Abstract

Estimation of overland precipitation using observations from the radar and passive microwave radiometer sensors onboard the current Global Precipitation Measurement (GPM) and predecessor Tropical Rainfall Measuring Mission (TRMM) satellites is constrained by the underlying surface variability. The factors controlling the multichannel microwave surface emissivity and radar surface backscatter are related to surface properties such as soil type and vegetation properties that vary with location and time. Variability due to slowly varying seasonal changes can be considered when simulating radar reflectivities and radiometer equivalent blackbody brightness temperatures for use with precipitation retrieval algorithms. However, over certain surfaces, a more transient, dynamic surface change is manifested upon the onset of intermittent rain events. In these situations, a timely update of the surface state prior to each satellite overpass, together with knowledge of the associated variability in the emissivity and radar surface backscatter, may be useful to improve the performance of the overland precipitation retrieval algorithms. In this study, the potential for wide-swath surface backscatter observations from the Ku-band, dual-beam OceanSat-2 scatterometer (OSCAT) is examined as a surface reference for tracking previous-time precipitation. Over certain surfaces, it is shown that a time-change detection approach is useful to isolate the change in radar backscatter owing to previous 3-h rainfall accumulations from the more slowly varying background state. A practical use of this method would be the production of an ancillary previous-time precipitation map, which could be consulted by retrieval algorithms to select (or adjust the weighting of) candidate solutions that represent the most current surface conditions.

1. Introduction

The joint National Aeronautics and Space Administration (NASA) and Japan Aerospace Exploration Agency (JAXA) Global Precipitation Mission (GPM) core satellite was successfully launched on 28 February 2014. GPM is a constellation mission, whereby the observations and precipitation profile estimates from the core satellite dual-frequency (Ku and Ka band) precipitation radar (DPR) and 13-channel (10–183 GHz) passive microwave (PMW) GPM imager (GMI) act as a reference for the other constellation PMW-only satellites [e.g., the Advanced Microwave Scanning Radiometer-2 (AMSR-2) onboard the Global Change Observation Mission–Water (GCOM-W) spacecraft, operated by JAXA; Hou et al. 2014].

With its 65° inclination, GPM observes more land surface area relative to the predecessor Tropical Rainfall Measuring Mission (TRMM) and associated extremes in seasonality, cold and snow-covered surfaces, inland water, and forest and vegetation. From space, the associated radar surface backscatter and the PMW surface emissivity signatures define the background state from which to carry out GPM precipitation retrievals from the radar and radiometer, respectively. Under nonraining conditions, the overland radar surface backscatter is highly variable, depending upon surface type and the viewing angle, which can impact the accuracy of path-integrated attenuation (PIA) radar retrieval methods (Meneghini and Jones 2011). At the center frequencies of the TRMM Microwave Imager (TMI) and GMI, the land surface emissivity varies with soil moisture and, to a lesser degree, the soil texture and surface roughness. Soil moisture increases the dielectric constant of the soil and water mixture and thus decreases the surface emissivity; the surface roughness increases scattering and surface area, resulting in an increasing surface emissivity (Li et al. 2010). The combination of the varying land surface emissivity and the surface skin temperature typically produces a radiometric equivalent blackbody brightness temperature (TB) background that is difficult to contrast against the desired PMW signal owing to the precipitation (Munchak and Skofronick-Jackson 2013; Ferraro et al. 2013).

During the TRMM era, the PMW overland precipitation products were largely empirically based (Gopalan et al. 2010), relying upon the lower-frequency channels (i.e., 37 GHz and below) for discriminating the presence of precipitation and precipitation-sized hydrometeor-induced TB depressions at or near 85 GHz to quantify the precipitation rate. With GPM, the design of the PMW precipitation retrieval is a Bayesian-based estimation for all surfaces, whose veracity and accuracy relies upon the capability to physically model the multichannel TB of all scenes, and under all atmospheric conditions (Kummerow et al. 2011). A probabilistic Bayesian type of estimate is appropriate for inversion of profiling radar and/or radiometric observations of precipitation, since the space–time variability of the factors influencing the observations (e.g., the precipitation microphysics) is described by a very large number of nonindependent physical parameters, and the Bayesian solution inherently provides a measure of uncertainty. However, it does place a heavy burden on the capability to forward model the expected range of scenes with adequate realism, taking into account natural variability describing the precipitation media as well as the underlying surface, for building the a priori dataset (Petty 2013; L’Ecuyer and Stephens 2002). Slowly varying changes to the surface, such as seasonal variations in vegetation, can be captured by land surface classification techniques. A well-known example is the Tool to Estimate Land Surface Emissivities at Microwave Frequencies (TELSEM) passive microwave–based surface classification (Prigent et al. 2006; Aires et al. 2011), which provides a lookup table method to interpolate the emissivity mean and variance at a specified incidence angle and frequency, using a precalculated monthly mean emissivity climatology derived from the seven-channel (19–85 GHz) Special Sensor Microwave Imager (SSM/I) observations. TELSEM has been successfully implemented into the current (as of March 2015, version 1) GPM radiometer algorithm (Kummerow et al. 2011), using a 15-class emissivity index to stratify the surface. With this algorithm, the Bayesian search is constrained to consider a priori candidate profiles that have the same classification index as the observation. In the radar case, the TRMM Precipitation Radar (PR) and GPM radar–based precipitation profile retrieval is improved with the surface reference technique (SRT; Meneghini et al. 2000) to establish the two-way path-integrated attenuation. The PIA estimation hinges upon estimation of the attenuation due to the presence of rain in the radar field of view, against the underlying natural normalized radar cross section (NRCS; typically denoted by ) variability owing to surface roughness, with the largest variability near nadir incidence. The variability is estimated from PR observations over rain-free land areas, using temporal (i.e., previous rain-free ) and spatial references (Meneghini et al. 2004), and variations on this method are being tested for the GPM radars.

In this manuscript, we examine the feasibility of a more timely update to the surface state prior to each satellite overpass, which, together with knowledge of the associated variability in the multichannel emissivity and , may improve the specification of the surface properties needed for the radiometer and radar overland precipitation retrieval algorithms. For example, the sudden appearance of late season snow cover over a region that is climatologically snow free will confuse a classification-based precipitation retrieval, since it will consider snow-free areas in its search for candidate solutions. Agreement between observations and simulations may not be reached, or may be reached for the wrong conditions (e.g., particle scattering processes within the precipitation media can result in TB values that appear radiometrically similar to snow-covered surfaces). For certain surfaces, the onset of rain tends to notably decrease the emissivity at the lower-frequency channels below 37 GHz. This dynamic emissivity change owing to rain events may appear quickly and persist for several days, depending on the precipitation timing and duration. Radar surface backscatter may also be affected, causing differences between the true surface backscatter and that available as a reference value in the SRT algorithm (Seto and Iguchi 2007). Other dynamic change examples include the appearance or disappearance of inland water bodies, or abrupt changes to vegetation owing to agricultural or forest practices.

