Monthly 1976-2005 mean daily minimum temperature: ANUClimate 1.0, 0.01 degree, Australian Coverage

Monthly mean daily minimum temperature for the Australian continent between 1976-2005. Daily minimum temperature regulates rates of plant growth. Modelled by fitting trivariate thin plate smoothing spline functions of longitude, latitude and vertically exaggerated elevation to observed and estimated monthly 1976-2005 daily minimum temperature means at 1541 Bureau of Meteorology stations. Missing monthly temperature values were estimated by regression with the long term station that provided estimates of the 1976-2005 monthly means with with the least standard error. This was applied to each station with at least 5 years of record between 1921 and 2012. Quality controls were applied to the regression and the surface fitting processes to remove poor quality data. Thus, of 1597 stations with at least 5 years of record, 38 stations had poor regressions with long term stations and a further 18 stations had extreme studentised residuals from initial spline analyses. These were commonly stations with an old or short period of record, or with an imprecise location. The spline analysis also incorporated the impact of distance from the coast as provided by eMAST_ANUClimate_fx_dist_v1m0. The root mean square error of individual cross validation residuals provided by the spline analysis at 225 stations with near complete records, of at least 28 years, is 0.47 degrees Celsius. The incorporation of distance from the coast reduced overall cross validation errors by around 10%. A comprehensive assessment of the analysis and the factors contributing to the quality of the final interpolated monthly mean daily minimum temperature grids is in preparation.

本数据集涵盖1976年至2005年澳大利亚大陆的月平均日最低气温数据。日最低气温是调控植物生长速率的关键因子。本研究通过构建三元薄板平滑样条（trivariate thin plate smoothing spline）函数开展空间插值建模，以经度、纬度与经垂直拉伸的高程为自变量，拟合1541个澳大利亚气象局（Bureau of Meteorology）站点的1976-2005年月度日最低气温均值观测值与估算值。
针对缺测的月度气温值，本研究采用回归方法进行补全：以1976-2005年月度均值估算标准误差最小的长期观测站作为参考站点，该方法适用于1921年至2012年间拥有至少5年观测记录的所有站点。
研究团队对回归与曲面拟合过程实施严格质量控制，以剔除低质量数据。因此，在1597个拥有至少5年观测记录的气象站中，38个站点与参考长期站的回归效果不佳，另有18个站点在初始样条分析中存在极端学生化残差。此类站点通常为观测记录年代久远、时长过短，或地理位置定位精度不足的站点。
本次样条分析还纳入了eMAST_ANUClimate_fx_dist_v1m0提供的海岸距离影响因子。对225个拥有至少28年近乎完整观测记录的站点开展交叉验证，结果显示样条分析得到的单个交叉验证残差均方根误差为0.47摄氏度。纳入海岸距离因子后，整体交叉验证误差降低了约10%。
目前，针对本次分析方法及最终插值得到的月平均日最低气温格点数据质量影响因子的全面评估工作正在筹备中。