Fig 1 - uploaded by Hyun-Jong Oh
Content may be subject to copyright.
Depiction of the Automated Weather Station operational rain gauge network operated by the Korean Meteorological Administration. The density is nearly homogeneous across the southern Korea peninsula with approximately 40 gauges per 1° box, each reporting at a 1-min time update
Source publication
In order to properly utilize remotely sensed precipitation estimates in hydrometeorological applications, knowledge of the
accuracy of the estimates are needed. However, relatively few ground validation networks operate with the necessary spatial
density and time-resolution required for validation of high-resolution precipitation products (HRPP) ge...
Contexts in source publication
Context 1
... KMA maintains an operational, densely spaced AWS network over the southern Korean Peninsula, consisting of nearly 500 tipping-buckets, uniformly spaced, 1-min updating raingauges (approximately 40 gauges per 1° box). Figure 1 depicts the AWS grid (not all stations are shown). AWS data were collected during June to August 2000 along with the individual hourly, instantaneous rainfall datasets produced by the blended satellite technique (the GMS-5 satellite was the operational geostationary satellite during this time and its routine refresh rate was hourly beginning at 30 min after each hour, and the Korean Peninsula was imaged about 8 min after the frame start time). ...
Context 2
... not yet incorporated into the NRL-Blend technique, giving a LEO revisit over Korea of about 4 h on average and 10 h worst-case. Considering that an individual satellite observation represents a 0.1° (approximately 10 km 9 10 km area) area average whereas the gauge measurement represents a small area less than 1 m 2 , South Korea is divided into smaller boxes, ranging from 0.1° to 3° on a side, where relatively homogeneous gauge distri- butions are found ( Fig. 1 depicts the 1° box sizes). Because of the inhomogeneity of the rain within the spatial aver- aging box and very small areas represented by individual gauges, a direct comparison of instantaneous (i.e., sensor scan level) satellite-based retrievals and gauges is inher- ently limited. ...
Citations
... At the short periods (,4 h) the average phase is, however, slightly higher than 0, which reveals that when IMERG does not perfectly capture the timing of precipitation it is more likely to detect precipitation a few minutes early than late. This tendency has been documented for several satellite QPEs (Turk et al. 2009;Utsumi et al. 2019) and can be explained by the falling time of the hydrometeors between the upper cloud layers to which passive spaceborne radiometric measurements are sensitive and the near-surface gauge-radar measurements. ...
Satellite precipitation products, as all quantitative estimates, come with some inherent degree of uncertainty. To associate a quantitative value of the uncertainty to each individual estimate, error modeling is necessary. Most of the error models proposed so far compute the uncertainty as a function of precipitation intensity only, and only at one specific spatiotemporal scale. We propose a spectral error model that accounts for the neighboring space–time dynamics of precipitation into the uncertainty quantification. Systematic distortions of the precipitation signal and random errors are characterized distinctively in every frequency–wavenumber band in the Fourier domain, to accurately characterize error across scales. The systematic distortions are represented as a deterministic space–time linear filtering term. The random errors are represented as a nonstationary additive noise. The spectral error model is applied to the IMERG multisatellite precipitation product, and its parameters are estimated empirically through a system identification approach using the GV-MRMS gauge–radar measurements as reference (“truth”) over the eastern United States. The filtering term is found to be essentially low-pass (attenuating the fine-scale variability). While traditional error models attribute most of the error variance to random errors, it is found here that the systematic filtering term explains 48% of the error variance at the native resolution of IMERG. This fact confirms that, at high resolution, filtering effects in satellite precipitation products cannot be ignored, and that the error cannot be represented as a purely random additive or multiplicative term. An important consequence is that precipitation estimates derived from different sources shall not be expected to automatically have statistically independent errors.
Significance Statement
Satellite precipitation products are nowadays widely used for climate and environmental research, water management, risk analysis, and decision support at the local, regional, and global scales. For all these applications, knowledge about the accuracy of the products is critical for their usability. However, products are not systematically provided with a quantitative measure of the uncertainty associated with each individual estimate. Various parametric error models have been proposed for uncertainty quantification, mostly assuming that the uncertainty is only a function of the precipitation intensity at the pixel and time of interest. By projecting satellite precipitation fields and their retrieval errors into the Fourier frequency–wavenumber domain, we show that we can explicitly take into account the neighboring space–time multiscale dynamics of precipitation and compute a scale-dependent uncertainty.
... but the uncertainty does not shrink to zero. Using a one-minute updating rain gauge network over Korea, Turk et al. (2009) examined the performance of the NRL-Blend fast-update precipitation product across telescoping space-time averaging scales. The spacetime root mean square (RMS) error, mean bias, and correlation matrices were computed using various time windows for the gauge averaging, centered about the satellite observation time (this is necessary since the satellite measurement responds to the precipitation before it has fallen to the ground, where the gauges measure). ...