Characterization of land surfaces by examining the imprint that rainfall leaves upon the associated PMW TB extends back to early studies with the SSM/I antecedent precipitation index (API) (Teng et al. 1993). More recently, C- and Ku-band scatterometer observations have been studied for estimating soil moisture (Mladenova et al. 2009), vegetation (Lu et al. 2013), detection of long periods of no precipitation for drought-monitoring indices (Nghiem et al. 2012), freeze–thaw conditions (Colliander et al. 2012) and snowmelt (Wang et al. 2008), and prior rainfall accumulations (Brocca et al. 2014). In particular, the wide-swath (1800 km), multiple-look pencil-beam scatterometer observations of the Ku-band, dual-beam OceanSat-2 scatterometer (OSCAT) are studied for their sensitivity to globally capturing and tracking short-period rain events. Active radar backscatter measurements are less sensitive to the surface temperature, but more affected by surface roughness and length scales, which change over longer time scales than the diurnally fluctuating surface temperature. Designed to capture small signals arising from ocean wind direction asymmetry, a scatterometer is by necessity very stable and well calibrated, to within tenths of a decibel (Bhowmick et al. 2014). There is a very wide body of work that has been done on radar backscattering over different surface types, and it is not the intent of this study to examine the physical modeling of vegetation canopies or soil properties.

The approach begins by first analyzing high-resolution (in space and time) ground radar precipitation products over the continental United States to examine the sensitivity of the joint TRMM radar surface backscatter and lower-frequency TMI surface emissivity under nonraining conditions as a function of the previous 3-hourly and longer (antecedent) rain accumulations and the PR incidence angle. This sensitivity study is needed to establish the magnitude of any displacement in the mean and covariance of these surface properties between scenes where no previous rainfall was noted, and those that experienced various amounts of rainfall. The hypothesis is made, similar to Turk et al. (2014a), that the surface variability that would be experienced during actual raining conditions is similar to the surface variability after a period antecedent precipitation. We are not attempting to estimate previous rainfall totals such as is done in Brocca et al. (2014) but rather to identify threshold conditions. A similar analysis is performed for the OSCAT backscatter cross sections during the same time period. It was noted that over certain surface types, a simple OSCAT time-change detection approach was capable of detecting locations that had experienced precipitation within the previous 3 h with a false alarm rate near 0.1 and detection rate near 0.8, with the veracity dropping over the more vegetated land surface classes. A practical use of this method would be the production of an ancillary, near-real-time updated previous-time precipitation map, which could be consulted by GPM and TRMM retrieval algorithms to adjust or modify the weighting of candidate solutions to better represent the most up-to-date surface conditions.

2. Description of datasets

The analyses will focus on OSCAT and TRMM PR and TMI for a 2-yr period (2010–11) over the continental United States. To capture the rapid time evolution of precipitation during the 24-h period just prior to each of these satellite overpasses, the high-refresh, time-coincident precipitation derived from NEXRAD data is used. In this section, we provide a brief description of each of these datasets.

a. OceanSat-2 over land

The Indian Space Research Organisation (ISRO) OceanSat-2 spacecraft was launched on 23 September 2009 from the Satish Dhawan Space Center in southeastern India. OceanSat-2 orbits with a local time on ascending node (LTAN) of 1200 local time ±10 min and ground tracks have a controlled revisit track pattern repeating every two days, in order to allow frequent observations from the Ocean Color Monitor (OCM) under similar solar illumination conditions. The OceanSat-2 Ku-band (13.5 GHz) scatterometer payload successfully operated from 29 September 2009 to 20 February 2014. OSCAT is a conically scanning, pencil-beam scatterometer whose primary task is the estimation of ocean surface wind vectors (Gohil et al. 2013). Its dual-beam, conically scanning antenna system is very similar to the two NASA SeaWinds scatterometers, the first of which operated in tandem with the Advanced Microwave Scanning Radiometer onboard the Advanced Earth Observing Satellite (ADEOS) spacecraft for eight months during 2003. The second SeaWinds was the sole payload onboard the QuikSCAT spacecraft, which operated from 19 July 1999 to 21 November 2009, allowing a brief two-month overlap between QuikSCAT and OSCAT. The OSCAT viewing geometry is illustrated in Fig. 1. The inner (outer) beam transmits and receives at horizontal H (vertical V) polarization with a 48.9° (57.6°) Earth incidence angle. The OSCAT level-1 product reports the inner and outer scan at two levels, the slice level and the footprint level (slices weighted by the corresponding X factor), which are subsequently used for the level-2 ocean vector winds processing. This provides four views of (approximately) the same grid cell from different azimuth viewing directions within the narrower 1400-km inner swath, but only two in the narrow strip self-contained by the outer scan. Over ocean, the relative separation of these two or four looks determines the wind vector quality for each grid cell. Over land, there can be considerable azimuthal anisotropy from certain types of land surfaces, which needs to be taken into consideration for geophysical retrievals (Bartalis et al. 2006).

Fig. 1.

Depiction of the OSCAT dual-beam conical scanning. The horizontal-polarized inner (vertical-polarized outer) beam provides an Earth incidence angle of 48.9° (57.6°) from nadir, with an 1820 (1400)-km swath width. The ISRO level 1B datasets contain 282 outer beam footprint aggregated grid cells (281 for inner beam), spaced by ~15 km across and along track. Figure courtesy of ISRO.

Fig. 1.

Depiction of the OSCAT dual-beam conical scanning. The horizontal-polarized inner (vertical-polarized outer) beam provides an Earth incidence angle of 48.9° (57.6°) from nadir, with an 1820 (1400)-km swath width. The ISRO level 1B datasets contain 282 outer beam footprint aggregated grid cells (281 for inner beam), spaced by ~15 km across and along track. Figure courtesy of ISRO.

