数与参数间的二阶或多阶交互作用的间接影响, 而阔叶树ET则主要是受到敏感参数直接效应的影响。(3)除了上述影响红松和阔叶树碳水通量的共性参数外, 诸如核酮糖-1,5-二磷酸羧化酶中叶氮含量、叶片与细根周转率、所有叶面积与投影叶面积之比等也是对模拟结果有影响的重要参数, 但是其敏感程度随物种不同和研究区不同而不同, 所以这类参数可以根据具体情况进行参数本地化, 对于其他不敏感参数则可以采用模型缺省值。     

阔叶红松林净初级生产力模型参数的敏感性

PnET-Ⅱ(photosynthesis and evapotranspiration)模型是生态系统过程模型,运行过程中所需的参数较多, 包括植被、土壤和气候参数等.本文估计了丰林自然保护区阔叶红松林中红松和阔叶树的总净初级生产力（NPP）和枝干NPP对PnET-Ⅱ模型参数变化的敏感程度.结果表明: PnET-Ⅱ模型的植被参数中,林冠参数变化对模拟结果影响较大,且红松总NPP对植被参数的敏感性大于阔叶树;红松和阔叶树NPP对土壤持水量变化敏感性较小,且红松NPP对土壤持水量的敏感性略小于阔叶树;在气候情景范围内,气温变化对红松和阔叶树NPP的影响最大,降水和光合有效辐射次之.不同气候情景对NPP模拟结果的影响不同.红松和阔叶树的总NPP和枝干NPP对各输入参数的敏感程度并不完全一致.     

基于BIOME-BGC模型的长白落叶松林净初级生产力模拟参数敏感性

基于植被生理生态过程的模型包含较多参数,合理的参数取值能够极大地提高模型的模拟能力.参数敏感性分析可以全面分析模型参数对模拟结果的影响程度,在筛选模型敏感参数过程中起到重要作用.本研究以模拟吉林省汪清林业局长白落叶松林净初级生产力（NPP）为例,分析了BIOME-BGC模型的参数敏感性.首先利用样地实测NPP数据与模拟值进行对比分析,检验模型对长白落叶松林NPP的模拟能力;然后利用Morris法和EFAST法筛选出BIOME-BGC模型中对长白落叶松林NPP影响较大的敏感参数.在此基础上,通过EFAST法对所有筛选出的参数进行定量的敏感性分析,计算了敏感参数的全局敏感性指数、一阶敏感性指数和二阶敏感性指数.结果表明： BIOME-BGC模型能够较好地模拟研究区内长白落叶松林NPP的变化趋势;Morris法可以在样本量较少的情况下实现对BIOME-BGC模型敏感参数的筛选,而EFAST法可以定量分析BIOME-BGC模型中单个参数以及不同参数之间交互作用对模拟结果的影响程度;BIOME-BGC模型中对长白落叶松林NPP影响较大的敏感参数为新生茎与叶片的碳分配比和叶片碳氮比,且二者之间的交互作用明显大于其他参数之间的交互作用.     

择伐对阔叶红松林碳密度和净初级生产力的影响

准确量化森林碳密度和净初级生产力(NPP)对于评价森林生态系统在全球碳循环中的作用至关重要.本研究以小兴安岭原始阔叶红松林和择伐(择伐强度30%,择伐对象为大径级红松)34年后的阔叶红松林为对象,采用样地清查和异速生长方程法测定了不同林分的碳密度和NPP.结果表明: 原始林和择伐林的碳密度总量分别为(397.95±93.82)和(355.61±59.37) t C·hm^-2,其中植被碳密度、碎屑碳密度、土壤碳密度分别占总碳库的31.0%、3.1%、65.9%和31.7%、2.9%、65.4%,两者的总碳密度和各组分的分配比例均无显著差异. 原始林和择伐林的NPP总量分别为(6.27±0.36 )和(6.35±0.70) t C·hm^-2·a^-1,乔木层、灌木和草本层、细根所占比例分别为60.3%、2.0%、37.7%和66.1%、2.0%、31.2%,两者的总NPP和各组分的贡献率均无显著差异.而原始林和择伐林中针、阔叶的NPP比例分别为4724∶52.76和20.48∶79.52,两者差异显著.择伐34年后阔叶红松林的碳密度和NPP均达到了择伐前水平.     

