Model identification for hydrological forecasting under uncertainty

Methods for the identification of models for hydrological forecasting have to consider the specific nature of these models and the uncertainties present in the modeling process. Current approaches fail to fully incorporate these two aspects. In this paper we review the nature of hydrological models and the consequences of this nature for the task of model identification. We then continue to discuss the history (“The need for more POWER‘’), the current state (“Learning from other fields”) and the future (“Towards a general framework”) of model identification. The discussion closes with a list of desirable features for an identification framework under uncertainty and open research questions in need of answers before such a framework can be implemented.     

Stochastic Modelling of the Impact of Flood Protection Measures Along the River Waal in the Netherlands

River flooding is a problem of international interest. In the past few years many countries suffered from severe floods. A large part of the Netherlands is below sea level and river levels. The Dutch flood defences along the river Rhine are designed for water levels with a probability of exceedance of 1/1250 per year. These water levels are computed with a hydrodynamic model using a deterministic bed level and a deterministic design discharge. Traditionally, the safety against flooding in the Netherlands is obtained by building and reinforcing dikes. Recently, a new policy was proposed to cope with increasing design discharges in the Rhine and Meuse rivers. This policy is known as the Room for the River (RfR) policy, in which a reduction of flood levels is achieved by measures creating space for the river, such as dike replacement, side channels and floodplain lowering. As compared with dike reinforcement, these measures may have a stronger impact on flow and sediment transport fields, probably leading to stronger morphological effects. As a result of the latter the flood conveyance capacity may decrease over time. An a priori judgement of safety against flooding on the basis of an increased conveyance capacity of the river can be quite misleading. Therefore, the determination of design water levels using a fixed-bed hydrodynamic model may not be justified and the use of a mobile-bed approach may be more appropriate. This problem is addressed in this paper, using a case study of the river Waal (one of the Rhine branches in the Netherlands). The morphological response of the river Waal to a flood protection measure (floodplain lowering in combination with summer levee removal) is analysed. The effect of this measure is subject to various sources of uncertainty. Monte Carlo simulations are applied to calculate the impact of uncertainties in the river discharge on the bed levels. The impact of the “uncertain” morphological response on design flood level predictions is analysed for three phenomena, viz. the impact of the spatial morphological variation over years, the impact of the seasonal morphological variation and the impact of the morphological variability around bifurcation points. The impact of seasonal morphological variations turns out to be negligible, but the other two phenomena appear to have each an appreciable impact (order of magnitude 0.05–0.1 m) on the computed design water levels. We have to note however, that other sources of uncertainty (e.g. uncertainty in hydraulic roughness predictor), which may be of influence, are not taken into consideration. In fact, the present investigation is limited to the sensitivity of the design water levels to uncertainties in the predicted bed level.     

时间尺度重构EEMD-GRNN改进模型预测PM2.5的研究

合理构建PM_2.5浓度预测模型是科学、准确地预测PM_2.5浓度变化的关键。传统PM_2.5预测EEMD-GRNN模型具有较好的预测精度,但是存在过于关注研究数据本身而忽略其物理意义的不足。本研究基于南京市2014-2017年PM_2.5浓度时间序列数据,分析PM_2.5浓度多尺度变化特征及其对气象因子和大气污染因子的尺度响应,基于时间尺度重构进行EEMD-GRNN模型的改进与实证研究。南京市样本数据PM_2.5浓度变化表现为明显的天际尺度和月际尺度,从重构尺度（天际、月际）构建GRNN模型更具有现实意义;同时,PM_2.5对PM₁₀、NO₂、O₃、RH、MinT等因子存在多尺度响应效应,以其作为GRNN模型中的输入变量更具有时间序列上的解释意义。改进后的EEMD-GRNN模型具有更高的PM_2.5浓度预测精度,MAE、MAPE、RMSE和R²分别为6.17、18.41%、8.32和0.95,而传统EEMD-GRNN模型的模型有效性检验结果分别为8.37、27.56%、11.56、0.91。对于高浓度天（PM_2.5浓度大于100 μg/m³）的预测,改进模型更是全面优于传统EEMD-GRNN模型,MAPE为12.02%,相较于传统模型提高了9.03%。     

Ocean current estimation using a Multi-Model Ensemble Kalman Filter during the Grand Lagrangian Deployment experiment (GLAD)