During 2010–11, approximately 5 million (M) ascending and 5M descending nonraining OSCAT footprint pixels were gathered over the continental United States (CONUS) from the inner swath region, where all four azimuthal views were available. A unique benefit of OSCAT is the relative stability of the four azimuthal viewing directions, which will be further explained below when these data are analyzed over land.

b. TRMM PR and TMI observations over land

The TRMM Precipitation Radar operates at Ku-band (13.8 GHz), with an electronically controlled cross-track scan encompassing 245 km (from the 405-km altitude after the 2001 orbit boost), as the PR scans its ±18° swath. Across this swath, the radar processor apportions the received power at 49 beam positions, covering 25 [i.e., (49 − 1)/2 + 1] unique viewing angles ranging from near nadir to 18°, each approximately 4 km × 4 km resolution. The companion passive microwave radiometer for PR is the TRMM Microwave Imager, which provides equivalent blackbody brightness temperatures at nine channels between 10 and 85 GHz in a conically scanning mode, covering an 830-km swath, with on-Earth resolutions ranging from 72 km × 43 km at 10.7 GHz to 8 km × 6 km at 85.5 GHz. To account for the on-Earth resolution difference between the PR and TMI fields of view, the matched PR–TMI dataset procedure averages the PR radar beams in a 3 × 3 region centered at each TMI sample location, within the coincident PR swath. Since this analysis only considers nonraining scenes, this simplified averaging was assumed as a first-order adjustment for the on-Earth sensor resolution differences. The maximum latitude achieved by the PR limits the land area for this study, and the 2-yr analysis period (2010–11) provides a sufficient number of overpasses over the southern United States below 35°N latitude, where continuous ground-based radar coverage from the operational NEXRAD network is available.

c. NEXRAD precipitation mosaic

To analyze the subdaily precipitation time history prior to each overpass within the CONUS, two products from the National Mosaic and Multi-Sensor Quantitative Precipitation Estimation (NMQ) radar product (Zhang et al. 2011) were used: a 0.01° precipitation rate updated every 5 minutes over the CONUS land areas and an hourly accumulated precipitation product. The former was used to identify the precipitation conditions at the time (within 5 min) of each OSCAT overpass during 2010–11 to identify nonraining OSCAT scenes. For the nonraining pixels in each scene, the hourly NMQ products were backward time-integrated to obtain the accumulated precipitation at seven intervals (1, 3, 6, 12, 24, 48, and 72 h). For TRMM overpasses over the CONUS during these same two years, the accumulated precipitation at these same five intervals were added to each nonraining PR–TMI pixel (nonraining conditions directly inferred from the PR itself). This common reference precipitation dataset enables a self-consistent analysis to be carried out between the OSCAT and PR–TMI observations and any previous rainfall.

3. Response of PR–TMI to previous-time precipitation

A number of recent studies have highlighted the geographical variability in PR surface backscatter (Meneghini et al. 2000, 2004; Seto and Iguchi 2007) and the multichannel microwave surface emissivity (Turk et al. 2014b; Norouzi et al. 2012). Seto and Iguchi (2007) noted that the TRMM PR was higher under rainfall than under no rainfall conditions, which they attributed to soil moisture, but the effect was much less for land surfaces covered by dense vegetation. It is therefore helpful to have a global land surface classification to better identify locations on Earth where this effect is most notable. Commonly, satellite visible or infrared spectral vegetation indices are used to classify Earth surfaces or vegetation, but the classifications that result are representative of the scattering and emission interactions near the top of the vegetation canopy (Hunt et al. 2011; Glenn et al. 2008; Yilmaz et al. 2008). For this study, it is useful to have a land surface classification that is more directly related to soils and interactions deeper into the vegetation canopy, based on similar microwave frequencies.

The TRMM PR observations have been used previously to relate soil moisture over semiarid areas (Stephen et al. 2010) and for establishing water stage in wetland areas (Puri et al. 2011). To better separate land surfaces for use by the TRMM SRT, Durden et al. (2012) developed a surface classification that is based on a principal component analysis of the PR at the 25 incidence angles. A nine-class (eight land and one ocean) clustering was found to represent the variability on a 0.1° grid map. Their hypothesis was that by maintaining an SRT reference for each land class, the radar precipitation retrieval is able to better match the reference backscatter to the intrinsic backscatter within the raining area. They found that a similar clustering was obtained when analyzing a limited set of Ka-band radar data gathered from the airborne Advanced Precipitation Radar (APR-2). Figures 2 and 3 show the map of the land classes over the regions of the CONUS and South America that are covered by the TRMM orbit. While there is no direct physical quantity attached to each class index, land class 8 appears to cover dense vegetation, whereas land class 7 appears to represent less vegetated areas such as south-central Brazil and shrub land in the western United States and land class 3 includes regions with surface water or low-lying coastal areas. In terms of the PR variability, the lower-numbered land classes correspond to regions with high variability, and higher-number land classes are regions with low variability.

Fig. 2.

Map of the TRMM PR land surface classification (Durden et al. 2012) over the southern portion of the continental United States covered by the TRMM radar.

Fig. 2.

Map of the TRMM PR land surface classification (Durden et al. 2012) over the southern portion of the continental United States covered by the TRMM radar.

Fig. 3.

As in Fig. 2, but for continental South America.

Fig. 3.

As in Fig. 2, but for continental South America.

The merged PR–TMI dataset was separated by each of the land classes to examine the magnitude of the joint change of the PR surface backscatter and the nine-channel TMI surface emissivity vector , under various antecedent precipitation conditions. The TMI emissivities were computed from the principal component–based method for estimating the clear-scene emissivities directly from the TB, developed by Turk et al. (2014a). Many previous studies have shown that the 10-GHz horizontally polarized emissivity estimated from sensors such as TMI or AMSR-E exhibits the widest dynamic range and sensitivity to soil properties (Ringerud et al. 2014) and is therefore best suited to jointly adjust all emissivities via the emissivity covariance. For different NMQ-derived antecedent precipitation conditions, the variability of the PR at four increasing incidence angle ranges is summarized in the cumulative distribution functions (CDFs) of Fig. 4. Figure 4 (left) is for land class 4 and Fig. 4 (right) is for land class 8. The solid black line represents all scenes where PR indicated no rain. Then, these data are separated by previous-time history; the CDFs of all OSCAT grid cells where NMQ indicated no rain within the previous 24 and 72 h are shown by the green and blue lines, respectively. All three CDFs are nearly the same, and there was not much additional effect noted if the nonraining period was extended past 24 h. Next, the OSCAT PR scenes where the NMQ analysis exceeded 10- and 25-mm accumulations over the previous 24 and 72 h are shown with red and magenta lines, respectively.

Fig. 4.

Histogram of the PR backscatter cross section for land classes (left) 4 and (right) 8, for four different PR zenith angles ranges (0°–4°, 4°–8°, 8°–12°, and 12°–18°), over the southern United States covered by the TRMM PR, for all TRMM overpasses during 2010 and 2011, under nonraining conditions (PR = 0). The line colors indicate various NMQ-observed antecedent precipitation conditions. Black line: all PR = 0 scenes; red line: scenes where NMQ observes >10 mm accumulated rain during the previous 3 h; magenta line: scenes where NMQ observes >25 mm accumulated rain during the previous 24 h; green line: scenes where NMQ = 0 during the previous 24 h; and blue line: scenes where NMQ = 0 during the previous 72 h.

Fig. 4.

