Search

Given the increasing impact of extreme rainfall and flooding on human life, studying and predicting changes in atmospheric water vapor (AWV) becomes particularly important. This paper analyzes the moderate-resolution imaging spectroradiometer (MODIS) data of the East Asian region from January 2003 to February 2023. The AWV data are examined in the time and frequency domain using methods such as empirical orthogonal function (EOF), Mann–Kendall (MK) analysis, and others. Additionally, four prediction models are applied to forecast the monthly average AWV data for the next year. The accuracy of these models is evaluated using metrics such as mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). The findings reveal several key insights: (1) The East Asian region exhibits highly variable seasonal variability in AWV, with identified mutation points after the MK test. (2) Spatial analysis shows high AWV data in the southern coastal areas of China, Thailand, Myanmar, Nansha Islands, and other regions during winter, while the Qinghai-Tibet Plateau region experiences low AWV during summer. (3) The first mode obtained through EOF decomposition contributes over 60% of the variance. Analysis of this mode reveals an increasing trend in AWV data for regions such as the Indian peninsula, Mongolia, and central and northeastern China over the past nine years. Conversely, the Bay of Bengal, Spratly Islands, eastern coast, and certain areas display a decreasing trend. (4) Employing the ensemble empirical mode decomposition (EEMD), the study identifies AWV data as a non-stationary series with an overall decreasing trend from 2003 to 2022. The filtered AWV series undergoes fast Fourier transform (FFT), uncovering periodicities of 2.6 years, 5 years, and 19 years. (5) Among the four forecasting models compared, the seasonal autoregressive integrated moving average model (SARIMA) demonstrates superior performance with the smallest MSE of 0.00782, MAE of 0.06977, RMSE of 0.08843, and the largest R2 value of 0.98454. These results clearly indicate that the SARIMA model provides the best fit. Therefore, the SARIMA forecasting model can be effectively utilized for forecasting AWV data, offering valuable insights for studying weather variability. [ABSTRACT FROM AUTHOR]

The total column water vapor (TCWV) is a relatively active component in the atmosphere and an important detection object of climate change. Exploring the spatiotemporal modes characteristics of TCWV and predicting its changing trends can provide a reference for human beings to deal with climate change and formulate corresponding countermeasures. The TCWV data over China region by using the Atmospheric Infrared Sounder (AIRS) dataset from 2002 to 2022 were obtained. The empirical orthogonal function (EOF) analysis, linear regression, Mann-Kendall (M-K) mutation test, Seasonal Autoregressive Integrated Moving Average (SARIMA) model and other methods were used to discuss the spatiotemporal modes characteristics of TCWV in the China region on the monthly, seasonal, and annual scales and verify the rationality of the forecast of the monthly average trend of TCWV in the next year. The obtained results show that: (1) The annual and seasonal scales spatial distributions of TCWV in China are roughly consistent, with obvious latitudinal distribution characteristics. That is, the TCWV in the low latitude region, especially in the tropical region, is larger, and it gradually decreases with the increase of the latitude. Furthermore, the TCWV in the eastern region is higher than that in the western region at the same latitude; (2) The EOF analysis results show that its first mode can better reflect the typical distribution characteristics of the southeast-northwest positive distribution in China; (3) From 2002 to 2022, the TCWV in China shows an upward trend and the TCWV increases at a rate of 0.0413 kg/m2 per year, which may be related to the long-term increase of air temperature in recent years; (4) The inter-monthly variation of TCWV shows a slightly positive skewed 'bell-shaped' curve, with the maximum in summer, the minimum in winter and the similar distribution in spring and autumn. As can be seen from the M-K curves of the four seasons, each season has different mutation points; (5) Forecasting the TCWV was done using time series monthly average values from September 2002 to February 2022. SARIMA (3, 1, 3) × (0, 1, 1, 12) was identified as the best model. This model passed the residual normality test and the forecasting evaluation statistics show that MAPE = 2.65%, MSE = 0.3229 and the R2-score = 0.9949. As demonstrated by the results, the SARIMA model is a good model for forecasting TCWV in the China region. [ABSTRACT FROM AUTHOR]