In the summer and fall of 2012, during the GLAD experiment in the Gulf of Mexico, the Consortium for Advanced Research on Transport of Hydrocarbon in the Environment (CARTHE) used several ocean models to assist the deployment of more than 300 surface drifters. The Navy Coastal Ocean Model (NCOM) at 1 km and 3 km resolutions, the US Navy operational NCOM at 3 km resolution (AMSEAS), and two versions of the Hybrid Coordinates Ocean Model (HYCOM) set at 4 km were running daily and delivering 72-h range forecasts. They all assimilated remote sensing and local profile data but they were not assimilating the drifter’s observations. This work presents a non-intrusive methodology named Multi-Model Ensemble Kalman Filter that allows assimilating the local drifter data into such a set of models, to produce improved ocean currents forecasts. The filter is to be used when several modeling systems or ensembles are available and/or observations are not entirely handled by the operational data assimilation process. It allows using generic in situ measurements over short time windows to improve the predictability of local ocean dynamics and associated high-resolution parameters of interest for which a forward model exists (e.g. oil spill plumes). Results can be used for operational applications or to derive enhanced background fields for other data assimilation systems, thus providing an expedite method to non-intrusively assimilate local observations of variables with complex operators. Results for the GLAD experiment show the method can improve water velocity predictions along the observed drifter trajectories, hence enhancing the skills of the models to predict individual trajectories.     

Efficient estimation of flood forecast prediction intervals via single‐ and multi‐objective versions of the LUBE method

Prediction intervals (PIs) are commonly used to quantify the accuracy and precision of a forecast. However, traditional ways to construct PIs typically require strong assumptions about data distribution and involve a large computational burden. Here, we improve upon the recent proposed Lower Upper Bound Estimation method and extend it to a multi‐objective framework. The proposed methods are demonstrated using a real‐world flood forecasting case study for the upper Yangtze River Watershed. Results indicate that the proposed methods are able to efficiently construct appropriate PIs, while outperforming other methods including the widely used Generalized Likelihood Uncertainty Estimation approach. Copyright © 2016 John Wiley & Sons, Ltd.     

HHT-EEMD用于IGS站高程时间序列分析

利用经验模态分解（EMD）和整体经验模态分解（EEMD）方法，将BJFS站2000~2015年高程时间序列进行分解，发现其不仅存在1 a、0.5 a、0.25 a、2个月、1个月以及长周期等周期项，还存在以往方法很难探测出来的近似2 a周期信号。与EMD分解结果对比，整体经验模态分解可以明显减弱模式混叠现象。对各分量进行Hilbert 变换（HHT），得到时间-频率-能量的Hilbert 频谱图。结果表明，年周期和2 a周期变化是高程运动的主要贡献项。利用小波变换方法对比验证EEMD的分解结果表明，与小波分析相比，EEMD重构信号与高程序列差异的RMS更小，证明了HHT-EEMD方法在数据资料分析过程中的有效性。对环境负载及GRACE负载造成的测站位移进行功率谱分析得出，环境负载确实会造成IGS站高程时间序列的1 a、0.5 a以及季节性运动，GRACE负载还验证了2 a信号的存在。     

保持局域特征的多源海冰图像融合与分类

本文提出一种保持局域特征的多源海冰图像融合方法，并在此基础上进行海冰分类。本文提出的多源海冰图像融合方法包括保持空间局域融合和保持特征局域融合两方面。首先，通过学习得到投影矩阵和相似矩阵。投影矩阵将多源像素进行投影变换，得到保留像素空间局域特性的融合向量。相似矩阵度量像素特征间的相似性，通过拉普拉斯特征分解，得到保留像素特征局域相似性的融合向量。然后，将空间融合向量和特征融合向量进行像素综合，得到融合图像。在此基础上，本文设计一种滑动集成分类方法进行融合图像像素分类。提出的分类方法利用滑动集成的特点，在分类时增强刻画了海冰局域特性。由于本文的保持局域融合框架不仅刻画了海冰在物理空间中的邻接关系，而且考虑不同海冰类型的特征关系，因此其在多源图像（多光谱和合成孔径雷达（SAR）图像）的海冰分类任务中表现优异。实验结果表明本文提出的基于保持局域特征融合的多源海冰图像分类方法有效提升了海冰分类精度。     