... Space-time contour plots of the correlation coefficient, root mean square error and mean bias for the rain gauge network time window average of ±10 minutes, centered about the time of the GMS satellite observation of Korea. The abscissa and ordinate of each contour plot denotes the spatial and temporal scales, respectively, used to average the rain gauge data and the NRL-blended satellite technique estimated rain (figure adapted fromTurk et al., 2009). ...
Level-3 “sub-daily global merged satellite precipitation products” are typically reported on a fixed rectangular latitude-longitude grid at high spatial and temporal resolution (respectively 0.1º and ~0.5 hour). This section is specifically concerned with the uncertainties in these products at their reported resolution. The discussion reviews the uncertainties that are inherent in the retrieval and processing steps that are used to produce the Level-3 estimates. These include the detection error, the passive microwave (MW) and infrared (IR) estimation errors, and the error incurred when using frequent IR information to fill long revisit gaps between passive MW estimates. Advantages and disadvantages of Level-3 products are also summarized.
... Each co-located data increments the histograms of TB and RR within a latitude-longitude box 2.5 • wide (i.e., a 2.5 • × 2.5 • box), as well as the eight surrounding boxes (this overlap ensures a fairly smooth transition in the histogram shape between neighboring boxes). The rationale behind these threshold values for time collocation and box size is discussed by [80]. ...
In this paper, precipitation estimates derived from the Italian ground radar network (IT GR) are used in conjunction with Spinning Enhanced Visible and InfraRed Imager (SEVIRI) measurements to develop an operational oriented algorithm (RAdar INfrared Blending algorithm for Operational Weather monitoring (RAINBOW)) able to provide precipitation pattern and intensity. The algorithm evaluates surface precipitation over five geographical boxes (in which the study area is divided). It is composed of two main modules that exploit a second-degree polynomial relationship between the SEVIRI brightness temperature at 10.8 µm TB10.8 and the precipitation rate estimates from IT GR. These relationships are applied to each acquisition of SEVIRI in order to provide a surface precipitation map. The results, based on a number of case studies, show good performance of RAINBOW when it is compared with ground reference (precipitation rate map from interpolated rain gauge measurements), with high Probability of Detection (POD) and low False Alarm Ratio (FAR) values, especially for light to moderate precipitation range. At the same time, the mean error (ME) values are about 0 mmh−1, while root mean square error (RMSE) is about 2 mmh−1, highlighting a limited variability of the RAINBOW estimations. The precipitation retrievals from RAINBOW have been also compared with the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) Satellite Application Facility on Support to Operational Hydrology and Water Management (H SAF) official microwave (MW)/infrared (IR) combined product (P-IN-SEVIRI). RAINBOW shows better performances than P-IN-SEVIRI, in terms of both detection and estimates of precipitation fields when they are compared to the ground reference. RAINBOW has been designed as an operational product, to provide complementary information to that of the national radar network where the IT GR coverage is absent, or the quality (expressed in terms of Quality Index (QI)) of the RAINBOW estimates is low. The aim of RAINBOW is to complement the radar and rain gauge network supporting the operational precipitation monitoring.
... Many studies have demonstrated that the performance of satellite estimation products strongly depends on the spatial and temporal scales at which they are evaluated (Hossain and Huffman 2008;Turk et al. 2009;Sohn et al. 2010;Scheel et al. 2011). When evaluated at spatio-temporal scales approaching their full nominal resolution, finely-gridded products may show mediocre performances, to the point that the variance of the retrieval error may be in the same order of magnitude as the statistical variance of the reference precipitation (Shen et al. 2010;Haile et al. 2013;Rios Gaona et al. 2016). ...
... The validation of satellite products is typically performed by comparison with a trusted reference dataset. The most straightforward way to perform the multiscale evaluation is to coarsen the compared fields by aggregation at multiple scales and to perform a complete analysis at each scale (Turk et al. 2009;Sohn et al. 2010;Scheel et al. 2011). However, this analysis is highly redundant: all the information contained in the coarse-resolution fields is necessarily present in the fine-resolution fields too. ...
Satellite precipitation products are essential for global analysis of water cycle dynamics as well as for regional analyses in regions where no ground observations are available. For any climatic or hydrologic application, it is important to know down to which scale a gridded satellite precipitation product can accurately resolve the spatial patterns of precipitation. This scale, which we call “effective resolution”, is a complex combination of the instrument resolution (especially so for multisensor products such as IMERG), the multi-sensor retrieval or merging algorithm, and the type of the precipitating system, and it can differ substantially from the grid size of the satellite product. Here, we use a wavelet-based framework to quantitatively define the effective resolution of the IMERG multi-satellite product by comparison with the MRMS ground radar product at the hourly time scale over the continental United States. Our findings show that the effective resolution varies across geographical areas, seasons and types of precipitation and provide insight for the use of those products in hydrologic applications and for algorithmic improvements.