Histogram of the PR backscatter cross section for land classes (left) 4 and (right) 8, for four different PR zenith angles ranges (0°–4°, 4°–8°, 8°–12°, and 12°–18°), over the southern United States covered by the TRMM PR, for all TRMM overpasses during 2010 and 2011, under nonraining conditions (PR = 0). The line colors indicate various NMQ-observed antecedent precipitation conditions. Black line: all PR = 0 scenes; red line: scenes where NMQ observes >10 mm accumulated rain during the previous 3 h; magenta line: scenes where NMQ observes >25 mm accumulated rain during the previous 24 h; green line: scenes where NMQ = 0 during the previous 24 h; and blue line: scenes where NMQ = 0 during the previous 72 h.

For land class 4, and near-nadir observations (PR zenith 0°–4°), varies over the widest range and the value increases for the previous-rain cases relative to the no-previous-rain cases (the CDF is displaced to the right by about 5 dB). Away from nadir, varies less as expected, but the cumulative magnitude of the rain effect is about the same value. Contrasting this behavior with land class 8, in Fig. 4 (right), the overall previous rain effect is still present in the displacement of the CDFs, but by only about 1–2 dB.

Figures 5 and 6 (land classes 4 and 8, respectively) depict results from these same experiments, but the joint variability between PR Ku-band and 10H-GHz surface emissivity is plotted in a 2D histogram format. Figures 5 and 6 (left) represent all scenes where NMQ indicated no rain in the previous 24 h, and Figs. 5 and 6 (right) represent where NMQ indicated >25 mm in the previous 24 h. Most notably, the land class 4 (less vegetated) scenes exhibit a wider displacement for the previous-rain cases than does land class 8. The associated range of is between 0.75 and 0.95 independent of the PR zenith angle (the TMI incidence angle is nearly constant across the scan) for land class 4, and about 0.85–0.95 for land class 8. The locations with less dense overall vegetation water content (e.g., central United States, eastern South America) are also the locations where the relationship between soil moisture and previous TRMM 5-day rainfall amounts were found to be most self-consistent (Brocca et al. 2014).

Fig. 5.

Joint 2D histogram of the TMI 10H-GHz surface emissivity and the PR backscatter cross section for land class 4, and four different PR zenith angles ranges (0°–4°, 4°–8°, 8°–12°, and 12°–18°), under nonraining conditions over the southern United States covered by the TRMM radar, for all TRMM overpasses during 2010 and 2011. (left) Scenes where NMQ = 0 during the previous 24 h. (right) Scenes where NMQ observes >25 mm accumulated rain during the previous 24 h.

Fig. 5.

Joint 2D histogram of the TMI 10H-GHz surface emissivity and the PR backscatter cross section for land class 4, and four different PR zenith angles ranges (0°–4°, 4°–8°, 8°–12°, and 12°–18°), under nonraining conditions over the southern United States covered by the TRMM radar, for all TRMM overpasses during 2010 and 2011. (left) Scenes where NMQ = 0 during the previous 24 h. (right) Scenes where NMQ observes >25 mm accumulated rain during the previous 24 h.

Fig. 6.

As in Fig. 5, but for TRMM land class 8.

Fig. 6.

As in Fig. 5, but for TRMM land class 8.

Since the natural variability of the nonraining PR is already known, these analyses have provided an indication of the magnitude of any additional variability due to the presence of antecedent precipitation, and locations where such added information is likely to be beneficial to radar and/or radiometer precipitation retrieval techniques, for knowing how much to vary the radar backscatter or surface emissivity . However, to keep this study focused, we did not go into additional detail as to the overall timing of the rain and the duration of the time interval.

4. OSCAT surface backscatter and time change

In this section, the ability of OSCAT observations to indicate the occurrence and amount of previous-time precipitation is studied using the same TRMM surface classes and NMQ analyses as section 3.

Figure 7 illustrates the OSCAT for the four views (clockwise from upper left: outer fore, inner fore, inner aft, and outer aft), for the midnight local time (near 0600 UTC) overpass on 20 May 2011 over the central United States. The horizontal-polarized inner beam is larger than the corresponding vertical-polarized outer beam, and shows a wider dynamic range over nonurban areas. Differences between the corresponding fore and aft views are more apparent near the center of the scans, where the relative azimuth difference between the fore and aft views is larger than for scenes near the swath edges. This feature is especially notable for large urban areas such as Kansas City (near swath center) and Denver (near edge of the swath), which are all highly reflective. Therefore, some means to account for or remove azimuthally dependent terrestrial signatures will be required to contrast any change in OSCAT due to precipitation-related effects. The overland azimuth viewing angle dependence of land targets has been examined by (Bartalis et al. 2006), using European Remote Sensing (ERS) scatterometer observations, a fan-beam system that transmits three fixed-angle beams pointing in different directions relative to the spacecraft motion. They found the azimuthal asymmetry to be over 2 dB in urban and agricultural areas. One advantage of OSCAT for the purposes of this study is its very stable orbit, providing nearly constant viewing geometry in both elevation and azimuth for each of the four views, which reduces the overall variability among the inner/outer and fore/aft for each scene. This characteristic is helpful for time-change detection approaches, which will be studied in the next section.

Fig. 7.

OSCAT footprint (dB) from the overpass at 0600 UTC 20 May 2011 over the central United States, from each of the four viewing geometries: (top left) outer fore, (top right) inner fore, (bottom left) outer aft, and (bottom right) inner aft. Major cities are labeled to the right of their location.

Fig. 7.

OSCAT footprint (dB) from the overpass at 0600 UTC 20 May 2011 over the central United States, from each of the four viewing geometries: (top left) outer fore, (top right) inner fore, (bottom left) outer aft, and (bottom right) inner aft. Major cities are labeled to the right of their location.

To demonstrate the time change for a particular region, Fig. 8 (top and middle) shows the 2-yr (2010–11) time series from the OSCAT descending and ascending orbits, which cover a location (42.1°N, 92.3°W) in central Iowa. Each symbol and color represents one of the four OSCAT viewing geometries shown in Fig. 7. The blue impulses denote the daily previous 24-h NMQ precipitation accumulations. For reference, the green symbols denote the daily WindSat-derived vegetation water content (VWC; kg m−2) derived by Turk et al. (2014b) and Li et al. (2010), here scaled by 6 to fit the range of the left y axis. The cycle of due to agricultural growing in the summer months exceeds 6 dB peak to peak and very closely tracks the VWC. On top of this, the effects due to precipitation events are not readily apparent. Figure 8 (bottom) shows these same quantities (descending and ascending together for more points) over a 4-month period beginning on 1 April 2010. The sudden increase in immediately following precipitation events in April and May can exceed 4 dB (not accounting for Ku-band propagation effects, which are not corrected for in these data), especially for the horizontal-polarized inner beam. But after early June, the rain effect on the surface is less apparent, masked by the presence of the larger VWC. In the next section, we examine the time change over the larger CONUS area.

