Montagnani Leonardo, Zanotelli Damiano, Tagliavini Massimo, Tomelleri Enrico
Faculty of Science and Technology Free University of Bolzano Bolzano Italy.
Forest Services Autonomous Province of Bolzano Bolzano Italy.
Ecol Evol. 2017 Nov 29;8(1):416-434. doi: 10.1002/ece3.3633. eCollection 2018 Jan.
Model parameterization and validation of earth-atmosphere interactions are generally performed using a single timescale (e.g., nearly instantaneous, daily, and annual), although both delayed responses and hysteretic effects have been widely recognized. The lack of consideration of these effects hampers our capability to represent them in empirical- or process-based models. Here we explore, using an apple orchard ecosystem in the North of Italy as a simplified case study, how the considered timescale impacts the relative importance of the single environmental variables in explaining observed net ecosystem exchange (NEE) and evapotranspiration (ET). Using 6 years of eddy covariance and meteorological information as input data, we found a decay of the relative importance of the modeling capability of photosynthetically active radiation in explaining both NEE and ET moving from half-hourly to seasonal timescale and an increase in the relative importance of air temperature () and VPD. Satellite NDVI, used as proxy of leaf development, added little improvement to overall modeling capability. Increasing the timescale, the number of variables needed for parameterization decreased (from 5 to 1), while the proportion of variance explained by the model increased ( from 0.56-0.78 to 0.85-0.90 for NEE and ET respectively). The wavelet coherence and the phase analyses showed that the two variables that increased their relative importance when the scale increased (, VPD) were not in phase at the correlation peak of both ET and NEE. This phase shift in the time domain corresponds to a hysteretic response in the meteorological variables domain. This work confirms that the model parameterization should be performed using parameters calculated at the appropriate scale. It suggests that in managed ecosystems, where the interannual variability is minimized by the agronomic practices, the use of timescales large enough to include hysteretic and delayed responses reduces the number of required input variables and improves their explanatory capacity.
地球 - 大气相互作用的模型参数化和验证通常使用单一时间尺度（例如，近乎瞬时、每日和每年）来进行，尽管延迟响应和滞后效应已得到广泛认可。对这些效应的忽视阻碍了我们在基于经验或过程的模型中对它们进行表征的能力。在这里，我们以意大利北部的一个苹果园生态系统作为简化案例研究，探讨所考虑的时间尺度如何影响单个环境变量在解释观测到的生态系统净交换（NEE）和蒸散（ET）方面的相对重要性。使用6年的涡度相关和气象信息作为输入数据，我们发现光合有效辐射在解释NEE和ET的建模能力方面，其相对重要性从半小时尺度到季节尺度逐渐衰减，而气温（）和水汽压差（VPD）的相对重要性增加。用作叶片发育指标的卫星归一化植被指数（NDVI）对整体建模能力的提升作用不大。随着时间尺度的增加，参数化所需的变量数量减少（从5个降至1个），而模型解释的方差比例增加（NEE和ET分别从0.56 - 0.78增至0.85 - 0.90）。小波相干分析和相位分析表明，当尺度增加时相对重要性增加的两个变量（，VPD）在ET和NEE的相关峰值处并不同相。这种时域中的相位偏移对应于气象变量域中的滞后响应。这项工作证实，模型参数化应使用在适当尺度上计算的参数来进行。这表明在管理生态系统中，由于农艺措施使年际变异性最小化，使用足够大的时间尺度以纳入滞后和延迟响应会减少所需输入变量的数量并提高其解释能力。
