Accurate rainfall forecasting in watersheds is of indispensable importance for predicting streamflow and flash floods. This paper investigates the accuracy of several forecasting technologies based on Wavelet Packet Decomposition (WPD) in monthly rainfall forecasting. First, WPD decomposes the observed monthly rainfall data into several subcomponents. Then, three data-based models, namely Back-propagation Neural Network (BPNN) model, group method of data handing (GMDH) model, and autoregressive integrated moving average (ARIMA) model, are utilized to complete the prediction of the decomposed monthly rainfall series, respectively. Finally, the ensemble prediction result of the model is formulated by summing the outputs of all submodules. Meanwhile, these six models are employed for benchmark comparison to study the prediction performance of these conjunction methods, which are BPNN, WPD-BPNN, GMDH, WPD-GMDH, ARIMA, and WPD-ARIMA models. The paper takes monthly data from Luoning and Zuoyu stations in Luoyang city of China as the case study. The performance of these conjunction methods is tested by four quantitative indexes. Results show that WPD can efficiently improve the forecasting accuracy and the proposed WPD-BPNN model can achieve better prediction results. It is concluded that the hybrid forecast model is a very efficient tool to improve the accuracy of mid- and long-term rainfall forecasting.

Gridded gross domestic product (GDP) data are a crucial land surface parameter for many geoscience applications. Recently, machine learning approaches have become powerful tools in generating gridded GDP data. However, most machine learning approaches for gridded GDP estimation seldom consider the geographical properties of input variables. Therefore, in this study, a geographically weighted stacking ensemble learning approach was developed to generate gridded GDP data. Three algorithms—random forest, XGBoost, and LightGBM—were used as base models, and the linear regression in stacking ensemble learning was replaced by geographically weighted regression to locally fuse the three predictions. A case study was conducted in China to demonstrate the effectiveness of the proposed approach. The results showed that the proposed GDP downscaling approach outperformed the three base models and traditional stacking ensemble learning. Meanwhile, it had good predictive power on county-level GDP test data with R² of 0.894, 0.976, and 0.976 for the primary, secondary, and tertiary sectors, respectively. Moreover, the predicted 1 km gridded GDP data had a high accuracy (R² = 0.787) when evaluated by town-level GDP data. Hence, the proposed GDP downscaling approach provides a valuable option for generating gridded GDP data. The generated 1 km gridded GDP data of China from 2020 are of great significance for other applications.

In the past decade, more than 300 people have died per year on average due to mountain torrents in China. Mountain torrents mostly occur in ungauged small and medium-sized catchments, so it is difficult to maintain high accuracy of flood prediction. In order to solve the problem of the low accuracy of flood simulation in the ungauged areas, this paper studies the influence of different methods on the parameter regionalization of distributed hydrological model parameters in hilly areas of Hunan Province. According to the terrain, landform, soil and land use characteristics of each catchment, we use Shortest Distance, Attribute Similarity, Support Vector Regression, Generative Adversarial Networks, Classification and Regression Tree and Random Forest methods to create parameter regionalization schemes. In total, 426 floods of 25 catchments are selected to calibrate the model parameters, and 136 floods of 8 catchments are used for verification. The results showed that the average values of the Nash–Sutcliffe coefficients of each scheme were 0.58, 0.64, 0.60, 0.66, 0.61 and 0.68, and the worst values were 0.27, 0.31, 0.25, 0.43, 0.35 and 0.59. The random forest model is the most stable solution and significantly outperforms other methods. Using the random forest model to regionalize parameters can improve the accuracy of flood simulation in ungauged areas, which is of great significance for flash flood forecasting and early warning.

Digital Surface Model (DSM) is the fundamental data in various geoscience applications, such as city 3D modeling and urban environment analysis. The freely available DSM often suffers from limited spatial resolution. Super-resolution (SR) is a promising technique to increase the spatial resolution of DSM. However, most existing SR models struggle to reconstruct spatial details such as buildings, valleys, and ridges. This paper proposes a novel DSM super-resolution (DSMSR) model that integrates high-resolution remote sensing imagery using generative adversarial networks. The generator in DSMSR contains three modules. The first DSM feature extraction module uses the residual-in-residual dense block (RRDB) to extract features from low-resolution DSM. The second multiscale attention feature extraction module employs the pyramid convolutional residual dense (PCRD) blocks to capture spatial details of ground objects at multiple scales from remote sensing imagery. The third DSM reconstruction module uses a squeeze-and-excitation (SE) block to fuse the extracted features from low-resolution DSM and high-resolution remote sensing imagery for generating SR DSM. The discriminator of DSMSR uses the relativistic average discriminator (RaD) for adversarial learning. The slope loss is further introduced to ensure the accurate representation of topographic features. We evaluate DSMSR on four different terrain regions in the UK to downscale the 30-m AW3D30 DSM to 5-m DSM. The experimental results indicate that DSMSR outperforms traditional interpolation algorithms and four existing deep learning-based SR models. The DSMSR restores more spatial detail of topographic features and generates more accurate image quality, elevation and terrain metrics.