Fig. 8.

Time series of each OSCAT overpass over a location (42.1°N, 92.3°W) in central Iowa, where each symbol and color represents a different OSCAT viewing geometry. The blue impulses denote the daily, previous 24-h NMQ precipitation accumulations. The green symbols denote the WindSat-derived VWC (Turk et al. 2014a), scaled by 6 to fit the range of the left axis. (top) All descending (midnight local time) overpasses during 2010–11. (middle) As in (top), but for ascending (noon local time) overpasses. (bottom) Close-up of the 4-month period from April to July 2010, showing the increase in the inner beam following rain events, which is less noticeable during the heavier VWC conditions noted after mid-June.

Fig. 8.

Time series of each OSCAT overpass over a location (42.1°N, 92.3°W) in central Iowa, where each symbol and color represents a different OSCAT viewing geometry. The blue impulses denote the daily, previous 24-h NMQ precipitation accumulations. The green symbols denote the WindSat-derived VWC (Turk et al. 2014a), scaled by 6 to fit the range of the left axis. (top) All descending (midnight local time) overpasses during 2010–11. (middle) As in (top), but for ascending (noon local time) overpasses. (bottom) Close-up of the 4-month period from April to July 2010, showing the increase in the inner beam following rain events, which is less noticeable during the heavier VWC conditions noted after mid-June.

a. Establishing the daily background

The relationship between the radar in relation to the surface soil or vegetation properties is a complex interaction and scattering process, and there is a large body of published studies, especially within the L-band synthetic aperture radar area (Huang et al. 2010). Given the wide variability in these properties at the time of any satellite overpass, a time rate of change approach was utilized to isolate precipitation events. Time-change approaches are commonly used for studies of vegetation and soil dynamics (Wagner et al. 1999; Kimball et al. 2004).

A previous-time history of all OSCAT data was prepared for each box on a 0.2° latitude–longitude grid. Judging by the time history effect arising from precipitation events in Fig. 8 (bottom), a 15-day moving median filter was independently applied to each of the four viewing geometries [a median filter is less affected by precipitation events (outliers) than a moving average filter]. The data were processed sequentially from January 2010 through the end of 2011, and the global map was written as a daily netCDF background file (denoted as ) at 0000 UTC each day. Figure 9 (top) shows the map of the inner and outer descending value for the 15-day period ending on 3 March 2011. The major urban areas noted in Fig. 7 are captured, as well as snow-covered areas in higher elevations and northern extremes. Owing to the smaller swath of the OSCAT inner beam, its overall revisit time is occasionally insufficient to capture short-term snow events (e.g., swath edge artifacts in northern Montana). Figure 9 (bottom) is the same, but for the period ending on 1 August 2011.

Fig. 9.

(top) Maps of the OSCAT 15-day median filtered background values (dB), for the (left) inner and (right) outer beams for 3 Mar 2011. (bottom) As in (top), but for 1 Aug 2011.

Fig. 9.

(top) Maps of the OSCAT 15-day median filtered background values (dB), for the (left) inner and (right) outer beams for 3 Mar 2011. (bottom) As in (top), but for 1 Aug 2011.

The application of these daily background datasets to the overpass of Fig. 7 is demonstrated in Fig. 10. Since OSCAT does not carry a companion radar, the nearest 5-min rain rate (mm h−1) from the NEXRAD NMQ analysis is shown in Fig. 10 (top left) to determine the precipitation during each OSCAT overpass. Figure 10 (top right) and Fig. 10 (middle left) depict the NMQ accumulations (mm) during the 6 and 24 h prior to the OSCAT overpass, respectively. The regions of enhanced that lie within the precipitating regions in Fig. 7 (top left; e.g., the line of heavy precipitation extending southwest of Oklahoma City, labeled “B” in Fig. 10) are manifestations of the joint effects of the surface-related precipitation effects (which tend to increase ), the Ku-band propagation effects (which decreases ), and volume backscatter from precipitation (which increases ).

Fig. 10.

(top left) Nearest 5-min rain rate from NEXRAD NMQ during the OSCAT overpass shown in Fig. 7. (top right) NMQ accumulations during the 6 h prior to the OSCAT overpass. (middle left) NMQ accumulations during the 24 h prior to the OSCAT overpass. (middle right) Outer beam residual (dB). (bottom left) Inner beam residual (dB). (bottom right) WindSat-derived VWC from Turk et al. (2014) on this date or during the 5 days prior.

Fig. 10.

(top left) Nearest 5-min rain rate from NEXRAD NMQ during the OSCAT overpass shown in Fig. 7. (top right) NMQ accumulations during the 6 h prior to the OSCAT overpass. (middle left) NMQ accumulations during the 24 h prior to the OSCAT overpass. (middle right) Outer beam residual (dB). (bottom left) Inner beam residual (dB). (bottom right) WindSat-derived VWC from Turk et al. (2014) on this date or during the 5 days prior.

Using from 0000 UTC on this day, Fig. 10 (middle right) shows the outer beam residual, and Fig. 10 (bottom left) shows the inner beam residual, both in decibels. The scale is intentionally compressed to highlight regions below 5 dB. From a qualitative standpoint, the map of the residual value exceeding ~2 dB coincides with regions indicating precipitation within the previous 3 h, and as far back as 24 h prior to the OSCAT overpass. Figure 10 (middle left) shows that the region in southern Missouri (labeled “C”) experienced substantial precipitation in the previous 24 h, but the resultant signature is not readily apparent in any increased residual noted in regions A and B under similar conditions. Note that the land near region C is characterized by a more vegetated, mountainous terrain east of the Kansas–Missouri state border, which is readily apparent in the high VWC [Fig. 10 (bottom right)]. The decreased sensitivity of the residual over vegetated areas was also noted in the mid to late summer months in the case of Fig. 8, and is consistent with previous TRMM PR soil moisture sensitivity studies (Seto and Iguchi 2007).

b. Overall detection performance statistics

The previous example was designed to illustrate the regions and conditions whereby the relative change in the OSCAT could be attributed to dynamic land surface changes from effects of previous precipitation. To assess this more comprehensively, OSCAT overpasses over the continental United States east of 100°W longitude (where NEXRAD has nearly contiguous coverage) during 2010–11 were analyzed in the same fashion as above, using NMQ analyses to trace the previous precipitation time history every hour, in the 24 h prior to each overpass. Histograms and CDFs similar to those shown in Fig. 4 were prepared for four azimuthal viewing conditions (inner beam descending, outer beam descending, inner beam ascending, outer beam ascending). To carry more points in the statistics, the fore and aft views were combined, but the ascending–descending distinction was maintained in order to assess any possible local time-of-day effects. These cases are shown in Fig. 11, for all eight land classes (2–9) together. The overall displacement in the residual is between 2 and 4 dB for either of the previous rain conditions (accumulated R > 10 mm in the previous 3 h, or R > 25 mm in the previous 24 h). Also, the descending passes (midnight local time) have a slightly larger shift in the residual CDF relative to the no-rain CDF than the ascending passes (noon local time). Since much of the overland precipitation develops later in the day, one possible explanation is that midnight local time passes captured precipitation-induced surface events more “quickly” than the orbits at noon local time (ascending orbit) the next day.