IBIS模拟东北东部森林NPP主要影响因子的敏感性

IBIS (Integrated Biosphere Simulator) 模型是生态系统过程模型,运行过程需要大量输入变量,包括植被、土壤和气候等。本文基于2004 — 2005年帽儿山不同植被类型净初级生产力（NPP）实测数据,修改了模型内部参数,使其更适合于我国东北东部地区NPP模拟。本文还展开了NPP对模型主要输入变量变化敏感程度的研究。结果表明：IBIS模型的输入变量中,叶面积指数（LAI）对模拟结果影响较大,且NPP对LAI 增加/降低5%敏感性最强;在气候情景范围内,仅考虑单一气象因子作用,NPP对温度变化敏感性最小;对降水量反应敏感,并随降水量增多而升高。     

改进Biome-BGC模型模拟哈佛森林地区水、碳通量

Biome-BGC模型通过耦合植被、土壤与大气间的水分与CO2交换过程,实现植被生产力的模拟,但土壤水平衡模块的不够完善,导致在长时间无降水情况下植被生产力模拟存在较大误差.针对这一问题,本文对Biome-BGC模型中土壤水分胁迫气孔导度方程、蒸散计算公式及土壤水分流失过程等3方面进行了改进和调整,利用改进的Biome-BGC模型模拟美国哈佛森林地区蒸散、植被生产力,并与地面通量观测值进行了比较.结果表明,改进后模拟精度有明显的提高,蒸散、植被生态系统生产力(NEE)与观测值间的决定系数分别由0.483和0.658提高到0.617和O.813,蒸散逐年均方根误差平均下降了48.7％,NEE逐年误差平方和平均下降了39.8％.改进后的模型模拟结果更接近观测值.     

基于Biome-BGC模型的西双版纳橡胶林碳收支模拟

以西双版纳橡胶适宜种植区(海拔550—600m)的橡胶林(Hevea brasiliensis)为研究对象,应用参数同化后的Biome-BGC模型模拟了1959—2012年橡胶林的碳循环。结果表明,(1)与涡度相关监测结果相比,橡胶林年总初级生产力(Gross Primary Productivity,GPP)、年总呼吸(Total Respiration,Rt)的模拟精度分别为98.37%和90%。由于对年GPP的过低估计和对年Rt的过高估计,年净生态系统交换量(Net Ecosystem Exchange,NEE)的模拟值比实测值低157.35 g C m~(-2)a~(-1)。但若考虑干胶碳(139g C m~(-2)a~(-1)),模拟值与实测值十分接近;(2)橡胶林在模拟进行的前8年里因异养呼吸较高,以碳排放为主,NEE平均约357 g C m~(-2)a~(-1);之后转为以碳固定为主,NEE平均约~(-1)46 g C m~(-2)a~(-1);(3)橡胶林在40年的更新周期中可固定碳1835 g C m~(-2),是一个弱的碳汇。但与热带雨林相同周期固碳6720 g C m~(-2)相比,仍为碳源。以上结果为深入了解橡胶种植对区域碳循环的影响提供了科学依据,建议当地政府一方面要有计划的对老胶林进行更新,以维持当前橡胶林生态系统中的碳平衡;另一方面要注重对热带雨林的保护,从而实现区域经济和生态环境保护的协调发展。     

