目前数值模式物理过程参数化的一个核心发展方向是在参数化方案中发展与分辨率无关的物理参数。参数化方案中大多数定常参数都是基于观测数据的统计结果或者数值试验的统计结论而得到的不随时空变化而改变的常数参数。本研究针对现有参数化方案中的定常参数提出了基于物理过程的动态计算的新方法，研发了基于物理过程的动态化参数方案改进现有的定常参数方案。
      本研究针对目前数值模拟/预测偏差较大的云和降水过程，选择了CAM的Zhang-McFarlane（ZM）深对流参数化方案和Park云宏物理参数化方案开展了相关数值试验，基于对流和云形成的物理过程，提出两个重要常数参数的动态计算公式，利用对流有效位能CAPE对特征调整时间尺度（τ）进行动态化，为ZM方案引入垂直速度项。同时，根据不同温度下，云内相对湿度和云的频率分布的时空对应关系的统计分析结果，通过曲线拟合得到成云临界相对湿度阈值（RHc）基于温度的动态计算公式，实现基于物理过程对现有参数化方案中常数参数的动态化，并在全球模式中进行数值模拟验证了该方法的正确性，结果表明，动态参数方案可以有效提高模式对云和降水的模拟效果。
      本文的主要结论如下：
    （1）本文在全球气候模式中构建了基于物理过程的特征调整时间尺度τ的动态参数新方案，实现了对原有深对流方案的改进。利用CAPE对τ进行动态计算，对ZM深对流方案引入垂直速度项，并将其应用于NCAR CAM6。CAM6模拟结果表明动态参数τ不但有效地改善了地形陡峭的地区（如青藏高原、安第斯山脉）、新几内亚及其周边岛屿等地区的降水偏差，而且减少了北半球夏季印度洋-太平洋暖池中部和赤道北侧的降水偏差，以及北半球冬季新几内亚和印度洋-太平洋暖池中部等地区的降水量正偏差。动态参数τ通过增强西侧上升气流和东侧下沉气流改善了太平洋沃克环流的模拟效果。动态参数τ使大部分地区的深对流降水的强度显著增加至少1mm/day，最大超过3mm/day，强深对流降水频率的最大增加量达到参照试验最大频率的50%。在CAPE值和时空变率相对较大的区域，动态参数τ对深对流降水的影响较大，如热带海洋和陡峭地形地区。根据对模式在1º和2º水平分辨率下模拟的年平均降水量的比较，总降水量差异主要集中在热带海洋上；在不同分辨率下，模式模拟的全球海洋、全球陆地的年平均总降水量差值绝对值仅为0.15和0.03mm/day。
    （2）本文基于云形成机理研发了成云临界相对湿度阈值RHc的动态参数新方案，改进了模式对降水和云量的模拟效果。利用温度来调制云形成的相对湿度阈值RHc，将RHc从常数改为由温度计算的动态参数，并将其应用于CAM6模式。CAM6的模拟结果表明，与原方案相比，敏感试验显著降低了副热带东部海洋上空的低云的负偏差。低云的增加伴随着液态水路径的增加，这有助于减少亚热带的短波辐射强迫偏差。同时，动态参数RHc主要增加了700hPa以下的云量，减少了400hPa以上的云量。动态参数RHc方案改善了非洲东南部、澳大利亚北部等地区的年降水量的正偏差，对陆地降水量的高估有一定的改善作用，这表明原方案中热带大陆的云形成阈值太低。在不同水平分辨率下的模拟结果表明，该方法对分辨率不太敏感。与原方案相比，动态RHc不但对700hPa附近的低空冷云有显著的增强作用，而且还减小了273.15K以上的RHc，增加了低空暖云。
    （3）本文采用动态参数组合方案改善了数值模式对云和降水模拟效果，并探究了其物理过程相互作用引起的联合效应。将动态参数组合试验与参照试验和仅使用一个动态参数的敏感试验结果进行比较，比较结果表明同时应用两个动态参数的组合试验对年平均总降水量的模拟效果比单独使用一个动态参数的效果更好，两个参数间具有联合效应，对云和降水的改善具有相互作用。对于太平洋沃克环流的模拟，组合试验更接近观测/再分析数据结果。组合试验的深对流的降水频率和降水强度的模拟依然有显著改善，且同时使用两个参数使强深对流降水的频率增加更多，弱深对流降水的频率减少更多。在云模拟方面，组合试验中的低云增量虽然略微减少，但最大增量仍能达到20%以上。组合试验在一定程度上减轻了单一参数对于一些地区的偏差矫枉过正的现象。此外，组合试验对水平分辨率的敏感性较低，在不同分辨率下都可以提高模式的模拟效果。
      由于深对流和云存在明显空间差异性，目前参数化中基于观测统计数据分析得到的定常参数缺乏物理基础，会导致云和降水的模拟偏差较大，因此本研究旨在提出应用于全球模式中基于物理的动态参数，有效改进模式对云和降水的模拟偏差问题。以上研究结果说明了基于物理过程对动态参数设计的有效性和正确性，以及将通过对不同动态参数进行组合可以有效发挥两者对模式不同方面的改进效果，有效避免单一参数对模式的矫枉过正，且这两种动态参数的模拟效果并不会因为分辨率不同而产生太大差异。这为今后模式的发展提供了一个新的方向。

      At present, the core of the development path of parameterization is to develop resolution independent physical parameters in parameterization schemes. Most of the constant parameters in parameterization scheme are based on statistical results of observed data or statistical conclusions of numerical experiments, and do not change with time and space. This study proposes dynamic calculations based on physics for the constant parameters in existing parameterization schemes, realizing the dynamic calculation of physical parameters for constant parameters in existing parameterization schemes based on physics.