The data‐driven approach as an operational real‐time flood forecasting model

Accurate water level forecasts are essential for flood warning. This study adopts a data‐driven approach based on the adaptive network–based fuzzy inference system (ANFIS) to forecast the daily water levels of the Lower Mekong River at Pakse, Lao People's Democratic Republic. ANFIS is a hybrid system combining fuzzy inference system and artificial neural networks. Five ANFIS models were developed to provide water level forecasts from 1 to 5 days ahead, respectively. The results show that although ANFIS forecasts of water levels up to three lead days satisfied the benchmark, four‐ and five‐lead‐day forecasts were only slightly better in performance compared with the currently adopted operational model. This limitation is imposed by the auto‐ and cross‐correlations of the water level time series. Output updating procedures based on the autoregressive (AR) and recursive AR (RAR) models were used to enhance ANFIS model outputs. The RAR model performed better than the AR model. In addition, a partial recursive procedure that reduced the number of recursive steps when applying the AR or the RAR model for multi‐step‐ahead error prediction was superior to the fully recursive procedure. The RAR‐based partial recursive updating procedure significantly improved three‐, four‐ and five‐lead‐day forecasts. Our study further shows that for long lead times, ANFIS model errors are dominated by lag time errors. Although the ANFIS model with the RAR‐based partial recursive updating procedure provided the best results, this method was able to reduce the lag time errors significantly for the falling limbs only. Improvements for the rising limbs were modest. Copyright © 2011 John Wiley & Sons, Ltd.     

Ensemble simulation of land evapotranspiration in China based on a multi-forcing and multi-model approach

In order to reduce the uncertainty of offline land surface model(LSM) simulations of land evapotranspiration(ET), we used ensemble simulations based on three meteorological forcing datasets [Princeton, ITPCAS(Institute of Tibetan Plateau Research, Chinese Academy of Sciences), Qian] and four LSMs(BATS, VIC, CLM3.0 and CLM3.5), to explore the trends and spatiotemporal characteristics of ET, as well as the spatiotemporal pattern of ET in response to climate factors over mainland China during 1982–2007. The results showed that various simulations of each member and their arithmetic mean(Ens Mean) could capture the spatial distribution and seasonal pattern of ET sufficiently well, where they exhibited more significant spatial and seasonal variation in the ET compared with observation-based ET estimates(Obs MTE). For the mean annual ET, we found that the BATS forced by Princeton forcing overestimated the annual mean ET compared with Obs MTE for most of the basins in China, whereas the VIC forced by Princeton forcing showed underestimations. By contrast, the Ens Mean was closer to Obs MTE, although the results were underestimated over Southeast China. Furthermore, both the Obs MTE and Ens Mean exhibited a significant increasing trend during 1982–98; whereas after 1998, when the last big EI Ni ?no event occurred, the Ens Mean tended to decrease significantly between 1999 and 2007, although the change was not significant for Obs MTE. Changes in air temperature and shortwave radiation played key roles in the long-term variation in ET over the humid area of China, but precipitation mainly controlled the long-term variation in ET in arid and semi-arid areas of China.     

A timescale decomposed threshold regression downscaling approach to forecasting South China early summer rainfall

A timescale decomposed threshold regression(TSDTR) downscaling approach to forecasting South China early summer rainfall(SCESR) is described by using long-term observed station rainfall data and NOAA ERSST data. It makes use of two distinct regression downscaling models corresponding to the interannual and interdecadal rainfall variability of SCESR.The two models are developed based on the partial least squares(PLS) regression technique, linking SCESR to SST modes in preceding months on both interannual and interdecadal timescales. Specifically, using the datasets in the calibration period 1915–84, the variability of SCESR and SST are decomposed into interannual and interdecadal components. On the interannual timescale, a threshold PLS regression model is fitted to interannual components of SCESR and March SST patterns by taking account of the modulation of negative and positive phases of the Pacific Decadal Oscillation(PDO). On the interdecadal timescale, a standard PLS regression model is fitted to the relationship between SCESR and preceding November SST patterns. The total rainfall prediction is obtained by the sum of the outputs from both the interannual and interdecadal models. Results show that the TSDTR downscaling approach achieves reasonable skill in predicting the observed rainfall in the validation period 1985–2006, compared to other simpler approaches. This study suggests that the TSDTR approach,considering different interannual SCESR-SST relationships under the modulation of PDO phases, as well as the interdecadal variability of SCESR associated with SST patterns, may provide a new perspective to improve climate predictions.