Objective This study aims to enhance the prediction accuracy of agricultural drought in Guizhou Province to address the increasing drought risks in the context of climate change. Methods For this purpose, Extreme Gradient Boosting tree (XGBoost) and Long Short-Term Memory network (LSTM) models were constructed based on the historical data of standardized precipitation evapotranspiration index (SPEI3) in Guizhou Province from 1979 to 2023. In this study, a hybrid XGBoost-ConvLSTM model was proposed for the first time, which integrated XGBoost, a Convolutional Neural Network (CNN) and LSTM, and could capture the spatiotemporal characteristics of droughts more accurately. A combination of K-means clustering and Thiessen polygons was used to calculate the rainfall area and divide the 84 meteorological stations in Guizhou Province into three subregions to assess the potential of each model for agricultural drought early warning. Results (1) The results showed that the predicted R2 of the XGBoost-ConvLSTM model for subregions Ⅰ, Ⅱ, and Ⅲ were 0.916, 0.877, and 0.901, respectively, which were better than those of XGBoost (0.760, 0.853, 0.735) and LSTM (0.760, 0.778, 0.710).(2) The characteristics of agricultural drought conditions in Guizhou Province in the next 30 years showed significant spatiotemporal variations, with 2024—2026 being a period of frequent droughts, peaking in 2024; and the drought situation was expected to be alleviated after 2030. (3) The centroid model was used to analyze the spatiotemporal migration of the drought centroid in Guizhou Province over the next decade, which showed that the drought centroid exhibited phased and abrupt spatiotemporal migration characteristics. Overall, the drought centroid in Guizhou Province gradually expanded from the south-central part of the province toward the north. Conclusion The XGBoost-ConvLSTM proposed in this study can be used for agricultural drought prediction in Guizhou Province, and its performance is better than traditional models. Agricultural drought in Guizhou Province exhibits significant phased and spatial migration characteristics, reflecting the complexity of its spatiotemporal evolution. The proposed XGBoost-ConvLSTM model demonstrates superior capability in capturing these characteristics and outperforms traditional models in agricultural drought prediction.
American Meteorological Society. Bulletin of the american meteorological society[M]. Boston, USA: American Meteorological Society, 1920.
[2]
DixitS, JayakumarK V. A study on copula-based bivariate and trivariate drought assessment in Godavari River basin and the teleconnection of drought with large-scale climate indices[J]. Theoretical and Applied Climatology, 2021,146(3):1335-1353.
[3]
DaiA. Increasing drought under global warming in observations and models[J]. Nature Climate Change, 2013,3(1):52-58.
FanL, LüC H, WangX C, et al. Applicability of EPIC model on simulation of growth and yield of winter wheat and summer maize in the North China Plain[J]. Journal of Triticeae Crops, 2014,34(12):1677-1684.
[8]
陈鹏,潘锋,吴麟.遥感技术在干旱监测中的应用研究分析[J].华东科技,2023(7):51-53.
[9]
ChenP, PanF, WuL. Research and analysis on application of remote sensing technology in drought monitoring[J]. East China Science & Technology, 2023(7):51-53.
[10]
BelaynehA, AdamowskiJ, KhalilB, et al. Long-term SPI drought forecasting in the Awash River Basin in Ethiopia using wavelet neural network and wavelet support vector regression models[J]. Journal of Hydrology, 2014,508:418-429.
[11]
LiuX F, ZhuX F, ZhangQ, et al. A remote sensing and artificial neural network-based integrated agricultural drought index: index development and applications[J]. Catena, 2020,186:104394.
[12]
WuJ F, ChenX H, YuZ X, et al. Assessing the impact of human regulations on hydrological drought development and recovery based on a ‘simulated-observed’ comparison of the SWAT model[J]. Journal of Hydrology, 2019,577:123990.
ZhangY L, WuZ Y, HeH. Agricultural drought assessment based on the coupled hydrology-crop growth model and CWAPI[J]. Journal of Hydraulic Engineering, 2022,53(10):1168-1179,1193.
