Meteorological drought forecasting based on a statistical model with machine learning techniques in Shaanxi province, China

Zhang, Rong; Chen, Zhaoyue; Xu, Lijun; Ou, Chun Quan

doi:10.1016/j.scitotenv.2019.01.431

Cited by 150 publications

(68 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Influenced by multiple climate factors, drought often displays nonlinear, nonstationary complex processes with periodic oscillations [6][7][8]. However, previous research mostly focused on linear variations and neglected the nonlinear responses to climate change [9][10][11].…”

Section: Introductionmentioning

confidence: 99%

Spatiotemporal Variation of Drought and Associated Multi-Scale Response to Climate Change over the Yarlung Zangbo River Basin of Qinghai–Tibet Plateau, China

Liu

Shan

et al. 2019

Remote Sensing

View full text Add to dashboard Cite

Drought is one of the most widespread and threatening natural disasters in the world, which has terrible impacts on agricultural irrigation and production, ecological environment, and socioeconomic development. As a critical ecologically fragile area located in southwest China, the Yarlung Zangbo River (YZR) basin is sensitive and vulnerable to climate change and human activities. Hence, this study focused on the YZR basin and attempted to investigate the spatiotemporal variations of drought and associated multi-scale response to climate change based on the scPDSI (self-calibrating Palmer drought severity index) and CRU (climate research unit) data. Results showed that: (1) The YZR basin has experienced an overall wetting process from 1956 to 2015, while a distinct transition period in the mid 1990s (from wet to dry) was detected by multiple statistical methods. (2) Considering the spatial variation of the scPDSI, areas showing the significantly wetting process with increasing scPDSI values were mostly located in the arid upstream and midstream regions, which accounted for over 48% area of the YZR basin, while areas exhibiting the drying tendency with decreasing scPDSI values were mainly concentrated in the humid southern part of the YZR basin, dominating the transition period from wet to dry, to which more attention should be paid. (3) By using the EEMD (ensemble empirical mode decomposition) method, the scPDSI over the YZR basin showed quasi-3-year and quasi-9-year cycles at the inter-annual scale, while quasi-15-year and quasi-56-year cycles were detected at the inter-decadal scale. The reconstructed inter-annual scale showed a better capability to represent the abrupt change characteristic of drought, which was also more influential to the original time series with a variance contribution of 55.3%, while the inter-decadal scale could be used to portray the long-term drought variation process with a relative lower variance contribution of 29.1%. (4) The multi-scale response of drought to climate change indicated that changes of precipitation (PRE) and diurnal temperature range (DTR) were the major driving factors in the drought variation at different time scales. Compared with potential evapotranspiration (PET), DTR was a much more important climate factor associated with drought variations by altering the energy balance, which is more obvious over the YZR basin distributed with extensive snow cover and glaciers. These findings could provide important implications for ecological environment protection and sustainable socioeconomic development in the YZR basin and other high mountain regions.

show abstract

Section: Introductionmentioning

confidence: 99%

Spatiotemporal Variation of Drought and Associated Multi-Scale Response to Climate Change over the Yarlung Zangbo River Basin of Qinghai–Tibet Plateau, China

Liu

Shan

et al. 2019

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Although many studies have investigated the SPEI variation over China [12][13][14][15][16]20], little attention has been focused on the detailed spatial variability and attribution of SPEI. This is largely because previous studies employed climate data from low-density stations or low-spatial-resolution reanalysis data, in addition to the computational complexity of differentiation in the equations used to calculate SPEI.…”

Section: Summary and Discussionmentioning

confidence: 99%

“…At present, SPEI is often estimated globally through station observations [12][13][14] or climate datasets released by climate research organizations [1,9,10,[15][16][17][18][19][20]. However, the detailed spatial patterns of SPEI cannot be obtained from these sources because of the current low densities of weather stations and the low spatial resolutions (e.g., >50 km per grid cell) of climate datasets [21,22].…”

Section: Introductionmentioning

confidence: 99%

“…Several studies have investigated climatic drought trends using SPEI over all or a part of China. However, such studies were carried out based on either station observations [12][13][14] or climatic proxy datasets with low spatial resolutions [15,16,20]. Because there is little information regarding the detailed spatial patterns of trends in SPEI, the development of sustainable strategies at a fine geographic scale to cope with the adverse effects of global warming in this region has been stymied.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Spatiotemporal Trends and Attribution of Drought across China from 1901–2100

