Search

In climate models, subgrid parameterizations of convection and clouds are one of the main causes of the biases in precipitation and atmospheric circulation simulations. In recent years, due to the rapid development of data science, machine learning (ML) parameterizations for convection and clouds have been demonstrated to have the potential to perform better than conventional parameterizations. Most previous studies were conducted on aqua-planet and idealized models, and the problems of simulation instability and climate drift still exist. Developing an ML parameterization scheme remains a challenging task in realistically configured models. In this paper, a set of residual deep neural networks (ResDNNs) with a strong nonlinear fitting ability is designed to emulate a super-parameterization (SP) with different outputs in a hybrid ML–physical general circulation model (GCM). It can sustain stable simulations for over 10 years under real-world geographical boundary conditions. We explore the relationship between the accuracy and stability by validating multiple deep neural network (DNN) and ResDNN sets in prognostic runs. In addition, there are significant differences in the prognostic results of the stable ResDNN sets. Therefore, trial and error is used to acquire the optimal ResDNN set for both high skill and long-term stability, which we name the neural network (NN) parameterization. In offline validation, the neural network parameterization can emulate the SP in mid- to high-latitude regions with a high accuracy. However, its prediction skill over tropical ocean areas still needs improvement. In the multi-year prognostic test, the hybrid ML–physical GCM simulates the tropical precipitation well over land and significantly improves the frequency of the precipitation extremes, which are vastly underestimated in the Community Atmospheric Model version 5 (CAM5), with a horizontal resolution of 1.9 ∘ × 2.5 ∘. Furthermore, the hybrid ML–physical GCM simulates the robust signal of the Madden–Julian oscillation with a more reasonable propagation speed than CAM5. However, there are still substantial biases with the hybrid ML–physical GCM in the mean states, including the temperature field in the tropopause and at high latitudes and the precipitation over tropical oceanic regions, which are larger than those in CAM5. This study is a pioneer in achieving multi-year stable climate simulations using a hybrid ML–physical GCM under actual land–ocean boundary conditions that become sustained over 30 times faster than the target SP. It demonstrates the emerging potential of using ML parameterizations in climate simulations. [ABSTRACT FROM AUTHOR]

The purpose of this study is to describe a new algorithm based on a neural network approach (Passive microwave Neural network Precipitation Retrieval – PNPR) for precipitation rate estimation from AMSU/MHS observations, and to provide examples of its performance for specific case studies over the European/Mediterranean area. The algorithm optimally exploits the different characteristics of Advanced Microwave Sounding Unit-A (AMSU-A) and the Microwave Humidity Sounder (MHS) channels, and their combinations, including the brightness temperature (TB) differences of the 183.31 channels, with the goal of having a single neural network for different types of background surfaces (vegetated land, snow-covered surface, coast and ocean). The training of the neural network is based on the use of a cloudradiation database, built from cloud-resolving model simulations coupled to a radiative transfer model, representative of the European and Mediterranean Basin precipitation climatology. The algorithm provides also the phase of the precipitation and a pixel-based confidence index for the evaluation of the reliability of the retrieval. Applied to different weather conditions in Europe, the algorithm shows good performance both in the identification of precipitation areas and in the retrieval of precipitation, which is particularly valuable over the extremely variable environmental and meteorological conditions of the region. The PNPR is particularly efficient in (1) screening and retrieval of precipitation over different background surfaces; (2) identification and retrieval of heavy rain for convective events; and (3) identification of precipitation over a cold/iced background, with increased uncertainties affecting light precipitation. In this paper, examples of good agreement of precipitation pattern and intensity with ground-based data (radar and rain gauges) are provided for four different case studies. The algorithm has been developed in order to be easily tailored to new radiometers as they become available (such as the cross-track scanning Suomi National Polar-orbiting Partnership (NPP) Advanced Technology Microwave Sounder (ATMS)), and it is suitable for operational use as it is computationally very efficient. PNPR has been recently extended for applications to the regions of Africa and the South Atlantic, and an extended validation over these regions (using 2 yr of data acquired by the Tropical Rainfall Measuring Mission precipitation radar for comparison) is the subject of a paper in preparation. The PNPR is currently used operationally within the EUMETSAT Hydrology Satellite Application Facility (H-SAF) to provide instantaneous precipitation from passive microwave cross-track scanning radiometers. It undergoes routinely thorough extensive validation over Europe carried out by the H-SAF Precipitation Products Validation Team. [ABSTRACT FROM AUTHOR]

