Special Sensor Microwave Imager Data Research Articles

Assimilation of SSMIS and ASCAT data and the replacement of highly uncertain estimates in the Environment Canada Regional Ice Prediction System

AbstractThis study describes the impact from three major modifications to an existing ice‐analysis system developed at Environment Canada. The analysis component of the Regional Ice Prediction System currently provides near real‐time gridded estimates of ice concentration for all ice‐affected areas around North America and Greenland, and is primarily aimed to satisfy the operational requirements of the Canadian Ice Service. The first modification is the assimilation of Special Sensor Microwave Imager/Sounder data from three satellite platforms to complement the already assimilated Special Sensor Microwave Imager data from one platform. The second change is the assimilation of ice‐concentration information derived from Advanced Scatterometer data. The third modification is to replace the ice concentration in the analysis with spatially interpolated values for all grid points where an estimated measure of uncertainty is above a specified threshold. Objective verification scores were computed from 1 year experiments spanning all of 2010 using independent verification data to evaluate the accuracy of the analyses. The incremental impact of adding each of the three modifications is examined along with the combined impact from the three modifications. It is demonstrated that the new version of the system produces consistently more accurate ice‐concentration analyses than the previous version, especially during the summer period and when the ice is refreezing.

What drives the southward drift of sea ice in the Sea of Okhotsk?

The Sea of Okhotsk is the southernmost sea-ice zone with sizable ice. It is widely believed that the prevailing northwesterly wind and the southward East Sakhalin Current (ESC) are the two main factors that drive the southward drift of sea ice. However, the relative contributions of these factors have not been understood. In this paper, by using the current and ice-drift data measured with the moored Acoustic Doppler Current Profiler, a 3-D ocean model simulation, objective analysis data of the wind, and satellite sea-ice data, we examine to what degree and how the ice drift is determined by the wind and ocean current. From a linear regression of the observed ice drift, ocean current, and wind, the wind-forced component of the ice drift was best fitted when sea ice is assumed to move with a speed of 1.6% of the geostrophic wind with a turning angle of 17.6° to the left of the wind. Such a relationship was adopted as the wind-drift component for all sea-ice pixels detected from Special Sensor Microwave Imager data. For the ocean-forced component of the ice drift, we adopted the current at 20m depth from a numerical model simulation that reproduces well the ESC and its variability. We then evaluated the sea-ice drift over 46–54°N during 1998–2005. For the southward drift of sea ice, the contribution of the wind component is found to be larger than the oceanic component, although the ocean contribution becomes larger, typically comparable to the wind contribution, near the coast and in the northern region where the ESC is stronger. We estimated the average annual cumulative southward ice transport to be 3.0±0.9×1011m3 at 53°N. This ice transport is comparable to the annual discharge of the Amur River. The ratio of wind to oceanic components in the transport is estimated to be ∼1.2–1.8. We also conducted ice-drift simulations based on the modeled current velocity and the assumed wind drift of 1.5% geostrophic wind with a turning angle of 15° to the left. The simulations reproduce well the ice drift north of 47°N but not south of 47°N, likely due to the poor representation of the current system at the latter, underevaluation of the wind factor near the ice edges, and the neglect of ice formation and melt.

Retrievals for the Rainfall Rate over Land Using Special Sensor Microwave Imager Data during Tropical Cyclones: Comparisons of Scattering Index, Regression, and Support Vector Regression

Abstract Tropical cyclones, also known as typhoons or hurricanes, are among the most devastating events in nature and often strike the western North Pacific region (including the Philippines, Taiwan, Japan, Korea, China, and others). This paper focuses on addressing the rainfall retrieval problem for quantitative precipitation forecast during tropical cyclones. In this study, Special Sensor Microwave Imager (SSM/I) data and Water Resources Agency (WRA) measurements of Taiwan were used to quantitatively estimate precipitation over the Tanshui River basin in northern Taiwan. Various retrievals for the rainfall rate over land are compared by five methods/techniques. They are the single-channel regression, multichannel linear regressions (MLR), scattering index over land approach (SIL), support vector regression (SVR), and the proposed SIL–SVR. This study collected 70 typhoons affecting the studied watershed over the past 12 years (1997–2008). The measurements of the SSM/I satellite comprise the brightness temperatures at 19.35, 22.23, 37.0, and 85.5 GHz. Overall, the results showed the approaches using the SVR and conjoined SVR and SIL performed better than regression and SIL methods according to their performances of the root-mean-square error (RMSE), bias ratio, and equitable threat score (ETS).

