Elsevier

Journal of Hydrology

Volume 522, March 2015, Pages 35-48
Journal of Hydrology

Differences in scale-dependent, climatological variation of mean areal precipitation based on satellite and radar-gauge observations

https://doi.org/10.1016/j.jhydrol.2014.11.077Get rights and content

Highlights

  • We compare scale-dependent variability of mean areal precipitation from CMORPH.

  • We analyze the S–R differences on the basis of fractional coverage and correlograms.

  • We found that CMORPH tends to overestimate the critical window size.

  • Underdetection/false alarm by CMORPH contributes to this overestimation.

Summary

This study compares the scale-dependent variation in hourly Mean Areal Precipitation (MAP) derived from a satellite (S) and a radar-gauge (R) Quantitative Precipitation Estimate (QPE), and seeks to explain the S–R differences on the basis of errors in the satellite QPE. This study employs an analytical framework to estimate the coefficient of variation (CV) of MAP for window sizes ranging from 4 km to 512 km, using the rainfall fields of the CPC MORPHing (CMORPH) satellite QPE and a radar-gauge Multisensor QPE (MQPE) over five domains centered in Texas, Oklahoma and New Mexico. CV values based on the analytical framework are first corroborated using empirical estimates. Then, S–R differences in CV are analyzed to determine the contributions of the S–R differences from empirical fractional coverage (FC) and spatial correlograms. Subsequently, sensitivity analyses are performed to isolate the impacts of false detections and long-term, magnitude-dependent bias in CMORPH on the inaccuracies in FC and correlograms. The results are stratified by domain and season (winter and summer) to highlight the impacts of differential accuracy of CMORPH under diverse rainfall regimes. Our analyses reveal that CMORPH-based CV tends to plateau at larger window sizes (referred to as critical window size, or CWS), and is broadly higher in magnitude. The mechanisms underlying the CV differences, however, differ between winter and summer. Over the winter, CMORPH suffers from severe underdetection, which yields suppressed FC across window sizes. This underestimation of FC, together with the lack of resolution of internal rainfall structure by CMORPH, leads to an magnification of both CWS and the magnitude of CV. By contrast, over the summer, widespread false detections in CMORPH lead to inflated FC, which tends to suppress CWS but this effect is outweighed by the opposing impacts of inflated outer and inner scales (i.e., distance parameters of indicator and conditional correlograms). Moreover, it is found that introducing false detection to MQPE via a simple expansion scheme is effective in increasing the FC and inner scale in tandem, and that histogram differences are a rather minor contributor to the S–R difference in inner scale. The implications of the findings for disaggregating climate model projection and data fusion are discussed.

Introduction

Satellite-based Quantitative Precipitation Estimates (QPEs), for their wide coverage and spatial continuity, have seen applications in water budget analysis, flood forecasting, soil moisture predictions and hydrologic model calibration for regions where ground sensors are lacking or deemed inadequate (Scofield and Kuligowski, 2003, Su et al., 2008, Tobin and Bennett, 2010, Habib et al., 2012b, Zhang et al., 2013, Wu et al., 2014). Evolving space-borne sensor technology and precipitation estimation techniques promise further refinement in the spatio-temporal resolutions of Satellite-based QPEs (henceforth referred to as SQPEs) and enhancements in their accuracy. For example, the recent launch of the Global Precipitation Measurement (GPM, Kidd and Huffman, 2011) satellite will refine the grid mesh of the multi-satellite products from 1/4 degree to 10 km, and improve their quality through cross-calibration of satellite sensors. Equally notable is that several satellite QPEs (e.g., the Tropical Rainfall Measurement Mission Multisatellite Precipitation Analysis, or TMPA; Huffman et al., 2007) have accumulated relatively long archives (>10 years), making them a potentially viable source of climate information, especially where a long-term, high resolution precipitation archive is absent.

Of the many opportunities presented by the improving quality and the expanding archive of SQPE data sets, yet to receive much attention is the use of SQPE for quantifying the long-term, scale-dependent temporal variability of Mean Areal Precipitation (MAP). Such information has two immediate downstream utilities, namely conditioning downscaled climate model outputs and fusing precipitation products of differing resolutions. Statistical downscaling of climate model projection, as pointed out in Fowler et al. (2007), requires long-term precipitation climatology for which variation on a subgrid scale is an important component (see Wood et al., 2004, Maraun et al., 2010 for related comments). To elaborate, inconsistency between the temporal variability represented by downscaled climate model outputs and that based on observations could hamper the former’s application to the analysis and prediction of floods and droughts. This can be addressed by adjusting the model-derived MAP so to match the variability based on observations. As for data fusion, there is a potential of blending a high resolution SQPE with coarser resolution gauge-only analysis (e.g., the 2 by 2.5 degree National Weather Service Climate Precipitation Center hourly Atlas No. 7; Higgins et al., 2000) to yield a product that combines the strengths of each product. For these applications, SQPE could be used to extrapolate the variance of the coarser spatial resolution product to the finer, target resolutions.

