The construction and operation of cascade reservoirs has changed the natural hydrological cycle in the Yalong River Basin, and reduced the accuracy of hydrological forecasting. The impact of cascade reservoir operation on the runoff of the Yalong River Basin is assessed, providing a theoretical reference for the construction and joint operation of reservoirs. In this paper, eight scenarios were set up, by changing the reservoir capacity, operating location, and relative location in the case of two reservoirs. The aim of this study is to explore the impact of the capacity and location of a single reservoir on runoff processes, and the effect of the relative location in the case of joint operation of reservoirs. The results show that: (1) the reservoir has a delay and reduction effect on the flood during the flood season, and has a replenishment effect on the runoff during the dry season; (2) the impact of the reservoir on runoff processes and changes in runoff distribution during the year increases with the reservoir capacity; (3) the mitigation of flooding is more obvious at the river basin outlet control station when the reservoir is further downstream; (4) an arrangement with the smaller reservoir located upstream and the larger reservoir located downstream can maximize the benefits of the reservoirs in flood control.

The China Meteorological Assimilation Driving Datasets for the Soil and Water Assessment Tool model (CMADS) have been widely applied in recent years because of their accuracy. An evaluation of the accuracy and efficiency of the Soil and Water Assessment Tool (SWAT) model and CMADS for simulating hydrological processes in the fan-shaped Lijiang River Basin, China, was carried out. The Sequential Uncertainty Fitting (SUFI-2) algorithm was used for parameter sensitivity and uncertainty analysis at the daily scale. The pair-wise correlation between parameters and the uncertainties associated with equifinality in model parameter estimation were investigated. The results showed that the SWAT model performed well in predicting daily streamflow for the calibration period (2009–2010). The correlation coefficient (R²) was 0.92, and the Nash-Sutcliffe model efficiency coefficient (NSE) was 0.89. For the validation period (2011–2018), R² = 0.89, NSE = 0.88, and reasonable values for the P-factor, R-factor, and percent bias (PBIAS) were obtained. In addition, the spatial and temporal variation of evapotranspiration (ET), surface runoff, and groundwater discharge were analyzed. The results clearly showed that spatial variation in surface runoff and groundwater discharge are strongly related to precipitation, while ET is largely controlled by land use types. The contributions to the water budget by surface runoff, groundwater discharge, and lateral flow were very different in flood years and dry years.

Existing water gauge reading approaches based on image analysis have problems such as poor scene adaptability and weak robustness. Here, we proposed a novel water level measurement method based on deep learning (YOLOv5s, convolutional neural network) to overcome these problems. The proposed method uses the YOLOv5s to extract the water gauge area and all scale character areas in the original video image, uses image processing technology to identify the position of the water surface line, and then calculates the actual water level elevation. The proposed method is validated with a video monitoring station on a river in Beijing, and the results show that the systematic error of the proposed method is only 7.7 mm, the error is within 1 cm/the error is between 1 cm and 3 cm, and the proportion of the number of images is 95%/5% (daylight), 98%/2% (infrared lighting at night), 97%/2% (strong light), 45%/44% (transparent water body), 91%/9% (rainfall), and 90%/10% (water gauge is slightly dirty). The results demonstrate that the proposed method shows good performance in different scenes, and its effectiveness has been confirmed. At the same time, it has a strong robustness and provides a certain reference for the application of deep learning in the field of hydrological monitoring.

Floating debris has a negative impact on the quality of the water as well as the aesthetics of surface waters. Traditional image processing techniques struggle to adapt to the complexity of water due to factors such as complex lighting conditions, significant scale disparities between far and near objects, and the abundance of small-scale floating debris in real existence. This makes the detection of floating debris extremely difficult. This study proposed a brand-new, effective floating debris detection approach based on YOLOv5. Specifically, the coordinate attention module is added into the YOLOv5 backbone network to help the model detect and recognize objects of interest more precisely so that feature information of small-sized and dense floating debris may be efficiently extracted. The previous feature pyramid network, on the other hand, summarizes the input features without taking into account their individual importance when fusing features. To address this issue, the YOLOv5 feature pyramidal network is changed to a bidirectional feature pyramid network with effective bidirectional cross-scale connection and weighted feature fusion, which enhances the model’s performance in terms of feature extraction. The method has been evaluated using a dataset of floating debris that we built ourselves (SWFD). Experiments show that the proposed method detects floating objects more precisely than earlier methods.

Inherent spectral characteristics of hyperspectral image (HSI) data are determined and need to be deeply mined. A convolution neural network (CNN) model of two-dimensional spectrum (2D spectrum) is proposed based on the advantages of deep learning to extract feature and classify HSI. First of all, the traditional data processing methods which use small area pixel block or one-dimensional spectral vector as input unit bring many heterogeneous noises. The 2D-spectrum image method is proposed to solve the problem and make full use of spectral value and spatial information. Furthermore, a batch normalization algorithm (BN) is introduced to address internal covariate shifts caused by changes in the distribution of input data and expedite the training of the network. Finally, Softmax loss models are proposed to induce competition among the outputs and improve the performance of the CNN model. The HSI datasets of experiments include Indian Pines, Salinas, Kennedy Space Center (KSC), and Botswana. Experimental results show that the overall accuracies of the 2D-spectrum CNN model can reach 98.26%, 97.28%, 96.22%, and 93.64%. These results are higher than the accuracies of other traditional methods described in this paper. The proposed model can achieve high target classification accuracy and efficiency.