... A similar exercise was conducted by Kubota et al. (2009) over Japan showing better performances of the products over ocean, reduced skill over the mountains, and overall poorer results over coastlines and small islands. Turk et al. (2009) and Sohn et al. (2010) compared high-resolution products over the South Korea dense gauge network pointing out the need for accurate PMW measurements as a prerequisite for better estimates by HRPP algorithms. The results of Roca et al. (2010) over West Africa show that the new generation of combined IR-PMW satellite products describes the rain variability similarly to ground measurements while the seasonal variability of the diurnal scale as well as its relative daily importance is only captured by some products. ...
The International Precipitation Working Group (IPWG) is a permanent International Science Working Group (ISWG) of the Coordination Group for Meteorological Satellites (CGMS), co-sponsored by CGMS and the World Meteorological Organization (WMO). The IPWG provides a focal point and forum for the international scientific community to address the issues and challenges of satellite-based quantitative precipitation retrievals, and for the operational agencies to access and make use of precipitation products. Through partnerships and biannual meetings, the group supports the exchange of information on techniques for retrieving and measuring precipitation and for enhancing the impact of space borne precipitation retrievals in numerical weather and hydro-meteorological prediction and climate studies. The group furthers the refinement of current estimation techniques and the development of new methodologies for improved global precipitation measurements, together with the validation of the derived precipitation products with ground-based precipitation measurements. The IPWG identifies critical issues, provides recommendations to the CGMS and supports upcoming precipitation oriented missions. Training activities on precipitation retrieval from space are also part of the IPWG mandate in cooperation with WMO and other bodies.
... Hossain and Huffman (2008) recommended to systematically analyze the dependence of error metrics to scale when assessing satellite rainfall data. To this end, Turk et al. (2009) and Sohn et al. (2010) aggregated satellite rain fields at various spatiotemporal resolutions to compare them with ground data. In this paper, the rain/no rain discrimination ability of a suite of highresolution products derived from spaceborne passive sensors is evaluated in West Africa. ...
... At each scale, the variances and the covariance of the coefficients are computed. This is equivalent to analyzing the masks through various bandpass filters and comparable to what is done in Turk et al. (2009) and Sohn et al. (2010), where the aggregation can be seen as a low-pass filtering. ...
... They account for only 7% of the covariance and up to 54% in the energy of the difference between the two masks. The performance of satellite products' detection generally decreases as the temporal and spatial scales decrease, consistent with the well-known improvement of the performance by spatial or temporal averaging (Turk et al. 2009;Sohn et al. 2010;Hossain and Huffman 2008). Note that the scale/correlation dependency is not perfectly monotonic, as shown in Fig. 8. ...
Validation studies have assessed the accuracy of satellite-based precipitation estimates at coarse scale
(18 and 1 day or coarser) in the tropics, but little is known about their ability to capture the finescale variability
of precipitation. Rain detection masks derived from four multisatellite passive sensor products [Tropical
Amount of Precipitation with an Estimate of Errors (TAPEER), PERSIANN-CCS, CMORPH, and GSMaP]
are evaluated against ground radar data in Burkina Faso. The multiscale evaluation is performed down to
2.8 km and 15 min through discrete wavelet transform. The comparison of wavelet coefficients allows iden-
tification of the scales for which the precipitation fraction (fraction of space and time that is rainy) derived
from satellite observations is consistent with the reference. The wavelet-based spectral analysis indicates that
the energy distribution associated with the rain/no rain variability throughout spatial and temporal scales in
satellite products agrees well with radar-based precipitation fields. The wavelet coefficients characterizing
very finescale variations (finer than 40 km and 2 h) of satellite and ground radar masks are poorly correlated.
Coarse spatial and temporal scales are essentially responsible for the agreement between satellite and radar
masks. Consequently, the spectral energy of the difference between the two masks is concentrated in fine
scales. Satellite-derived multiyear mean diurnal cycles of rain occurrence are in good agreement with gauge
data in Benin and Niger. Spectral analysis and diurnal cycle computation are also performed in the West
Africa region using the TRMM Precipitation Radar. The results at the regional scale are consistent with the
results obtained over the ground radar and gauge sites.