Quantifying the role of atmospheric rivers in the interior western United States

AbstractAtmospheric rivers (ARs) have increasingly been recognized for their contribution to high‐impact weather and climate variability. A recent investigation based on observations located primarily in lowland valleys and basins of the western United States suggests that 10–50% of the total cool season (November to April) precipitation between water years 1998 and 2008 occurred on the day of and day following AR landfall (hereafter the AR fraction), as identified using Special Sensor Microwave Imager data. However, these results are based only on ARs crossing the North American west coast between 32.5°N and 52.5°N, which excludes those crossing the west coast of the Baja Peninsula. Here, we identify ARs in the ERA‐Interim reanalysis and examine the AR fraction at high‐elevation observational sites and in the NOAA/CPC Unified Daily Precipitation Analysis. At high‐elevation snowpack telemetry sites, we find good agreement with the AR fraction obtained previously for valley and basin locations. We also show that including ARs crossing the west coast of the Baja Peninsula (as far south as 24°N) substantially increases the AR fraction over the southwestern United States. Copyright © 2012 Royal Meteorological Society

Intersensor Calibration Between F13 SSMI and F17 SSMIS for Global Sea Ice Data Records

An intercalibration between F13 Special Sensor Microwave Imager (SSMI) and F17 Special Sensor Microwave Imager Sounder (SSMIS) sea ice extents and areas for a full year of overlap is undertaken preparatory to extending the 1979-2007 National Aeronautics and Space Administration (NASA) Goddard Space Flight Center NASA Team algorithm time series of global sea ice extents and areas. The 1979-2007 time series was created from Scanning Multichannel Microwave Radiometer (SMMR) and SSMI data. After intercalibration, the yearly mean F17 and F13 difference in northern hemisphere (NH) sea ice extents is - 0.0156%, with a standard deviation (SD) of the differences of 0.6204%, and the yearly mean difference in NH sea ice areas is 0.5433%, with an SD of 0.3519%. For the southern hemisphere, the yearly mean difference in sea ice extents is 0.0304% ±0.4880%, and the mean difference in sea ice areas is 0.1550% ±0.3753%. This F13/F17 intercalibration enables the extension of the 29-year 1979-2007 SMMR/SSMI sea ice time series for as long as there are stable F17 SSMIS brightness temperatures available.

Spatiotemporal variations of snowmelt in Antarctica derived from satellite scanning multichannel microwave radiometer and Special Sensor Microwave Imager data (1978–2004)

We derived the extent, onset date, end date, and duration of snowmelt in Antarctica from 1978 to 2004 using satellite passive microwave scanning multichannel microwave radiometer (SMMR) and Special Sensor Microwave Imager (SSM/I) data. A wavelet‐transform‐based method was developed to determine and characterize melt occurrences. About 9–12% of the Antarctic surface experiences melt annually. This is more than twice the surface melt extent measured in Greenland. Seasonally, surface melt primarily takes place in December, January, and February and peaks in early January. Regression analysis over the 25 year period of study reveals a negative interannual trend in surface melt. Nevertheless, the trend inference is not statistically significant. Large year‐to‐year fluctuations characterize the interannual variability. Extremely high melt occurred in the 1982/1983 and 1991/1992 summers, while extremely weak melt occurred in the 1999/2000 summer. A strong correlation with air temperature suggests that the melt index can serve as a diagnostic indicator for regional temperature variations. Periodic melting has been observed over Ross Ice Shelf, Ronne‐Filchner Ice Shelf, the West Antarctic ice streams, and outlet glaciers in the Transantarctic Mountains. The Antarctic Peninsula, West Ice Shelf, Shackleton Ice Shelf, Amery Ice Shelf, and the ice shelf along the Princess Ragnhild Coast experienced the most persistent and intensive melt and should be closely monitored for their stability in the future, given the recent disintegration of the Larsen Ice Shelf A and B.