Fig. 11.

Histograms (left y axis) and CDFs (right y axis) of the residual (observed-minus-background) (dB) for all nonraining OSCAT overpasses during 2010–11 over the CONUS. The black lines indicate all OSCAT cells where the NMQ radar analysis indicated no rain in the previous 24 h. The red lines indicate all OSCAT cells where the NMQ radar analysis indicated more than 10 mm of accumulated rainfall in the 3 h prior to the overpass. The magenta lines indicate all OSCAT cells where the NMQ radar analysis indicated more than 25 mm of accumulated rainfall in the 24 h prior to the overpass. Shown are (top left) inner descending, (top right) outer descending, (bottom left) inner ascending, and (bottom right) outer ascending.

Fig. 11.

Histograms (left y axis) and CDFs (right y axis) of the residual (observed-minus-background) (dB) for all nonraining OSCAT overpasses during 2010–11 over the CONUS. The black lines indicate all OSCAT cells where the NMQ radar analysis indicated no rain in the previous 24 h. The red lines indicate all OSCAT cells where the NMQ radar analysis indicated more than 10 mm of accumulated rainfall in the 3 h prior to the overpass. The magenta lines indicate all OSCAT cells where the NMQ radar analysis indicated more than 25 mm of accumulated rainfall in the 24 h prior to the overpass. Shown are (top left) inner descending, (top right) outer descending, (bottom left) inner ascending, and (bottom right) outer ascending.

Figure 12 shows these same analyses, but only for land classes 4 (solid lines) and 8 (dotted lines). Similar to the PR analyses shown in Fig. 4, the shift in the residual CDF toward larger values is more substantial for land class 4 than 8. For the descending (midnight local time) orbits, the shift is much larger for the shorter-duration precipitation (accumulated R > 10 mm in the previous 3 h; red line) than for the longer-duration precipitation (accumulated R > 25 mm in the previous 24 h; magenta line). Similar to the reasoning above, one explanation could be that there is a response time whereby particular surface types more quickly respond to precipitation (via changes in emissive or reflective characteristics) over shorter time scales. Of course, the efficacy of this time-change approach relies upon the capability of the background analysis to accurately capture longer-period changes among the presence of perturbations from rain events. The 15-day period used for the moving median filter was selected purely ad hoc from Fig. 8, by no means a rigorous determination.

Fig. 12.

CDFs of the residual (observed-minus-background) (dB) for all nonraining OSCAT overpasses during 2010–11 over the CONUS for land classes 4 and 8. The layout and color scale is the same as Fig. 11. Solid (dotted) lines indicate land class 4 (land class 8).

Fig. 12.

CDFs of the residual (observed-minus-background) (dB) for all nonraining OSCAT overpasses during 2010–11 over the CONUS for land classes 4 and 8. The layout and color scale is the same as Fig. 11. Solid (dotted) lines indicate land class 4 (land class 8).

For the inner swath region, and scenes where all four observations were available at each overpass, a simple linear discriminant was applied to the collection of four residual values, similar to Turk et al. (2014a). These data were separated into two populations (no rain in the preceding 24 h, and accumulated R > 10 mm in the previous 3 h). Figure 13 (left) shows how these two populations separate and Fig. 13 (right) shows the corresponding relative operating characteristic (ROC) shape (Wilks 2011). The better the separation, the closer the inflection point on the curve in Fig. 13 (right) comes to the upper-left corner, where the combination of low false alarm rate and larger hit rate is optimized. The best performance was noted for land class 4, which exceeds 80% hit rate with less than 10% false alarm rate. The worst performance is noted for land class 3, which represents the scenes where surface water is likely present, and additional rainfall on top of this is unlikely to exhibit much additional change to the emissivity or radar backscatter.

Fig. 13.

(left) Histogram of the four-element discriminant value. The black line represents the discriminant value calculated for all OSCAT cells where the NMQ radar analysis indicated no rain in the previous 24 h. The red line represents the discriminant value for all OSCAT cells where the NMQ radar analysis indicated more than 10 mm of accumulated rainfall in the 3 h prior to the overpass. (right) Corresponding ROC curve, plotted separately for land classes 3, 4, 5, 6, and 8.

Fig. 13.

(left) Histogram of the four-element discriminant value. The black line represents the discriminant value calculated for all OSCAT cells where the NMQ radar analysis indicated no rain in the previous 24 h. The red line represents the discriminant value for all OSCAT cells where the NMQ radar analysis indicated more than 10 mm of accumulated rainfall in the 3 h prior to the overpass. (right) Corresponding ROC curve, plotted separately for land classes 3, 4, 5, 6, and 8.

5. Conclusions

In this study, a 2-yr set of radar backscatter cross-sectional observations from the Ku-band, dual-beam OceanSat-2 scatterometer (OSCAT) were examined for their capability to track dynamic surface changes owing to previous-time (antecedent) precipitation. The tightly maintained OSCAT ground repeat track pattern maintains nearly constant azimuthal viewing directions for the inner fore, inner aft, outer fore, and outer aft viewing directions for each ascending or descending overpass. An eight-class surface classification derived from the zenith variability in the TRMM Precipitation Radar (PR) surface backscatter cross-sectional was used separate surface types, based upon their versus incidence angle variability. Further analysis of radar surface backscatter cross-sectional observations from nonraining TRMM overpasses covering the continental United States during 2010–11 indicated larger joint variability between the PR and the TMI 10H emissivity for less vegetated surfaces, such as the central United States. This joint variability was more pronounced when these same observations were further separated by the amount of previous (antecedent) accumulated precipitation. For land class 4 (the most surface-sensitive class), the associated range of and was between 0.85 and 0.95 and −7 and −2 dB, respectively. When the scenes where more than 10 mm of precipitation had fallen in the previous 3 h were examined, these same ranges expanded to between 0.75 and 0.95, and increased to between −5 and +5 dB, respectively, for PR zenith angles between 8° and 12°.