This paper investigates the skill of 90-day low-flow forecasts using two conceptual hydrological models and one data-driven model based on Artificial Neural Networks (ANNs) for the Moselle River. The three models, i.e. HBV, GR4J and ANN-Ensemble (ANN-E), all use forecasted meteorological inputs (precipitation P and potential evapotranspiration PET), whereby we employ ensemble seasonal meteorological forecasts. We compared low-flow forecasts for five different cases of seasonal meteorological forcing: (1) ensemble P and PET forecasts; (2) ensemble P forecasts and observed climate mean PET; (3) observed climate mean P and ensemble PET forecasts; (4) observed climate mean P and PET and (5) zero P and ensemble PET forecasts as input for the models. The ensemble P and PET forecasts, each consisting of 40 members, reveal the forecast ranges due to the model inputs. The five cases are compared for a lead time of 90 days based on model output ranges, whereas the models are compared based on their skill of low-flow forecasts for varying lead times up to 90 days. Before forecasting, the hydrological models are calibrated and validated for a period of 30 and 20 years respectively. The smallest difference between calibration and validation performance is found for HBV, whereas the largest difference is found for ANN-E. From the results, it appears that all models are prone to over-predict runoff during low-flow periods using ensemble seasonal meteorological forcing. The largest range for 90-day low-flow forecasts is found for the GR4J model when using ensemble seasonal meteorological forecasts as input. GR4J, HBV and ANN-E under-predicted 90-day-ahead low flows in the very dry year 2003 without precipitation data. The results of the comparison of forecast skills with varying lead times show that GR4J is less skilful than ANN-E and HBV. Overall, the uncertainty from ensemble P forecasts has a larger effect on seasonal low-flow forecasts than the uncertainty from ensemble PET forecasts and initial model conditions. [ABSTRACT FROM AUTHOR]

Rainfall observed on the ground is dependent on the four dimensional structure of precipitation aloft. Scanning radars can observe the four dimensional structure of precipitation. Neural network is a nonparametric method to represent the nonlinear relationship between radar measurements and rainfall rate. The relationship is derived directly from a dataset consisting of radar measurements and rain gauge measurements. The performance of neural network based rainfall estimation is subject to many factors, such as the representativeness and sufficiency of the training dataset, the generalization capability of the network to new data, seasonal changes, and regional changes. Improving the performance of the neural network for real time applications is of great interest. The goal of this paper is to investigate the performance of rainfall estimation based on Radial Basis Function (RBF) neural networks using radar reflectivity as input and rain gauge as the target. Data from Melbourne, Florida NEXRAD (Next Generation Weather Radar) ground radar (KMLB) over different years along with rain gauge measurements are used to conduct various investigations related to this problem. A direct gauge comparison study is done to demonstrate the improvement brought in by the neural networks and to show the feasibility of this system. The principal components analysis (PCA) technique is also used to reduce the dimensionality of the training dataset. Reducing the dimensionality of the input training data will reduce the training time as well as reduce the network complexity which will also avoid over fitting. [ABSTRACT FROM AUTHOR]

We investigate techniques for using deep neural networks to produce surrogate models for short-term climate forecasts. A convolutional neural network is trained on 97 years of monthly precipitation output from the 1pctCO2 run (the CO2 concentration increases by 1 % per year) simulated by the second-generation Canadian Earth System Model (CanESM2). The neural network clearly outperforms a persistence forecast and does not show substantially degraded performance even when the forecast length is extended to 120 months. The model is prone to underpredicting precipitation in areas characterized by intense precipitation events. Scheduled sampling (forcing the model to gradually use its own past predictions rather than ground truth) is essential for avoiding amplification of early forecasting errors. However, the use of scheduled sampling also necessitates preforecasting (generating forecasts prior to the first forecast date) to obtain adequate performance for the first few prediction time steps. We document the training procedures and hyperparameter optimization process for researchers who wish to extend the use of neural networks in developing surrogate models. [ABSTRACT FROM AUTHOR]