[15]
GuptaH V, WagenerT, LiuY Q. Reconciling theory with observations: elements of a diagnostic approach to model evaluation[J]. Hydrological Processes:2008,22(18):3802-3813.
WuZ Y, ChengD D, HeH, et al. Research progress of composite drought index[J]. Water Resources Protection, 2021,37(1):36-45.
[18]
YangT T, AsanjanA A, FaridzadM, et al. An enhanced artificial neural network with a shuffled complex evolutionary global optimization with principal component analysis[J]. Information Sciences, 2017,418/419:302-316.
[19]
BrownJ F, WardlowB D, TadesseT, et al. The vegetation drought response index (VegDRI): a new integrated approach for monitoring drought stress in vegetation[J]. GIScience & Remote Sensing, 2008,45(1):16-46.
ZhangJ H, ZhangQ, XuD H, et al. Temporal and spatial distribution prediction of meteorological drought based on EBK and LSTM model[J]. Yellow River, 2020,42(8):77-82.
DuJ, DongS J, HeY. Drought prediction model for the Xilingol grassland based on BO-CNN-LSTM[J]. Grassland and Turf, 2024,44(4):64-75.
[24]
HouZ J, WangB B, ZhangY C, et al. Drought prediction in Jilin Province based on deep learning and spatio-temporal sequence modeling[J]. Journal of Hydrology, 2024,642:131891.
[25]
MokhtarA, JalaliM, HeH M, et al. Estimation of SPEI meteorological drought using machine learning algorithms[J]. IEEE Access, 2021,9:65503-65523.
[26]
RheeJ, ImJ. Meteorological drought forecasting for ungauged areas based on machine learning: using long-range climate forecast and remote sensing data[J]. Agricultural and Forest Meteorology, 2017,237/238:105-122.
[27]
贵州省水利厅.贵州省防汛抗旱统计年鉴(2011—2020)[M].贵阳:贵州省水利厅,2022.
[28]
Guizhou Provincial Department of Water Resources. Guizhou flood control and drought relief statistical yearbook(2011—2020)[M]. Guiyang: Guizhou Provincial Department of Water Resources, 2022.
ChenL H, HeZ H, ZhangY B, et al. Spatial-temporal characteristics analysis of agricultural drought based on vegetation supply water index in Guizhou[J]. China Rural Water and Hydropower, 2023(4):158-166,176.
[31]
ChenT Q, GuestrinC. XGBoost: a scalable tree boosting system[C]∥Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. New York: Association for Computing Machinery,2016.
[32]
SchmidhuberJ, HochreiterS. Long short-term memory[J]. Neural Computation, 1997,9(8):1735-1780.
YanX D, LiY, XiaY, et al. Analysis of characteristics and causes of persistent drought in Guizhou during 1961-2016[J]. Mid-Low Latitude Mountain Meteorology, 2019,43(3):1-7.
WangX J, BaiH, ZhouW Y, et al. Comparative analysis of drought in Guizhou from July to August between the year of 2011 and 2013 and the effect on agriculture[J]. Tianjin Agricultural Sciences, 2014,20(11):118-124.
DingY, XuD H, CaoL H, et al. Applicability of the LSTM and ARIMA model in drought prediction based on CEEMD: a case study of Xinjiang[J]. Arid Zone Research, 2022,39(3):734-744.
[39]
KattenbornT, LeitloffJ, SchieferF, et al. Review on Convolutional Neural Networks (CNN) in vegetation remote sensing[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2021,173:24-49.
[40]
YanY L, WangH, LiG P, et al. Projection of future extreme precipitation in China based on the CMIP6 from a machine learning perspective[J]. Remote Sensing, 2022,14(16):4033.
XieM, GaoJ L, SunJ Y, et al. Spatiotemporal variation of drought in Inner Mongolia estimated based on the standardized precipitation evapotranspiration index[J]. Journal of Irrigation and Drainage, 2022,41(6):140-146.