Over certain classes, it was shown that a time-change detection approach, based on a daily updated OSCAT background analysis applied to each of the four OSCAT azimuthal viewing geometries, is useful to isolate the change in owing to previous 3-h rainfall accumulations, from the change of owing to seasonality and other longer-term surface changes. A 15-day period was chosen for the moving median filter length. The magnitude of the change in the resulting OSCAT residual was between 2 and 4 dB, and was shown to separate certain surface classes with no previous precipitation, from conditions where the accumulated precipitation exceeded 10 mm in the previous 3 h. This suggests that a simple dynamic map of scatterometer time change could be a useful ancillary dataset to dynamically adjust either (or both of) the PR and the 10H GHz surface emissivity, the latter of which is useful to adjust the other PMW emissivities via known cross-channel covariance matrices. More specifically, GPM and TRMM retrieval algorithms could consult this type of dynamic map to better adjust or modify the weighting of candidate solutions, to reflect more up-to-date surface conditions.

The analysis was done largely from a qualitative method, the rationale being that it is important to first identify and quantify the response of the surface to the precipitation, from the observations themselves, before moving forward into more detailed physical surface modeling and forward simulations. Many choices were not fully explored, such as the moving average median filtering and further discretization of the spatial and time scales of the previous rain accumulations, and no attempt was made to assign physical meanings to the land surface classes.

While not investigated, this type of analysis is applicable to observations from the recently activated (October 2014) Rapid Scatterometer (RapidScat) onboard the International Space Station, to apply to overland precipitation retrievals from the GMI and/or DPR sensors, thereby maintaining continuity between the 17+-yr TRMM data record and GPM.

Acknowledgments

This work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. OceanSat-2 data were provided courtesy of the Indian Space Research Organisation, and TRMM data via the NASA Precipitation Processing System (PPS). FJT and SD acknowledge support from the NASA Precipitation Measuring Mission (PMM) science team.