The Impact of Amazonian Deforestation on Dry Season Rainfall

Many modeling studies have concluded that widespread deforestation of Amazonia would lead to decreased rainfall. Geosynchronous visible and infrared satellite data over southwest Brazil are analyzed with respect to percent cloudiness, and rain estimates are analyzed from both the Tropical Rainfall Measuring Mission and Special Sensor Microwave Imager. The studies conclude that in the dry season, when the effects of the surface are not overwhelmed by synoptic-scale weather disturbances, shallow cumulus cloudiness, deep convective cloudiness, and rainfall occurrence all are larger over the deforested and nonforested (savanna) regions than over areas of dense forest. This paper speculates that this difference is in response to a local circulation initiated by the differential heating of the region's varying forestation. Analysis of the diurnal cycle of cloudiness reveals a shift in the onset of convection toward afternoon hours in the deforested and toward the morning hours in the savanna regions when compared to the neighboring forested regions. Analysis of 14 years of monthly estimates from the Special Sensor Microwave Imager data revealed that in August there was a pattern of higher monthly rainfall amounts over the deforested region. Analysis of available rain gauge data showed an increase in regional rainfall since deforestation began around 1978.

Large‐scale drift of Arctic Sea ice retrieved from passive microwave satellite data

A method of determining the large‐scale sea ice drift using 85.5 GHz Special Sensor Microwave Imager data are presented. A cross‐correlation method is applied to sequential images of gridded data covering the entire Arctic. Individual correlation results are validated with ice velocities derived from buoy data. The satellite‐derived mean drift values and the variabilities of the ice drift correspond closely with the buoy data. Similarly, time series of buoy data and associated satellite data are in good agreement even over large time periods. An example of a satellite‐retrieved 3 day mean drift field demonstrates the potential of the method for providing large‐scale ice circulation patterns. Mean drift fields of the winter periods 1987–1988 and 1992–1993 indicate a considerable interannual variability of the sea ice drift pattern in the Arctic Ocean. The Arctic region is divided into seven larger areas, and the area flux between these regions has been derived. The Kara Sea and the Laptev Sea show the largest area ice export with 0.02 and 0.015 km2 s−1, respectively. The central Arctic export through Fram Strait amounts to 0.12 Sv during the winter of 1992–1993 with a maximum of 0.15 Sv in January.

Data assimilation in sea-ice monitoring

AbstractThe detection of small changes in concentration or thickness in the Arctic or Antarctic ice cover is an important topic in the current global-climate-change debate. Change detection using satellite data alone requires rigorous error analysis for their derived ice products, including inter-satellite validation for long time series. All models of physical processes are only approximations, and the best models of complicated physical processes have errors and uncertainties. A promising approach is data assimilation, combining model, in situ data and satellite remote-sensing data. Sea-ice monitoring from satellite, ice-model estimates, and the potential benefit of combining the two are discussed in some detail. In a case-study we demonstrate how the sea-ice backscatter for the Beaufort Sea region was derived using a backscattering model in combination with an ice model. We conclude that, for data assimilation, the first steps include the use of simple models, moving, with success at this level, to progressively more complex models. We also recommend reconfiguring the current remote-sensing data to include precise time tags with each pixel. For example, the current Special Sensor Microwave Imager data might be reissued in a time-tagged orbital (or gridded) format as opposed to the currently available daily averaged gridded data. Finally, error statistics and quality-control information also need to be readily available in a form useful for assimilation. The effectiveness of data-assimilation techniques is directly linked to the availability of data error statistics.