Precipitation estimation with high-accuracy and spatiotemporal resolution is the key for distributed hydrological modelling. Gauge-satellite precipitation merging provides an effect way to improve the spatial precipitation estimation. In spite of the encouraging outcomes of gauge-satellite merging, it is quite a challenging job to derive precipitation appropriate for distributed watershed hydrological modeling when facing the following problems: (1) the spatial resolution of satellite-based precipitation product is much coarser than the modeled one; (2) the spatial distribution information of daily/sub-daily precipitation input to distributed hydrological models cannot be well captured by a single satellite-based precipitation product. In allusion to these problems, a two-step strategy for multi-source information merging was proposed in this paper to improve the spatial resolution and accuracy of daily precipitation estimates. The strategy is based on geographically weighted ridge regression kriging (GWRRK) with emphasis on dealing with non-stationarity and local collinearity in the regression context. Firstly, several individual satellite-based precipitation products, e.g., Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Aanlysis (TMPA) 3B42RT, Climate Prediction Center MORPHing technique (CMORPH), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), and Global Satellite Mapping of Precipitation (GSMaP), were downscaled to 1km by employing the relationships between precipitation and other environmental factors such as topography and vegetation. And then, all these downscaled products were merged with gauge observations. Considering the strongly skewed distribution of daily precipitation, Box-Cox transformation was carried out prior to the GWRRK approach to give a more normal distribution. To our knowledge, this is the first research to downscale satellite-based precipitation product at daily scale using geographically weighted regression method and to merge gauge observations and multiple satellite-based precipitation products. The two-step merging strategy was applied to an experiment conducted in Xijiang Basin of China from 2010 to 2017. Results showed that: (1) the regression relationships in spatial downscaling and gauge-multisatellites merging were strongly spatially non-stationary; (2) local collinearity existed in both steps: this problem was not serious in spatial downscaling but serious enough to be the main barrier for the application of geographically weighted (GWR) in gauge-multisatellites merging; (3) the spatial trend of the downscaled data was consistent with that of the original data for all the four satellite-based daily precipitation products; (4) the merging of gauge observations and multiple satellite-based precipitation products provided more accurate estimates than the merging of gauge observations and a single satellite-based precipitation product. The GWRRK-based two-step merging strategy presented in this study has provided an efficient way to derive daily precipitation with high-accuracy and spatiotemporal resolution. [ABSTRACT FROM AUTHOR]

This study focuses on the Iberian Peninsula and investigates the propagation of precipitation uncertainty, and its interaction with hydrologic modeling, in global water resource reanalysis. Analysis is based on ensemble hydrologic simulations for a period spanning 11 years (2000–2010). To simulate the hydrological variables of surface runoff, subsurface runoff, and evapotranspiration, we used four land surface models (LSMs) – JULES (Joint UK Land Environment Simulator), ORCHIDEE (Organising Carbon and Hydrology In Dynamic Ecosystems), SURFEX (Surface Externalisée), and HTESSEL (Hydrology – Tiled European Centre for Medium-Range Weather Forecasts – ECMWF – Scheme for Surface Exchanges over Land) – and one global hydrological model, WaterGAP3 (Water – a Global Assessment and Prognosis). Simulations were carried out for five precipitation products – CMORPH (the Climate Prediction Center Morphing technique of the National Oceanic and Atmospheric Administration, or NOAA), PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks), 3B42V(7), ECMWF reanalysis, and a machine-learning-based blended product. As a reference, we used a ground-based observation-driven precipitation dataset, named SAFRAN, available at 5 km, 1 h resolution. We present relative performances of hydrologic variables for the different multi-model and multi-forcing scenarios. Overall, results reveal the complexity of the interaction between precipitation characteristics and different modeling schemes and show that uncertainties in the model simulations are attributed to both uncertainty in precipitation forcing and the model structure. Surface runoff is strongly sensitive to precipitation uncertainty, and the degree of sensitivity depends significantly on the runoff generation scheme of each model examined. Evapotranspiration fluxes are comparatively less sensitive for this study region. Finally, our results suggest that there is no single model–forcing combination that can outperform all others consistently for all variables examined and thus reinforce the fact that there are significant benefits to exploring different model structures as part of the overall modeling approaches used for water resource applications. [ABSTRACT FROM AUTHOR]

Over the past 2 decades, a wide range of studies have incorporated Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) products. Currently, PERSIANN offers several precipitation products based on different algorithms available at various spatial and temporal scales, namely PERSIANN, PERSIANN-CCS, and PERSIANN-CDR. The goal of this article is to first provide an overview of the available PERSIANN precipitation retrieval algorithms and their differences. Secondly, we offer an evaluation of the available operational products over the contiguous US (CONUS) at different spatial and temporal scales using Climate Prediction Center (CPC) unified gauge-based analysis as a benchmark. Due to limitations of the baseline dataset (CPC), daily scale is the finest temporal scale used for the evaluation over CONUS. Additionally, we provide a comparison of the available products at a quasi-global scale. Finally, we highlight the strengths and limitations of the PERSIANN products and briefly discuss expected future developments. [ABSTRACT FROM AUTHOR]