REFERENCES

REFERENCES
Aires
,
F.
,
C.
Prigent
,
F.
Bernardo
,
C.
Jiménez
,
R.
Saunders
, and
P.
Brunel
,
2011
:
A Tool to Estimate Land-Surface Emissivities at Microwave frequencies (TELSEM) for use in numerical weather prediction
.
Quart. J. Roy. Meteor. Soc.
,
137
,
690
699
, doi:.
Bartalis
,
Z.
,
K.
Scipal
, and
W.
Wagner
,
2006
:
Azimuthal anisotropy of scatterometer measurements over land
.
IEEE Trans. Geosci. Remote Sens.
,
44
,
2083
2092
, doi:.
Bhowmick
,
S. A.
,
R.
Kumar
, and
A. S. K.
Kumar
,
2014
:
Cross calibration of the OceanSAT-2 scatterometer with QuikSCAT scatterometer using natural terrestrial targets
.
IEEE Trans. Geosci. Remote Sens.
,
52
,
3393
3398
, doi:.
Brocca
,
L.
, and Coauthors
,
2014
:
Soil as a natural rain gauge: Estimating global rainfall from satellite soil moisture data
.
J. Geophys. Res. Atmos.
,
119
,
5128
5141
, doi:.
Colliander
,
A.
,
K.
McDonald
,
R.
Zimmermann
,
R.
Schroeder
,
J. S.
Kimball
, and
E. G.
Njoku
,
2012
:
Application of QuikSCAT backscatter to SMAP validation planning: Freeze/thaw state over ALECTRA sites in Alaska from 2000 to 2007
.
IEEE Trans. Geosci. Remote Sens.
,
50
,
461
468
, doi:.
Durden
,
S. L.
,
S.
Tanelli
, and
R.
Meneghini
,
2012
:
Using surface classification to improve surface reference technique performance over land
.
Indian J. Radio Space Phys.
,
41
,
403
410
.
Ferraro
,
R. R.
, and Coauthors
,
2013
:
An evaluation of microwave land surface emissivities over the continental United States to benefit GPM-era precipitation algorithms
.
IEEE Trans. Geosci. Remote Sens.
,
51
,
378
398
, doi:.
Glenn
,
E. P.
,
A. R.
Huete
,
P. L.
Nagler
, and
S. G.
Nelson
,
2008
:
Relationship between remotely-sensed vegetation indices, canopy attributes and plant physiological processes: What vegetation indices can and cannot tell us about the landscape
.
Sensors
,
8
,
2136
2160
, doi:.
Gohil
,
B. S.
,
R. M.
Gairola
,
A. K.
Mathur
,
A. K.
Varma
,
C.
Mahesh
,
R. K.
Gangwar
, and
P. K.
Pal
,
2013
:
Algorithms for retrieving geophysical parameters from the MADRAS and SAPHIR sensors of the Megha-Tropiques satellite: Indian scenario
.
Quart. J. Roy. Meteor. Soc.
,
139
,
954
963
, doi:.
Gopalan
,
K.
,
N. Y.
Wang
,
R.
Ferraro
, and
C.
Liu
,
2010
:
Status of the TRMM 2A12 land precipitation algorithm
.
J. Atmos. Oceanic Technol.
,
27
,
1343
1354
, doi:.
Hou
,
A. Y.
, and Coauthors
,
2014
:
The Global Precipitation Measurement (GPM) Mission
.
Bull. Amer. Meteor. Soc.
,
95
,
701
722
, doi: .
Huang
,
S.
,
L.
Tsang
,
E. G.
Njoku
, and
K. S.
Chan
,
2010
:
Backscattering coefficients, coherent reflectivities, and emissivities of randomly rough soil surfaces at L-band for SMAP applications based on numerical solutions of Maxwell equations in three-dimensional simulations
.
IEEE Trans. Geosci. Remote Sens.
,
48
,
2557
2568
, doi:.
Hunt
,
E. R.
,
L.
Li
,
M. T.
Yilmaz
, and
T. J.
Jackson
,
2011
:
Comparison of vegetation water contents derived from shortwave-infrared and passive-microwave sensors over central Iowa
.
Remote Sens. Environ.
,
115
,
2376
2383
, doi:.
Kimball
,
J. S.
,
K. C.
McDonald
,
S. W.
Running
, and
S. E.
Frolking
,
2004
:
Satellite radar remote sensing of seasonal growing seasons for boreal and subalpine evergreen forests
.
Remote Sens. Environ.
,
90
,
243
258
, doi:.
Kummerow
,
C. D.
,
S.
Ringerud
,
J.
Crook
,
D.
Randel
, and
W.
Berg
,
2011
:
An observationally generated a priori database for microwave rainfall retrievals
.
J. Atmos. Oceanic Technol.
,
28
,
113
130
, doi:.
L’Ecuyer
,
T. S.
, and
G. L.
Stephens
,
2002
:
An uncertainty model for Bayesian Monte Carlo retrieval algorithms: Application to the TRMM observing system
.
Quart. J. Roy. Meteor. Soc.
,
128
,
1713
1737
, doi:.
Li
,
L.
, and Coauthors
,
2010
:
WindSat global soil moisture retrieval and validation
.
IEEE Trans. Geosci. Remote Sens.
,
48
,
2224
2241
, doi:.
Lu
,
L.
,
H.
Guo
,
C.
Wang
, and
Q.
Li
,
2013
:
Assessment of the SeaWinds scatterometer for vegetation phenology monitoring across China
.
Int. J. Remote Sens.
,
34
,
5551
5568
, doi:.
Meneghini
,
R.
, and
J. A.
Jones
,
2011
:
Standard deviation of spatially averaged surface cross section data from the TRMM Precipitation Radar
.
IEEE Trans. Geosci. Remote Sens.
,
8
,
293
297
, doi:.
Meneghini
,
R.
,
T.
Iguchi
,
T.
Kozu
,
L.
Liao
,
K.
Okamoto
,
J. A.
Jones
, and
J.
Kwiatkowski
,
2000
:
Use of the surface reference technique for path attenuation estimates from the TRMM Precipitation Radar
.
J. Appl. Meteor.
,
39
,
2053
2070
, doi:.
Meneghini
,
R.
,
J. A.
Jones
,
T.
Iguchi
,
K.
Okamoto
, and
J.
Kwiatkowski
,
2004
:
A hybrid surface reference technique and its application to the TRMM Precipitation Radar
.
J. Atmos. Oceanic Technol.
,
21
,
1645
1658
, doi:.
Mladenova
,
I.
,
V.
Lakshmi
,
J. P.
Walker
,
D. G.
Long
, and
R.
De Jeu
,
2009
:
An assessment of QuikSCAT Ku-band scatterometer data for soil moisture sensitivity
.
IEEE Trans. Geosci. Remote Sens.
,
6
,
640
643
, doi:.
Munchak
,
S. J.
, and
G.
Skofronick-Jackson
,
2013
:
Evaluation of precipitation detection over various surfaces from passive microwave imagers and sounders
.
Atmos. Res.
,
131
,
81
94
, doi:.
Nghiem
,
S.
,
B. D.
Wardlow
,
D.
Allured
,
M. D.
Svoboda
,
D.
LeComte
,
M.
Rosencrans
,
S.
Chan
, and
G.
Neumann
,
2012
: Microwave remote sensing of soil moisture: Science and applications. Microwave Remote Sensing of Drought: Innovative Monitoring Approaches, B. D. Wardlow, M. C. Anderson, and J. P. Verdin, Eds., CRC Press, 197–226.
Norouzi
,
H.
,
W.
Rossow
,
M.
Temimi
,
C.
Prigent
,
M.
Azarderakhsh
,
S.
Boukabara
, and
R.
Khanbilvardi
,
2012
:
Using microwave brightness temperature diurnal cycle to improve emissivity retrievals over land
.
Remote Sens. Environ.
,
123
,
470
482
, doi:.
Petty
,
G. W.
,
2013
:
Dimensionality reduction in Bayesian estimation algorithms
.
Atmos. Meas. Tech. Discuss.
,
6
,
2327
2352
, doi:.
Prigent
,
C.
,
F.
Aires
, and
W. B.
Rossow
,
2006
:
Land surface microwave emissivities over the globe for a decade
.
Bull. Amer. Meteor. Soc.
,
87
,
1573
1584
, doi:.
Puri
,
S.
,
H.
Stephen
, and
S.
Ahmad
,
2011
:
Relating TRMM Precipitation Radar backscatter to water stage in wetlands
.
J. Hydrol.
,
401
,
240
249
, doi:.
Ringerud
,
S.
,
C.
Kummerow
,
C.
Peters-Lidard
,
Y.
Tian
, and
K.
Harrison
,
2014
:
A comparison of microwave window channel retrieved and forward-modeled emissivities over the U.S. southern Great Plains
.
IEEE Trans. Geosci. Remote Sens.
,
52
,
2395
2412
, doi:.
Seto
,
S.
, and
T.
Iguchi
,
2007
:
Rainfall-induced changes in actual surface backscattering cross sections and effects on rain-rate estimates by spaceborne precipitation radar
.
J. Atmos. Oceanic Technol.
,
24
,
1693
1709
, doi:.
Stephen
,
H.
,
S.
Ahmad
,
T. C.
Piechota
, and
C.
Tang
,
2010
:
Relating surface backscatter response from TRMM Precipitation Radar to soil moisture: Results over a semi-arid region
.
Hydrol. Earth Syst. Sci.
,
14
,
193
204
, doi:.
Teng
,
W. L.
,
J. R.
Wang
, and
P. C.
Doraiswamy
,
1993
:
Relationship between satellite microwave radiometric data, antecedent precipitation index, and regional soil moisture
.
Int. J. Remote Sens.
,
14
,
2483
2500
, doi:.
Turk
,
F. J.
,
Z. S.
Haddad
, and
Y.
You
,
2014a
:
Principal components of multifrequency microwave land surface emissivities. Part I: Estimation under clear and precipitating conditions
.
J. Hydrometeor.
,
15
,
3
19
, doi:.
Turk
,
F. J.
,
L.
Li
, and
Z. S.
Haddad
,
2014b
:
A physically based soil moisture and microwave emissivity data set for Global Precipitation Measurement (GPM) applications
.
IEEE Trans. Geosci. Remote Sens.
,
52
,
7637
7650
, doi:.
Wagner
,
W.
,
G.
Lemoine
,
M.
Borgeaud
, and
H.
Rott
,
1999
:
A study of vegetation cover effects on ERS scatterometer data
.
IEEE Trans. Geosci. Remote Sens.
,
37
,
938
948
, doi:.
Wang
,
L.
,
C.
Derksen
, and
R.
Brown
,
2008
:
Detection of pan-Arctic terrestrial snowmelt from QuikSCAT, 2000–2005
.
Remote Sens. Environ.
,
112
,
3794
3805
, doi:.
Wilks
,
D. S.
,
2011
: Statistical Methods in the Atmospheric Sciences. 3rd ed. Elsevier, 676 pp.
Yilmaz
,
M. T.
,
E. R.
Hunt
, and
T. J.
Jackson
,
2008
:
Remote sensing of vegetation water content from equivalent water thickness using satellite imagery
.
Remote Sens. Environ.
,
112
,
2514
2522
, doi:.
Zhang
,
J.
, and Coauthors
,
2011
:
National Mosaic and Multi-Sensor QPE (NMQ) system: Description, results, and future plans
.
Bull. Amer. Meteor. Soc.
,
92
,
1321
1338
, doi:.

Footnotes

This article is included in the Seventh International Precipitation Working Group (IPWG) Workshop special collection.