Search Results

You are looking at 1 - 10 of 19 items for

  • Author or Editor: Xiaogang Gao x
  • Refine by Access: All Content x
Clear All Modify Search
Xiaogang Gao and Soroosh Sorooshian

Abstract

In the surface hydrologic pararmeterization of general circulation models (GCMs), it is commonly assumed that the precipitation processes are homogeneous over a GCM grid square and that the precipitation intensity is uniformly distributed. Based on evidence that the spatial distribution of precipitation within a GCM grid square is crucial for the land surface hydrology parameterization, a few researchers have explored the impacts of assuming that the precipitation is exponentially distributed. This paper explores the suitability of the afore-mentioned assumptions. First, a statistical analysis is conducted of historical precipitation data for three GCM grids in different regions of the United States. The analysis suggests that neither the uniform nor the exponential distribution assumption may be suitable at the GCM grid scale and, that instead, the spatial variability in precipitation is characterized by statistical patterns that are inhomogeneous. These patterns vary from grid to grid and are induced by the interaction between atmospheric conditions and various land surface characteristics, such as topographical features, surface properties, etc. Within the same grid square, however, the statistical patterns are generally constant from year to year. Based on this analysis, a computationally viable (i.e., usable with GCMs) stochastic precipitation disaggregation scheme that utilizes these stable statistical patterns is proposed. The method was used to generate spatially distributed hourly rainfall for a summer season in the southwestern region of the continental United States. Analysis of the results shows that the methodology preserves the seasonal characteristics of spatial variability in precipitation that is observed in the long-term historical data.

Full access
Scott Lee Sellars, Xiaogang Gao, and Soroosh Sorooshian

Abstract

This manuscript introduces a novel computational science approach for studying the impact of climate variability on precipitation. The approach uses an object-oriented connectivity algorithm that segments gridded near-global satellite precipitation data into four-dimensional (4D) objects (longitude, latitude, time, and intensity). These precipitation systems have distinct spatiotemporal properties that are counted, tracked, described, and stored in a searchable database. A case study of western United States precipitation systems is performed, demonstrating the unique properties and capabilities of this object-oriented database. The precipitation dataset used in the case study is the University of California, Irvine, Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) from 1 March 2000 to 1 January 2011. A search of the database for all western United States precipitation systems during this time period returns 626 precipitation systems as objects. By analyzing these systems as segmented objects, joint interactions of the selected climate phenomena 1) Arctic Oscillation (AO), 2) Madden–Julian oscillation (MJO), and 3) El Niño–Southern Oscillation (ENSO) on precipitation can be shown. They directly show the increased/decreased likelihood of having precipitation systems occurring over the western United States (monthly count) during phases of these climate phenomena. It is found that specific climate phenomena impact the monthly count of the events differently, and that the joint interaction of climate phenomena of the AO–MJO and AO–ENSO is important, especially during certain months of the year. It is also found that these interactions impact the physical features of the precipitation systems themselves.

Full access
Hao Liu, Soroosh Sorooshian, and Xiaogang Gao

Abstract

Studies have been reported about the efficacy of satellites for measuring precipitation and about quantifying their errors. Based on these studies, the errors are associated with a number of factors, among them, intensity, location, climate, and season of the year. Several error models have been proposed to assess the relationship between the error and the rainfall intensity. However, it is unknown whether these models are adaptive to different seasons, different regions, or different types of satellite-based estimates. Therefore, how the error–intensity relationship varies with the season or region is unclear. To investigate these issues, a parametric joint pdf model is proposed to analyze and study the 9-yr satellite-derived precipitation datasets of Climate Prediction Center (CPC) morphing technique (CMORPH); PERSIANN; and the real-time TRMM product 3B42, version 7 (TRMM-3B42-RTV7). The NEXRAD Stage IV product is the ground reference. The adaptability of the proposed model is verified by applying it to three locations (Oklahoma, Montana, and Florida) and by applying it to cold season, warm season, and the entire year. Then, the heteroscedasticities in the errors of satellite-based precipitation measurements are investigated using the proposed model under those scenarios. The results show that the joint pdfs have the same formulation under these scenarios, whereas their parameter sets were adaptively adjusted. This parametric model reveals detailed information about the spatial and seasonal variations of the satellite-based precipitation measurements. It is found that the shape of the conditional pdf shifts across the intensity ranges. At the ~10–20 mm day−1 range, the conditional pdf is L shaped, while at the ~40–60 mm day−1 range, it becomes more bell shaped. It is also concluded that no single satellite-based precipitation product outperforms others with respect to the different scenarios (i.e., seasons, regions, and climates).

