当前位置: X-MOL 学术Comput. Geosci. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Effect of missing data on short time series and their application in the characterization of surface temperature by detrended fluctuation analysis
Computers & Geosciences ( IF 4.4 ) Pub Date : 2021-04-20 , DOI: 10.1016/j.cageo.2021.104794
J.L. López , S. Hernández , A. Urrutia , X.A. López-Cortés , H. Araya , L. Morales-Salinas

Climate change is deeply impacting the society on different scales. Decision making becomes a complex task when, in adverse weather conditions, the meteorological records show missing data due to failures of measuring instruments. Several investigations have proposed optimized regression methods, K-nearest-neighbor imputation, and multiple imputations for the treatment of missing data; however, there is less information about the application of imputation methods for the treatment of missing data on short meteorological records. Therefore, the expected confidence in the results requires using robust analysis methods that depend the least as possible on the length of the records and the number of missing data. In this research, the performance of detrended fluctuation analysis applied on temperature short record was studied when K-nearest-neighbor and neural networks are used as imputation techniques, and compared with the performance without data imputation. The results showed the robustness when it is applied to a short time series with missing data and without data imputation. In this aspect, the DFA method only requires removing the seasonality from the temperature records to get good performance.



中文翻译:

缺失数据对短时间序列的影响及其在去趋势波动分析中表征表面温度的应用

气候变化在不同程度上深深地影响着社会。当在不利的天气条件下,气象记录显示由于测量仪器故障而导致数据丢失时,决策就成为一项复杂的任务。一些研究提出了优化的回归方法,K近邻插补和多重插补来处理丢失的数据。但是,关于插补方法在处理短气象记录中丢失数据方面的应用信息较少。因此,对结果的预期置信度要求使用可靠的分析方法,该方法应尽可能少地依赖于记录的长度和丢失的数据的数量。在这项研究中 以K-近邻和神经网络为插补技术,研究了去趋势波动分析在温度短记录上的性能,并与无数据插补的性能进行了比较。结果表明,将其应用于具有缺失数据且没有数据插补的短时间序列时,具有鲁棒性。在这方面,DFA方法仅需要从温度记录中删除季节性即可获得良好的性能。

更新日期:2021-04-23
down
wechat
bug