These promising prospects notwithstanding, it needs to be noted that SQPEs remain susceptible to large biases and random errors in spite of their continuing improvements. Habib et al. (2012b), for example, illustrated the large bias in the SQPEs produced via the CPC MORPHing (CMORPH) algorithm, one of the most accurate products. Kuligowski et al. (2013) highlighted the contribution of error in microwave to that of Self-calibrating Multivariate Precipitation Retrieval (SCaMPR) SQPE. Mei et al. (2014) compared multiple SQPE products over mountainous basins and found that none of them was ideal in capturing heavy precipitation events. Despite these and related works dedicated to the evaluation of satellite QPEs (Xie et al., 2007, Sapiano and Arkin, 2009, Yilmaz et al., 2005, Bitew and Gebremichael, 2011, Tobin and Bennett, 2010, Pan et al., 2010, Zhang et al., 2013), it is yet unclear how these systematic and random errors in the SQPEs impact the scale-dependent variability of MAP as portrayed by SQPE. The present study serves precisely this purpose: it offers a dissection of the differences in the coefficient of variation (CV) of MAP derived from a SQPE and a high quality radar-gauge product, and it attempts to link these inaccuracies of the errors in the former product.

The goals of this work are threefold. The first is to compare the scale-dependent CV from CMORPH SQPE and the National Weather Service multisensor QPE that is based on radar and gauge data. Second, the work seeks to isolate the contributions of errors in scale-dependent rainfall intermittency characterized by FC, and those associated with spatial rainfall structure represented by spatial correlograms to the differences in CV based on CMORPH and MQPE (hereinafter referred to S–R difference, with S and R standing for satellite and radar, respectively). Third, the study attempts to quantify the impact of long-term bias and false rainfall detection on the fractional coverage (FC) of positive precipitation and correlograms. A unique aspect of this study lies in the use of an analytic framework developed by Seo and Smith (1996) that allows for a separation of impact of scale-dependent CV from FC and correlograms. Such a breakdown further permits an analysis of the sensitivity of CV to bias and false rainfall detection directly through perturbation experiments. This study complements the aforementioned body of literature on SQPE accuracy by offering new insights into the scale-dependent impact of SQPE inaccuracies on CV of MAP and by addressing potential impact of the inaccuracies on downscaled climate model outputs.

The remainder of the paper is organized as follows. Section 2 reviews the analytical framework of Seo and Smith (1996), and describes the data and methods. Section 3 summarizes the results. Section 4 discusses the results, and Section 5 presents the conclusions and future works.

Section snippets

Analytical framework

In this section, we review the analytical framework devised by Seo and Smith (1996). Detailed derivations are omitted here and interested readers are referred to the original paper.

Let R(x,t) denote the hourly rainfall amount at location x and hour t, where x is the location vector with x=(x1,x2). Let M and Z be the Mean Areal Precipitation (MAP) estimates and the FC of positive precipitation for domain A, respectively; thenM(A,t)=1||A||AR(x,t)dxZ(A,t)=1||A||AIR(x,t)dxwhere ||A|| is the size

Results

In this section, we first present the precipitation climatology as represented by S (CMORPH) and R (MQPE) across the five domains along with R and S-based parameter values. Then we characterize and corroborate the analytical CV estimates, and assess the sensitivity of the CV to the parameters. In the end, we present the impacts of introducing false detection and bias on the related parameters.

Discussion

This study offers a dissection of the difference in CMORPH and MQPE-based CV on the basis of an analytical framework of Seo and Smith (1996). In this section we provide a synthetic view of the relationship between the S–R difference in CV and errors in CMORPH, and discuss their implications for applications such as downscaling climate model projections and devising data fusion strategies.

Conclusions and further work

This study compares the coefficient of variation of mean areal precipitation as represented by CMORPH satellite QPE and a radar-gauge product over five domains in the southern US, and relates the differences to errors in CMORPH. Major findings include the following:

  • 1.

    For a majority of domains, CMORPH-based CV is associated with larger CWS and higher magnitude than that based on MQPE. This behavior is captured by the analytical framework.

  • 2.

    The mechanisms of the S–R difference differ in winter and

Acknowledgments

This work was in part supported by the NOAA US Weather Research Program through the Global Precipitation Measurement (GPM) effort. The work has benefited from discussions with Andy Wood at NCAR, Rob Cifelli at NOAA ESRL, and David Kitzmiller at OHD. We would also like to thank Pingping Xie and Shaorong Wu at NCEP for providing the CMORPH data and related documentations, and Ms. Alexis Cooley for proofreading the manuscript.

References (42)

  • D.J. Seo

    Real-time estimation of rainfall fields using radar rainfall and rain gage data

    J. Hydrol.

    (1998)
  • D.J. Seo

    Real-time estimation of rainfall fields using rain gage data under fractional coverage conditions

    J. Hydrol.

    (1998)
  • D.J. Seo et al.