Arctic sea ice drift from wavelet analysis of NSCAT and special sensor microwave imager data

Wavelet analysis of NASA scatterometer (NSCAT) backscatter and DMSP special sensor microwave imager (SSM/I) radiance data can be used to obtain daily sea ice drift information for the Arctic region. This technique provides improved spatial coverage over the existing array of Arctic Ocean buoys and better temporal resolution over techniques utilizing data from satellite synthetic aperture radars. Comparisons with ice motion derived from ocean buoys give good quantitative agreement. Both comparison results from NSCAT and SSM/I are compatible, and the results from NSCAT can definitely complement that from SSM/I when there are cloud or surface effects. Then, three sea ice drift daily results from NSCAT, SSM/I, and buoy data can be merged as a composite map by some data fusion techniques. The ice flow streamlines are highly correlated with surface air pressure contours. Examples of derived ice drift maps in December 1996 illustrate large‐scale circulation reversals over a period of 4 days. These calibrated/validated results indicate that NSCAT, SSM/I merged daily ice motions are suitably accurate to identify and closely locate sea ice processes and to improve our current knowledge of sea ice drift and related processes through the data assimilation of ocean‐ice numerical model.

Monitoring of boreal forests with multitemporal special sensor microwave imager data

The feasibility of multitemporal special sensor microwave imager (SSM/I) data for monitoring boreal forests was evaluated. The parameters of interest were forest coverage fraction and forest stem volume (biomass). The employed test sites covered almost the whole of Finland. Two measurement periods were used: July through September in 1993 and January through February in 1994. The apparent emissivities of various land cover types were determined with the mixed pixel approach under summer and winter conditions. The aim was to define the dominating factors on emissivity under winter and summer conditions. The mixed pixel approach was tested for the estimation of forest coverage fraction. The results with multitemporal SSM/I data show that the pixel‐wise fractions of water, nonforested, and forested area can be estimated with rms errors of around 10 percent units. The correlation between the estimates and the ground truth was over 0.85. A new inversion method for stem volume estimation was presented. At SSM/I frequencies the forest canopy and the snow‐covered ground dominate the emissivity behavior under winter conditions in the boreal forest zone. The method is based on the fact that the emissivity of forest canopy is close to 1, while that of dry snow cover is relatively low. The results with the method showed promising accuracies when wintertime SSM/I data were employed. The rms error was from 13 to 19 m3/ha per pixel (25 km by 25 km), which was 15–16% of the mean stem volume. In the test area the stem volume ranged from 40 to 160 m3/ha per pixel.

Special sensor microwave imager data in field experiments: FIFE-1987

Estimation of regional soil moisture is of importance to the hydrological modelling community for validating the accuracy of model predictions and to the climate community for use in regional and global climate models. Soil moisture estimation is done in two ways, viz., estimation using models and estimation using satellite data. Both these methods have to be corroborated using in situ field data. The present paper examines the feasibility of using observations of brightness temperature from the Special Sensor Microwave Imager (SSM/I) for small scale/catchment scale studies. The simulations of the 19 and 37 GHz brightness temperatures are carried out for a period of 13 days between 6 and 21 August 1987. The comparisons between the simulated and the observed brightness temperatures are good and the errors can be explained based on the atmospheric and surface conditions. Sensitivity analysis of the brightness temperatures to leaf area index, soil moisture and soil temperatures shows interesting characteristics. The errors in the simulated brightness temperatures can be explained by variability in the input parameters. This study shows that such an approach holds promise for local and regional studies of the land surface.