      This study specifically targets cloud and precipitation processes with large numerical simulation/prediction bias, and selects the Zhang-McFarlane (ZM) deep convection parameterization scheme and Park cloud macrophysical parameterization scheme in the CAM for testing. Based on the physical processes of convection and cloud formation dynamic calculation formulas of two important parameters are proposed. A dynamic calculation formula of characteristic adjustment time scale (τ) based on CAPE was carried out by introducing vertical velocity into the ZM scheme. At the same time, based on the statistical analysis results of the spatiotemporal correspondence between the relative humidity and the frequency distribution of cloud at different temperatures, a temperature based dynamic calculation formula for critical relative humidity (RHc) was fitted, achieving dynamic calculation of constant parameters based on physical processes in parameterization schemes. Numerical simulations were conducted in global models to verify the correctness of this method. The results showed that dynamic parameters can effectively improve the simulation of clouds and precipitation.

      The main conclusions of this study are as follows:

      (1) This study constructs a new dynamic parameter scheme of characteristic adjustment time scale τ based on physical processes in a global climate model, which improves the original deep convection scheme. By utilizing CAPE to calculate τ dynamically, a vertical velocity term is introduced into the ZM deep convection scheme, and this method is applied in the NCAR CAM6. The simulation results of the CAM6 model show that the dynamic parameter τ not only effectively improves the precipitation bias in steep terrain regions (such as the Tibet Plateau, the Andes Mountains), New Guinea, and its surrounding islands, but also reduces the precipitation bias in the central part of the Indian Ocean-Pacific warm pool and on the northern side of the equator in JJA, as well as the positive precipitation bias in New Guinea, the central part of the Indian Ocean-Pacific warm pool and other regions in DJF. The dynamic parameter τ improves simulation effect of the Pacific Walker circulation by enhancing the updrafts on the west side and downdrafts on the east side. The dynamic parameter τ significantly increases the intensity of deep convection precipitation in most regions by at least 1mm/day, with a maximum increase of more than 3mm/day, and the maximum increase in frequency of strong deep convection precipitation reaches 50% of the maximum frequency in control experiments. In areas with relatively large CAPE, temporal and spatial variability of CAPE (such as tropical ocean and steep terrain regions), the dynamic τ has a greater impact on deep convection precipitation. According to a comparison of annual mean precipitation simulated by the model at 1º and 2º horizontal resolutions, the total precipitation difference is mainly concentrated in tropical oceans. Under different resolutions, the absolute value of annual mean total precipitation difference simulated by the model is only 0.15 and 0.03mm/day for global oceans and global land.

      (2) This study develops a new dynamic parameter scheme for the critical relative humidity threshold (RHc) of cloud formation based on the mechanism of cloud formation, and improves the simulation effect of the model on precipitation and cloud fraction. Using temperature to modulate RHc for cloud formation, changing RHc from a constant to a dynamic parameter calculated from temperature, and applying it to the CAM6. The simulation results of CAM6 indicate that compared with the original scheme, the sensitivity test significantly reduces the negative bias of low clouds over the eastern subtropical ocean. The increase in low clouds is accompanied by an increase in LWP, which helps reduce the short-wave radiation forcing bias in the subtropics. Also, the dynamic parameter RHc mainly increases cloud fraction below 700hPa and reduces cloud fraction above 400hPa. Dynamic RHc scheme improves the positive bias of annual mean precipitation in southeastern Africa, northern Australia and other regions, which has a certain improvement effect on the overestimation of land precipitation. This indicates that the cloud formation threshold in the tropical continent is too low in the original scheme. The simulation results under different horizontal resolutions indicate that this method is not sensitive to resolution. Compared with the original scheme, dynamic RHc not only has a significant enhancement effect on low-altitude cold clouds near 700hPa, but also reduces RHc above 273.15K and increases low-altitude warm clouds.

      (3) This study adopts a dynamic parameter combination scheme to improve the simulation effect of clouds and precipitation in numerical models, and explores the joint effects caused by the interaction of their physical processes. Comparing the results of dynamic parameter combination experiments with those of control experiments and sensitivity experiments using only a single dynamic parameter, it is found that the combined experiment using both dynamic parameters simultaneously shows better simulation results for annual mean total precipitation than using single dynamic parameter. There is a joint effect between the two parameters, which has an interactive effect on the improvement of cloud and precipitation. For the simulation of the Pacific Walker circulation, the combined experiment result is closer to the observation/reanalysis data. The simulation of precipitation frequency and intensity in deep convection in the combined experiment still has significant improvement, and using both parameters simultaneously increases the frequency of strong deep convection precipitation more and reduces the frequency of weak deep convection precipitation more. In terms of cloud simulation, the increase in low cloud in the combined experiment is slightly reduced, but the maximum increase can still reach more than 20%. The combined experiment alleviates the phenomenon of overcorrection of bias by using single dynamic parameter in some regions to a certain extent. Besides, the combined experiment has low sensitivity to horizontal resolution, and can improve the simulation under different resolutions.

      Due to the spatial distribution differences of deep convection and clouds are significant, the constant parameters based on observation and statistical data analysis lack physical basis, which can lead to significant simulation biases in clouds and precipitation. This study aims to propose physical based dynamic parameters applied to global models to effectively improve the simulation bias of clouds and precipitation. The above research results verify the effectiveness and correctness of dynamic parameter design based on physical processes. Combining different dynamic parameters in simulation can effectively exert the improvement effect of both parameters on different aspects, effectively avoiding overcorrection of using a single parameter. Besides, the simulation effect of these two dynamic parameters does not vary greatly due to different resolutions. These provide a new direction for the future development of models.