    Real-time estimation of mean field bias in radar rainfall data

    J. Hydrol.

    (1999)
  • M.L. Baeck et al.

    Climatological analysis of manually digitized radar data for the United States east of the rocky mountains

    Water Resour. Res.

    (1995)
  • M.M. Bitew et al.

    Assessment of satellite rainfall products for streamflow simulation in medium watersheds of the Ethiopian highlands

    Hydrol. Earth Syst. Sci.

    (2011)
  • R.R. Ferraro

    An evaluation of microwave land surface emissivities over the continental United States to benefit GPM-era precipitation algorithms

    IEEE Trans. Geosci. Remote Sens.

    (2012)
  • H.J. Fowler et al.

    Linking climate change modelling to impacts studies: recent advances in downscaling techniques for hydrological modelling

    Int. J. Climatol.

    (2007)
  • R.A. Fulton et al.

    The WSR-88D rainfall algorithm

    Weather Forecast.

    (1998)
  • Greene, D., Hudlow, M., 1982. Hydrometeorologic grid mapping procedures. In: Int. Symp. on Hydrometeorology, AWRA....
  • E. Habib et al.

    Climatology-focused evaluation of CMORPH and TMPA satellite rainfall products over the Nile Basin

    J. Appl. Meteorol. Climatol.

    (2012)
  • E. Habib et al.

    Evaluation of the high-resolution CMORPH satellite-rainfall product using dense rain gauge observations and radar-based estimates

    J. Hydrometeorol.

    (2012)
  • E. Habib et al.

    Independent assessment of incremental complexity in the NWS multi-sensor precipitation estimator algorithms

    J. Hydrol. Eng.

    (2012)
  • Higgins, R.W., Shi, W., Yarosh, E., Joyce, R., 2000. Improved United States Precipitation Quality Control System and...
  • F. Hossain et al.

    A two-dimensional satellite rainfall error model

    IEEE Trans. Geosci. Remote Sens.

    (2006)
  • G.J. Huffman et al.

    The TRMM multisatellite precipitation analysis (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at fine scales

    J. Hydrometeorol.

    (2007)
  • R. Joyce et al.

    CMORPH: a method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution

    J. Hydrometeorol.

    (2004)
  • C. Kidd et al.

    Global precipitation measurement

    Meteorol. Appl.

    (2011)
  • D. Kitzmiller et al.

    Radar and multisensor precipitation estimation techniques in national weather service hydrologic operations

    J. Hydrol. Eng.

    (2013)
  • R. Kuligowski et al.

    Impact of TRMM data on a low-latency, high-resolution precipitation algorithm for flash flood forecasting

    J. Appl. Meteorol. Climatol.

    (2013)
  • L. Luo et al.

    Bayesian merging of multiple climate model forecasts for seasonal hydrological predictions

    J. Geophys. Res.

    (2007)
  • D. Maraun et al.

    Precipitation downscaling under climate change: recent developments to bridge the gap between dynamical models and the end user

    Rev. Geophys.

    (2010)
  • Cited by (4)

    • Improving multisensor estimation of heavy-to-extreme precipitation via conditional bias-penalized optimal estimation

      2018, Journal of Hydrology
      Citation Excerpt :

      Because it is impractical to estimate time-varying covariance structures due to lack of data and large computing requirements, climatological correlograms from the 7-year period of 2002–2008 in Table 1 were used in this study. The use of climatological (warm vs. cool season) correlation structure, as opposed to flow-dependent structure that varies dynamically, renders the OCK and CBPCK estimates less than optimal (Seo and Smith, 1996; Zhang et al., 2016, 2015b). The impact of such suboptimal modeling of correlation structure is generally much larger for unconstrained estimators such as simple kriging or simple cokriging whose weights are not constrained to sum to unity.

    • Evaluation of the TRMM multisatellite precipitation analysis and its applicability in supporting reservoir operation and water resources management in Hanjiang basin, China

      2017, Journal of Hydrology
      Citation Excerpt :

      Most of these satellite precipitation products have been widely used for hydrological studies and applications (Guetter et al., 1996; Grimes and Diop, 2003; Hughes, 2006; Wilk et al., 2006; Li et al., 2009; Islam et al., 2012; Gupta et al., 2013; Chen et al., 2015). However, the errors and uncertainties of satellite-based precipitation products always impose limitations on the full utilization of these products in hydrologic and water resources applications (Su et al., 2008; Srivastava et al., 2014; Zhang et al., 2015). Yilmaz et al. (2005) used the PERSIANN satellite precipitation algorithm to forecast streamflow using a lumped hydrologic model over several medium-size basins in the southeastern United States; they found that it showed poorer performance in the smaller basins than in the larger basins compared to the streamflow gauge observations.

    View full text