基于Biome-BGC模型的北方杨树人工林碳水通量对气候变化的响应研究

研究中国北方杨树人工林碳水通量对气候变化的响应,对于制定合理的经营管理措施以应对区域的气候变化具有重要意义。基于对杨树人工林碳水通量的连续监测数据和对Biome-BGC模型参数的校准,模拟分析杨树人工林碳水通量及水分利用效率(WUE)对气候变化(气温上升、降水变化和大气CO_2浓度上升)的响应规律。结果表明,Biome-BGC模型校准后显著提升了其对杨树人工林碳水通量的模拟精度,对GPP、ET模拟结果的Nash-Sutcliffe效率系数(NS)分别为0.69和0.63,各自提高了64.3%和80%,均方根误差(RMSE)则分别降低至1.94 g C m~(-2) d~(-1)和0.88 mm/d,分别下降了26.5%和25.4%。在未来气候变化情景中,单独的气温上升、降水增加和大气CO_2浓度上升分别造成GPP的降低、升高和升高,其中GPP对大气CO_2浓度上升的响应程度(28%—44%)远高于对气温上升(1%—5%)和降水变化(3%—10%)的,ET则主要受降水的影响,响应程度在5%—14%之间。GPP和ET对气候变化的响应则受不同水平的气温上升、降水变化和大气CO_2浓度上升三者综合作用的影响。基于GPP和ET对气候变化的响应,WUE随气温上升、降水增加表现为降低趋势,随降水减少和大气CO_2浓度升高则呈升高趋势;其对未来气候中大气CO_2浓度升高的响应程度为27.7%—43.6%,远高于对气温上升(1.2%—5.8%)和降水变化(1.2%—3.5%)的,说明未来气候变化中大气CO_2浓度上升是促进杨树生长的主要因素;其中相对于当前WUE(2.8 g C/kg H_2O),C2T2P1和C0T3P0情景下WUE的升高和降低幅度最大,分别为45.4%和5.8%。     

Postfire carbon pools and fluxes in semiarid ponderosa pine in Central Oregon

Forest fire dramatically affects the carbon storage and underlying mechanisms that control the carbon balance of recovering ecosystems. In western North America where fire extent has increased in recent years, we measured carbon pools and fluxes in moderately and severely burned forest stands 2 years after a fire to determine the controls on net ecosystem productivity (NEP) and make comparisons with unburned stands in the same region. Total ecosystem carbon in soil and live and dead pools in the burned stands was on average 66% that of unburned stands (11.0 and 16.5 kg C m⁻², respectively, P<0.01). Soil carbon accounted for 56% and 43% of the carbon pools in burned and unburned stands. NEP was significantly lower in severely burned compared with unburned stands (P<0.01) with an increasing trend from −125±44 g C m⁻² yr⁻¹ (±1 SD) in severely burned stands (stand replacing fire), to −38±96 and +50±47 g C m⁻² yr⁻¹ in moderately burned and unburned stands, respectively. Fire of moderate severity killed 82% of trees <20 cm in diameter (diameter at 1.3 m height, DBH); however, this size class only contributed 22% of prefire estimates of bole wood production. Larger trees (> 20 cm DBH) suffered only 34% mortality under moderate severity fire and contributed to 91% of postfire bole wood production. Growth rates of trees that survived the fire were comparable with their prefire rates. Net primary production NPP (g C m⁻² yr⁻¹, ±1 SD) of severely burned stands was 47% of unburned stands (167±76, 346±148, respectively, P<0.05), with forb and grass aboveground NPP accounting for 74% and 4% of total aboveground NPP, respectively. Based on continuous seasonal measurements of soil respiration in a severely burned stand, in areas kept free of ground vegetation, soil heterotrophic respiration accounted for 56% of total soil CO₂ efflux, comparable with the values of 54% and 49% previously reported for two of the unburned forest stands. Estimates of total ecosystem heterotrophic respiration (R_h) were not significantly different between stand types 2 years after fire. The ratio NPP/R_h averaged 0.55, 0.85 and 1.21 in the severely burned, moderately burned and unburned stands, respectively. Annual soil CO₂ efflux was linearly related to aboveground net primary productivity (ANPP) with an increase in soil CO₂ efflux of 1.48 g C yr⁻¹ for every 1 g increase in ANPP (P<0.01, r²= 0.76). There was no significant difference in this relationship between the recently burned and unburned stands. Contrary to expectations that the magnitude of NEP 2 years postfire would be principally driven by the sudden increase in detrital pools and increased rates of R_h, the data suggest NPP was more important in determining postfire NEP.     