Full access
Yang Hong, Kuo-Lin Hsu, Soroosh Sorooshian, and Xiaogang Gao

Abstract

A satellite-based rainfall estimation algorithm, Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) Cloud Classification System (CCS), is described. This algorithm extracts local and regional cloud features from infrared (10.7 μm) geostationary satellite imagery in estimating finescale (0.04° × 0.04° every 30 min) rainfall distribution. This algorithm processes satellite cloud images into pixel rain rates by 1) separating cloud images into distinctive cloud patches; 2) extracting cloud features, including coldness, geometry, and texture; 3) clustering cloud patches into well-organized subgroups; and 4) calibrating cloud-top temperature and rainfall (TbR) relationships for the classified cloud groups using gauge-corrected radar hourly rainfall data. Several cloud-patch categories with unique cloud-patch features and TbR curves were identified and explained. Radar and gauge rainfall measurements were both used to evaluate the PERSIANN CCS rainfall estimates at a range of temporal (hourly and daily) and spatial (0.04°, 0.12°, and 0.25°) scales. Hourly evaluation shows that the correlation coefficient (CC) is 0.45 (0.59) at a 0.04° (0.25°) grid scale. The averaged CC of daily rainfall is 0.57 (0.63) for the winter (summer) season.

Full access
Kou-lin Hsu, Xiaogang Gao, Soroosh Sorooshian, and Hoshin V. Gupta

Abstract

A system for Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) is under development at The University of Arizona. The current core of this system is an adaptive Artificial Neural Network (ANN) model that estimates rainfall rates using infrared satellite imagery and ground-surface information. The model was initially calibrated over the Japanese Islands using remotely sensed infrared data collected by the Geostationary Meteorological Satellite (GMS) and ground-based data collected by the Automated Meteorological Data Acquisition System (AMeDAS). The model was then validated for both the Japanese Islands (using GMS and AMeDAS data) and the Florida peninsula (using GOES-8 and NEXRAD data). An adaptive procedure is used to recursively update the network parameters when ground-based data are available. This feature dramatically improves the estimation performance in response to the diverse precipitation characteristics of different geographical regions and time of year. The model can also be successfully updated using only spatially and/or temporally limited observation data such as ground-based rainfall measurements. Another important feature is a procedure that provides insights into the functional relationships between the input variables and output rainfall rate.

Full access
Do Hyuk Kang, Xiaogang Shi, Huilin Gao, and Stephen J. Déry

Abstract

This paper presents an application of the Variable Infiltration Capacity (VIC) model to the Fraser River basin (FRB) of British Columbia (BC), Canada, over the latter half of the twentieth century. The Fraser River is the longest waterway in BC and supports the world’s most abundant Pacific Ocean salmon populations. Previous modeling and observational studies have demonstrated that the FRB is a snow-dominated system, but with climate change, it may evolve to a pluvial regime. Thus, the goal of this study is to evaluate the changing contribution of snow to the hydrology of the FRB over the latter half of the twentieth century. To this end, a 0.25° atmospheric forcing dataset is used to drive the VIC model from 1949 to 2006 (water years) at a daily time step over a domain covering the entire FRB. A model evaluation is first conducted over 11 major subwatersheds of the FRB to quantitatively assess the spatial variations of snow water equivalent (SWE) and runoff (R). The ratio of the spatially averaged maximum SWE to R (R SR) is used to quantify the contribution of snow to the runoff in the 11 subwatersheds of interest. From 1949 to 2006, R SR exhibits a significant decline in 9 of the 11 subwatersheds (with p < 0.05 according to the Mann–Kendall test statistics). To determine the sensitivity of R SR, the air temperature and precipitation in the forcing dataset are then perturbed. The ratio R SR decreases more significantly, especially during the 1990s and 2000s, when air temperatures have warmed considerably compared to the 1950s. On the other hand, increasing precipitation by a multiplicative factor of 1.1 causes R SR to decrease. As the climate continues to warm, ecological processes and human usage of natural resources in the FRB may be substantially affected by its transition from a snow to a hybrid (nival/pluvial) and even a rain-dominated system.

Full access
Yumeng Tao, Xiaogang Gao, Alexander Ihler, Soroosh Sorooshian, and Kuolin Hsu

Abstract

In the development of a satellite-based precipitation product, two important aspects are sufficient precipitation information in the satellite-input data and proper methodologies, which are used to extract such information and connect it to precipitation estimates. In this study, the effectiveness of the state-of-the-art deep learning (DL) approaches to extract useful features from bispectral satellite information, infrared (IR), and water vapor (WV) channels, and to produce rain/no-rain (R/NR) detection is explored. To verify the methodologies, two models are designed and evaluated: the first model, referred to as the DL-IR only method, applies deep learning approaches to the IR data only; the second model, referred to as the DL-IR+WV method, incorporates WV data to further improve the precipitation identification performance. The radar stage IV data are the reference data used as ground observation. The operational product, Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks–Cloud Classification System (PERSIANN-CCS), serves as a baseline model with which to compare the performances. The experiments show significant improvement for both models in R/NR detection. The overall performance gains in the critical success index (CSI) are 21.60% and 43.66% over the verification periods for the DL-IR only model and the DL-IR+WV model compared to PERSIANN-CCS, respectively. In particular, the performance gains in CSI are as high as 46.51% and 94.57% for the models for the winter season. Moreover, specific case studies show that the deep learning techniques and the WV channel information effectively help recover a large number of missing precipitation pixels under warm clouds while reducing false alarms under cold clouds.