2020

View full text Add to dashboard Cite

Investigating long-term drought trends is of great importance in coping with the adverse effects of global warming. However, little attention has been focused on studying the detailed spatial variability and attribution of drought variation in China. In this study, we first generated a 1 km resolution monthly climate dataset for the period 1901–2100 across China using the delta spatial downscaling method to assess the variability of the Standardized Precipitation Evaporation Index (SPEI). We then developed a simple approach to quantifying the contributions of water supply (precipitation) and demand (potential evapotranspiration, PET) on SPEI variability, according to the meaning of the differentiating SPEI equation. The results indicated that the delta framework could accurately downscale and correct low-spatial-resolution monthly temperatures and precipitation from the Climatic Research Unit and general circulation models (GCMs). Of the 27 GCMs analyzed, the BNU-ESM, CESM1-CAM5, and GFDL-ESM2M were found to be the most accurate in modeling future temperatures and precipitation. We also found that, compared with the past (1901–2017), the climate in the future (2018–2100) will tend toward significant droughts, although both periods showed a high spatial heterogeneity across China. Moreover, the proportion of areas with significantly decreasing SPEI trends was far greater than the proportion of those with increasing trends in most cases, especially for northwestern and northern China. Finally, the proposed approach to quantifying precipitation and PET contributions performed well according to logical evaluations. The percentage contributions of precipitation and PET on SPEI variability varied with study periods, representative concentration pathway scenarios, trend directions, and geographic spaces. In the past, PET contributions for significant downward trends and precipitation contributions for significantly upward trends accounted for 95% and 72%, while their future contributions were 57 ± 22%–149 ± 20% and 95 ± 27%–190 ± 58%, respectively. Overall, our results provide detailed insights for planning flexible adaptation and mitigation strategies to cope with the adverse effects of climate drought across China.

show abstract

“…XGBoost, an extension of gradient boosting decision tree (GDBT), is an ensemble learning algorithm proposed by Chen et al in 2016 [Zhang, Chen, Xu et al (2019)]. In recent years, XGBoost has demonstrated significant regression and classification performance in the Kaggle data-mining competition [Qi, Xu and Zhu (2019)].…”

Section: Xgboostmentioning

confidence: 99%

Novel Ensemble Modeling Method for Enhancing Subset Diversity Using Clustering Indicator Vector Based on Stacked Autoencoder

Wang¹,

Yan²

2019

Computer Modeling in Engineering &Amp; Sciences

View full text Add to dashboard Cite

A single model cannot satisfy the high-precision prediction requirements given the high nonlinearity between variables. By contrast, ensemble models can effectively solve this problem. Three key factors for improving the accuracy of ensemble models are namely the high accuracy of a submodel, the diversity between subsample sets and the optimal ensemble method. This study presents an improved ensemble modeling method to improve the prediction precision and generalization capability of the model. Our proposed method first uses a bagging algorithm to generate multiple subsample sets. Second, an indicator vector is defined to describe these subsample sets. Third, subsample sets are selected on the basis of the results of agglomerative nesting clustering on indicator vectors to maximize the diversity between subsets. Subsequently, these subsample sets are placed in a stacked autoencoder for training. Finally, XGBoost algorithm, rather than the traditional simple average ensemble method, is imported to ensemble the model during modeling. Three machine learning public datasets and atmospheric column dry point dataset from a practical industrial process show that our proposed method demonstrates high precision and improved prediction ability.

show abstract

Meteorological drought forecasting based on a statistical model with machine learning techniques in Shaanxi province, China

Cited by 150 publications

References 55 publications

Spatiotemporal Variation of Drought and Associated Multi-Scale Response to Climate Change over the Yarlung Zangbo River Basin of Qinghai–Tibet Plateau, China

Spatiotemporal Variation of Drought and Associated Multi-Scale Response to Climate Change over the Yarlung Zangbo River Basin of Qinghai–Tibet Plateau, China

Spatiotemporal Trends and Attribution of Drought across China from 1901–2100

Novel Ensemble Modeling Method for Enhancing Subset Diversity Using Clustering Indicator Vector Based on Stacked Autoencoder

Contact Info

Product

Resources

About