基于数据同化的哈佛森林地区水、碳通量模拟

模型模拟和站点观测是陆地生态系统水、碳循环研究最主要的两种手段,但各有优势和不足,若二者相互结合,则能更准确地反映生态系统水、碳通量的动态变化.数据同化为模型与观测结合提供了一条有效的途径.本文采用哈佛森林环境监测站相关数据,利用集合卡曼滤波同化算法,将实测叶面积指数(LAI)和遥感LAI同化进入Biome BGC模型中,对该地区水、碳通量进行模拟.结果表明：与未同化模拟相比,将1998、1999和2006年实测LAI数据同化后,模型模拟碳通量(NEE)与通量观测NEE的决定系数(R²)平均提升8.4%;蒸散发(ET)的R²平均提升10.6%;NEE的绝对误差和(SAE)和均方根误差(RMSE)平均下降17.7%和21.2%,ET的SAE和RMSE平均下降26.8%和28.3%.将2000-2004年MODIS LAI 产品与模型同化后,NEE、ET模拟值与观测值间的R²分别提升7.8%和4.7%;NEE的SAE和 RMSE分别下降21.9%和26.3%,ET的SAE和 RMSE分别下降24.5%和25.5%.无论实测LAI还是遥感观测LAI,同化进入模型都能不同程度地提高水碳通量的模拟精度.     

基于参数优化的人工灌丛生态系统碳水通量模拟

荒漠草原是陆地生态系统中最为脆弱且受人类干扰较为严重的生态类型之一,精准模拟其碳水通量及对人为干扰的响应,不仅能够揭示其复杂的生态学过程,而且还可为人为生态修复和保护提供决策依据。生态模型能够有效地模拟陆地生态系统的碳水循环过程,但模型众多的参数及其取值的合理性限制了其普遍应用,故探索参数优化是提升生态模型应用的有效途径。利用PEST参数优化方法和涡度相关观测数据对Biome-BGC模型的生理生态参数进行优化,在评估参数优化效果的基础上模拟了1986-2018年宁夏盐池荒漠草原区人工灌丛生态系统的总初级生产力（Gross primary productivity,GPP）和蒸散（Evapotranspiration,ET）。结果表明：（1）参数优化可以改善Biome-BGC模型对荒漠草原区人工灌丛生态系统GPP和ET的模拟效果,参数优化后模拟的GPP和ET均更接近于观测值,其中月尺度的模拟效果更佳;（2）基于PEST的Biome-BGC模型参数优化方法具有较强的普适性,优化后的参数可推广应用于荒漠草原区人工灌丛生态系统长时间序列的GPP和ET模拟;（3）宁夏盐池荒漠草原区人工灌丛生态系统的GPP在1986-2018年呈缓慢上升趋势,增幅为1.47 g C m^-2 a^-1,但ET的年际变化率较大,且无显著变化趋势。     

Seasonal carbon dynamics and water fluxes in an Amazon rainforest

Satellite‐based observations indicate that seasonal patterns in canopy greenness and productivity in the Amazon are negatively correlated with precipitation, with increased greenness occurring during the dry months. Flux tower measurements indicate that the canopy greening that occurs during the dry season is associated with increases in net ecosystem productivity (NEP) and evapotranspiration (ET). Land surface and terrestrial biosphere model simulations for the region have predicted the opposite of these observed patterns, with significant declines in greenness, NEP, and ET during the dry season. In this study, we address this issue mainly by developing an empirically constrained, light‐controlled phenology submodel within the Ecosystem Demography model version 2 (ED2). The constrained ED2 model with a suite of field observations shows markedly improved predictions of seasonal ecosystem dynamics, more accurately capturing the observed patterns of seasonality in water, carbon, and litter fluxes seen at the Tapajos National Forest, Brazil (2.86°S, 54.96°W). Long‐term simulations indicate that this light‐controlled phenology increases the resilience of Amazon forest NEP to interannual variability in climate forcing.     