A neural-network-based retrieval method to determine the snow ice water path (SIWP), liquid water path (LWP), and integrated water vapor (IWV) from millimeter and submillimeter brightness temperatures, measured by using airborne radiometers (ISMAR and MARSS), is presented. The neural networks were trained by using atmospheric profiles from the ICON numerical weather prediction (NWP) model and by radiative transfer simulations using the Atmospheric Radiative Transfer Simulator (ARTS). The basic performance of the retrieval method was analyzed in terms of offset (bias) and the median fractional error (MFE), and the benefit of using submillimeter channels was studied in comparison to pure microwave retrievals. The retrieval is offset-free for SIWP  > 0.01 kg m−2, LWP  > 0.1 kg m−2, and IWV  > 3 kg m−2. The MFE of SIWP decreases from 100 % at SIWP  =  0.01 kg m−2 to 20 % at SIWP  =  1 kg m−2 and the MFE of LWP from 100 % at LWP  = 0.05 kg m−2 to 30 % at LWP  =  1 kg m−2. The MFE of IWV for IWV  > 3 kg m−2 is 5 to 8 %. The SIWP retrieval strongly benefits from submillimeter channels, which reduce the MFE by a factor of 2, compared to pure microwave retrievals. The IWV and the LWP retrievals also benefit from submillimeter channels, albeit to a lesser degree. The retrieval was applied to ISMAR and MARSS brightness temperatures from FAAM flight B897 on 18 March 2015 of a precipitating frontal system west of the coast of Iceland. Considering the given uncertainties, the retrieval is in reasonable agreement with the SIWP, LWP, and IWV values simulated by the ICON NWP model for that flight. A comparison of the retrieved IWV with IWV from 12 dropsonde measurements shows an offset of 0.5 kg m−2 and an RMS difference of 0.8 kg m−2, showing that the retrieval of IWV is highly effective even under cloudy conditions. [ABSTRACT FROM AUTHOR]

High-resolution satellite precipitation products are very attractive for studying the hydrologic processes in mountainous areas where rain gauges are generally sparse. Four high-resolution satellite precipitation products are evaluated using gauge measurements over different climate zones of the Tibetan Plateau (TP) within a 6 yr period from 2004 to 2009. The four satellite-based precipitation data sets are: Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis 3B42 version 6 (TMPA) and its Real Time version (TMPART), Climate Prediction Center Morphing Technique (CMOPRH) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network (PERSIANN). TMPA and CMORPH, with higher correlation coefficients and lower root mean square errors (RMSEs), show overall better performance than PERSIANN and TMPART. TMPA has the lowest biases among the four precipitation data sets, which is likely due to the correction process against the monthly gauge observations from global precipitation climatology project (GPCP). TMPA also shows large improvement over TMPART, indicating the importance of gauge-based correction on accuracy of rainfall. The four products show better agreement with gauge measurements over humid regions than that over arid regions where correlation coefficients are less than 0.5. Moreover, the four precipitation products generally tend to overestimate light rainfall (0-10 mm) and underestimate moderate and heavy rainfall (>10 mm). Moreover, this study extracts 24 topographic variables from a DEM (digital elevation model) and uses a linear regression model to explore the bias--topography relationship. Results show that biases of TMPA and CMORPH present weak dependence on topography. However, biases of TMPART and PERSIANN present dependence on topography and variability of elevation and surface roughness plays important roles in explaining their biases. [ABSTRACT FROM AUTHOR]

This work addresses the issue of better considering the heterogeneity of precipitation fields within lumped rainfall-runoff models where only areal mean precipitation is usually used as an input. A method using a Kohonen neural network is proposed for the clustering of precipitation fields. The evaluation and improvement of the performance of a lumped rainfall-runoff model for one-day ahead predictions is then established based on this clustering. Multilayer perceptron neural networks are employed as lumped rainfall-runoff models. The Bas-en-Basset watershed in France, which is equipped with 23 rain gauges with data for a 21-year period, is employed as the application case. The results demonstrate the relevance of the proposed clustering method, which produces groups of precipitation fields that are in agreement with the global climatological features affecting the region, as well as with the topographic constraints of the watershed (i.e., orography). The strengths and weaknesses of the rainfall-runoff models are highlighted by the analysis of their performance vis-à-vis the clustering of precipitation fields. The results also show the capability of multilayer perceptron neural networks to account for the heterogeneity of precipitation, even when built as lumped rainfall-runoff models. [ABSTRACT FROM AUTHOR]