Full access
Liming Xu, Soroosh Sorooshian, Xiaogang Gao, and Hoshin V. Gupta

Abstract

A new cloud-patch method for the identification and removal of no-rain cold clouds from infrared (IR) imagery is presented. A cloud patch is defined as a cluster of connected IR imagery pixels that are colder than a given IR brightness temperature threshold. The threshold is derived through a combination of the rainfall field estimated from microwave observations and the IR data closely coincident with microwave sensor satellite overpasses. Seven cloud-patch features are used to describe cloud-top properties, including six IR based and one VIS based. The ID3 algorithm is used to extract structural knowledge from a training dataset and to produce classification rules expressed explicitly on the values of various patch features; these rules can be used to explain the physical principles underlying the cloud classification. The method was evaluated for the Japanese islands and surrounding oceans using AIP/1 data for June (training period) and July–August (evaluation period) 1989. The results of identifying no-rain cloud patches are very good for both periods in spite of the change in rainfall regime from frontal to subtropical convective. Nearly 20% of the total pixels and 60% of the no-rain cloud pixels were removed with negligible rain losses due to misclassification. Moreover, visible data were found to be useful for enhancing the no-rain cold patch identification and thereby reducing the rain loss.

Full access
Yumeng Tao, Kuolin Hsu, Alexander Ihler, Xiaogang Gao, and Soroosh Sorooshian

Abstract

Compared to ground precipitation measurements, satellite-based precipitation estimation products have the advantage of global coverage and high spatiotemporal resolutions. However, the accuracy of satellite-based precipitation products is still insufficient to serve many weather, climate, and hydrologic applications at high resolutions. In this paper, the authors develop a state-of-the-art deep learning framework for precipitation estimation using bispectral satellite information, infrared (IR), and water vapor (WV) channels. Specifically, a two-stage framework for precipitation estimation from bispectral information is designed, consisting of an initial rain/no-rain (R/NR) binary classification, followed by a second stage estimating the nonzero precipitation amount. In the first stage, the model aims to eliminate the large fraction of NR pixels and to delineate precipitation regions precisely. In the second stage, the model aims to estimate the pointwise precipitation amount accurately while preserving its heavily skewed distribution. Stacked denoising autoencoders (SDAEs), a commonly used deep learning method, are applied in both stages. Performance is evaluated along a number of common performance measures, including both R/NR and real-valued precipitation accuracy, and compared with an operational product, Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks–Cloud Classification System (PERSIANN-CCS). For R/NR binary classification, the proposed two-stage model outperforms PERSIANN-CCS by 32.56% in the critical success index (CSI). For real-valued precipitation estimation, the two-stage model is 23.40% lower in average bias, is 44.52% lower in average mean squared error, and has a 27.21% higher correlation coefficient. Hence, the two-stage deep learning framework has the potential to serve as a more accurate and more reliable satellite-based precipitation estimation product. The authors also provide some future directions for development of satellite-based precipitation estimation products in both incorporating auxiliary information and improving retrieval algorithms.

Full access
Newsha K. Ajami, Qingyun Duan, Xiaogang Gao, and Soroosh Sorooshian

Abstract

This paper examines several multimodel combination techniques that are used for streamflow forecasting: the simple model average (SMA), the multimodel superensemble (MMSE), modified multimodel superensemble (M3SE), and the weighted average method (WAM). These model combination techniques were evaluated using the results from the Distributed Model Intercomparison Project (DMIP), an international project sponsored by the National Weather Service (NWS) Office of Hydrologic Development (OHD). All of the multimodel combination results were obtained using uncalibrated DMIP model simulations and were compared against the best-uncalibrated as well as the best-calibrated individual model results. The purpose of this study is to understand how different combination techniques affect the accuracy levels of the multimodel simulations. This study revealed that the multimodel simulations obtained from uncalibrated single-model simulations are generally better than any single-member model simulations, even the best-calibrated single-model simulations. Furthermore, more sophisticated multimodel combination techniques that incorporated bias correction step work better than simple multimodel average simulations or multimodel simulations without bias correction.

Full access