Changes in carbon storage and fluxes in a chronosequence of ponderosa pine

Forest development following stand‐replacing disturbance influences a variety of ecosystem processes including carbon exchange with the atmosphere. On a series of ponderosa pine (Pinius ponderosa var. Laws.) stands ranging from 9 to> 300 years in central Oregon, USA, we used biological measurements to estimate carbon storage in vegetation and soil pools, net primary productivity (NPP) and net ecosystem productivity (NEP) to examine variation with stand age. Measurements were made on plots representing four age classes with three replications: initiation (I, 9–23 years), young (Y, 56–89 years), mature (M, 95–106 years), and old (O, 190–316 years) stands typical of the forest type in the region. Net ecosystem productivity was lowest in the I stands (?124 g C m^?2 yr^?1), moderate in Y stands (118 g C m^?2 yr^?1), highest in M stands (170 g C m^?2 yr^?1), and low in the O stands (35 g C m^?2 yr^?1). Net primary productivity followed similar trends, but did not decline as much in the O stands. The ratio of fine root to foliage carbon was highest in the I stands, which is likely necessary for establishment in the semiarid environment, where forests are subject to drought during the growing season (300–800 mm precipitation per year). Carbon storage in live mass was the highest in the O stands (mean 17.6 kg C m^?2). Total ecosystem carbon storage and the fraction of ecosystem carbon in aboveground wood mass increased rapidly until 150–200 years, and did not decline in older stands. Forest inventory data on 950 ponderosa pine plots in Oregon show that the greatest proportion of plots exist in stands ～ 100 years old, indicating that a majority of stands are approaching maximum carbon storage and net carbon uptake. Our data suggests that NEP averages ～ 70 g C m^?2 year^?1 for ponderosa pine forests in Oregon. About 85% of the total carbon storage in biomass on the survey plots exists in stands greater than 100 years, which has implications for managing forests for carbon sequestration. To investigate variation in carbon storage and fluxes with disturbance, simulation with process models requires a dynamic parameterization for biomass allocation that depends on stand age, and should include a representation of competition between multiple plant functional types for space, water, and nutrients.     

Emergent climate and CO2 sensitivities of net primary productivity in ecosystem models do not agree with empirical data in temperate forests of eastern North America

Ecosystem models show divergent responses of the terrestrial carbon cycle to global change over the next century. Individual model evaluation and multimodel comparisons with data have largely focused on individual processes at subannual to decadal scales. Thus far, data‐based evaluations of emergent ecosystem responses to climate and CO₂ at multidecadal and centennial timescales have been rare. We compared the sensitivity of net primary productivity (NPP) to temperature, precipitation, and CO₂ in ten ecosystem models with the sensitivities found in tree‐ring reconstructions of NPP and raw ring‐width series at six temperate forest sites. These model‐data comparisons were evaluated at three temporal extents to determine whether the rapid, directional changes in temperature and CO₂ in the recent past skew our observed responses to multiple drivers of change. All models tested here were more sensitive to low growing season precipitation than tree‐ring NPP and ring widths in the past 30 years, although some model precipitation responses were more consistent with tree rings when evaluated over a full century. Similarly, all models had negative or no response to warm‐growing season temperatures, while tree‐ring data showed consistently positive effects of temperature. Although precipitation responses were least consistent among models, differences among models to CO₂ drive divergence and ensemble uncertainty in relative change in NPP over the past century. Changes in forest composition within models had no effect on climate or CO₂ sensitivity. Fire in model simulations reduced model sensitivity to climate and CO₂, but only over the course of multiple centuries. Formal evaluation of emergent model behavior at multidecadal and multicentennial timescales is essential to reconciling model projections with observed ecosystem responses to past climate change. Future evaluation should focus on improved representation of disturbance and biomass change as well as the feedbacks with moisture balance and CO₂ in individual models.