... The validation of satellite cloud and precipitation retrieval algorithms is a very complex and still in progress activity, which requires dedicated airborne and in situ measurements to obtain the cloudy scenario characterization as well as reliable networks of rain gauges and radar to provide independent rainfall measurements. An exhaustive summary about the recent validation activities of several satellite precipitation products can be extracted from Ebert et al. (2007) and Turk et al. (2009) , whereas examples of performance evaluation of cloud retrieval algorithms can be found in Nauss et al. (2005) and Ham et al. (2009). ...
... Critically, the validation of instantaneous precipitation products over Europe was very scarce to date. However, dependency on accumulated time and space scales of validation results have been discussed in Turk et al., 2009, where averaging was done from 0.1°to 3°box sizes, and from 1 h to 30 days long periods. ...
... We have to emphasize that the comparison of radar and satellite data over longer time scales (1 h, 3 h, 6 h, up to 30 days) would provide better statistics as shown in Turk et al., 2009. They have validated the IR-MW-combined NRL-Blended (Naval Research Laboratory -blended) satellite retrieval technique against 1-min rain gauge data over the Korean Peninsula for June-August 2000). ...
Highlights
► Microwave and infrared-based satellite precipitation products have been validated. ► All three products give the best results in summer. ► The location of convective cells is determined well by microwave observations. ► Significant underestimation of heavy rain was found for the infrared-based product. ► Both statistical comparisons and case study analysis are important for validation.
... Rainfall retrieval techniques have made a significant advancement during the recent years. Studies by Wolff and Fisher (2009) and Turk et al. (2009) show that precipitation estimates aggregated over 1 day and and 1 • are nearly unbiased and have root mean square error of 1 mm day −1 or less. Same studies indicate that on a smaller spatial 20 and temporal scale the uncertainties are still large. ...
This study focus is set on quantifying sampling related uncertainty in
the satellite rainfall estimates. We conduct observing system simulation
experiment to estimate sampling error for various constellations of
Low-Earth orbiting and geostationary satellites. There are
two types of microwave instruments currently available: cross track
sounders and conical scanners. We evaluate the differences in sampling
uncertainty for various satellite constellations that carry instruments
of the common type as well as in combination with geostationary
observations. A precise orbital model is used to simulate
realistic satellite overpasses with orbital shifts taken into account.
With this model we resampled rain gauge timeseries to simulate
satellites rainfall estimates free of retrieval and calibration errors.
We concentrate on two regions, Germany and Benin, areas with different
precipitation regimes. Our results show that sampling
uncertainty for all satellite constellations does not differ greatly
depending on the area despite the differences in local precipitation
patterns. Addition of 3 hourly geostationary observations provides equal
performance improvement in Germany and Benin, reducing rainfall
undersampling by 20-25% of the total rainfall amount. Authors do not
find a significant difference in rainfall sampling between conical
imager and cross-track sounders.
... By applying the minimum sampling error selection method for time and space proposed by Bell and Kundu (2003), a 15-min averaging time for the rain gauge measurements is needed to minimize the sampling error for a given TMI 85-GHz footprint. The 15-min time average used to minimize the sampling errors appears to be consistent with results of Turk et al. (2009) for the statistics of satellite-rain gauge comparisons with various time windows for the average from 230 to 130 min. It has been shown that there is a sharp improvement in the statistics when the time window is widened from 0 to 65 min. ...
... It has been shown that there is a sharp improvement in the statistics when the time window is widened from 0 to 65 min. This improvement appears to slow down from 65 to 610 min (see Fig. 5 of Turk et al. 2009), however Thus, we used 14-min averages for the surface rains taken within 67 min centered at a TRMM overpass time. ...
Summer rainfall characteristics over the Korean Peninsula are examined using six years of Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) measurements and surface rain measurements from the densely populated rain gauges spread across South Korea. A comparison of the TMI brightness temperature at 85 GHz with the measured surface rain rate reveals that a significant portion of rainfall over the peninsula occurs at warmer brightness temperatures than would be expected from the Goddard profiling (GPROF) database. By incorporating the locally observed rain characteristics into the GPROF algorithm, efforts are made to test whether locally appropriate hydrometeor profiles may be used to improve the retrieved rainfall. Profiles are obtained by simulating rain cases using the cloud-resolving University of Wisconsin Nonhydrostatic Modeling System (UW-NMS) model and matching the calculated radar reflectivities to TRMM precipitation radar (PR) reflectivities. Selected profiles and the corresponding simulated TMI brightness temperatures (limited in this study to values that are larger than 235 K) are added to the GPROF database to form a modified database that is considered to be more suitable for local application over the Korean Peninsula. The rainfall retrieved from the new database demonstrates that heavy-rainfall events-in particular, those associated with warmer clouds-are better captured by the new algorithm as compared with the official TRMM GPROF version-6 retrievals. The results suggest that a more locally suitable rain retrieval algorithm can be developed if locally representative rain characteristics are included in the GPROF algorithm.