Analysis of merged SMMR‐SSMI time series of Arctic and Antarctic sea ice parameters 1978–1995

The most consistent means of investigating the global sea ice cover is by satellite passive microwave sensors, as these are independent of illumination and cloud cover. The Nimbus 7 Scanning Multichannel Microwave Radiometer (SMMR) and the Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave Imager (SSMI) provide information on the global sea ice cover from 1978 to present. The two instruments flew simultaneously during a 6‐week overlap period in July and August 1987, thus enabling intercomparison of the two sensors. Brightness temperatures are corrected for instrument drift and calibration differences in order to produce continuous time series of monthly averaged Arctic and Antarctic sea ice extent and sea ice area through the use of the NOrwegian Remote Sensing EXperiment (NORSEX) algorithm, which relates brightness temperatures to ice concentration. Statistical analysis on the time series estimates the decreases in Arctic ice extent and ice area to be 4.5% and 5.7%, respectively, during the 16.8‐year observation period. The overall trends established here serve to better define and strengthen earlier assertions of a reduced ice cover, based on analysis of SMMR and SSMI data taken separately. These results are consistent with GCM simulations that suggest retreat of the sea ice cover under global warming scenarios.

TOPEX/POSEIDON microwave radiometer performance and in‐flight calibration

Results of the in‐flight calibration and performance evaluation campaign for the TOPEX/POSEIDON microwave radiometer (TMR) are presented. Intercomparisons are made between TMR and various sources of ground truth, including ground‐based microwave water vapor radiometers, radiosondes, global climatological models, special sensor microwave imager data over the Amazon rain forest, and models of clear, calm, subpolar ocean regions. After correcting for preflight errors in the processing of thermal/vacuum data, relative channel offsets in the open ocean TMR brightness temperatures were noted at the ≈1 K level for the three TMR frequencies. Larger absolute offsets of 6–9 K over the rain forest indicated a ≈5% gain error in the three channel calibrations. This was corrected by adjusting the antenna pattern correction (APC) algorithm. A 10% scale error in the TMR path delay estimates, relative to coincident radiosondes, was corrected in part by the APC adjustment and in part by a 5% modification to the value assumed for the 22.235 GHz water vapor line strength in the path delay retrieval algorithm. After all in‐flight corrections to the calibration, TMR global retrieval accuracy for the wet tropospheric range correction is estimated at 1.1 cm RMS with consistent performance under clear, cloudy, and windy conditions.

The total mass of the atmosphere

Accurate but approximate formulae for determining the mass of the atmosphere in terms of the surface pressure ps are derived and applied to globally analyzed data from the European Centre for Medium‐Range Weather Forecasts (ECMWF) for 1985 through 1993. The formulae take into account effects of the shape of the Earth and variations in gravity with latitude and height. Variations in total mass occur because of changes in the water vapor loading of the atmosphere. Independent computations are made of the surface pressure due to water vapor pw, which is proportional to the precipitable water, using the ECMWF analyses of specific humidity. Spurious trends in both the mass of dry air and the atmospheric moisture are found to arise from changes in the analysis system at ECMWF, confounding attempts to seek real trends associated with climate change. For the recent 4‐year period 1990 to 1993 the mean annual ps was 984.76 mbar with a maximum in July of 984.98 mbar and a minimum in December of 984.61 mbar which correspond to a total mean mass of the atmosphere of 5.1441×1018 kg with a range of 1.93×1015 kg throughout the year associated with changes in water vapor in the atmosphere. The global mean pw for 1985–1993 is 2.58 mbar, but values are 5 to 10% lower after mid‐1992. Using the Special Sensor Microwave Imager data to make adjustments, the best estimate of the annual global pw is 2.4 mbar, corresponding to ∼2.5 cm of precipitable water. The total atmospheric moisture as given by pw varies with an annual cycle range of 0.36 mbar, a maximum in July, and a minimum in December. Thus the mean mass of water vapor is 1.25×1016 kg and the dry air mass is 5.132×1018 kg, corresponding to a mean surface pressure of 982.4 mbar. Overall uncertainties are ∼0.1 mbar or 0.5×1015 kg in total mass and about double those values for atmospheric moisture content. As well as the global means, hemispheric mean values and meridional profiles of ps and pw are presented for the mean annual cycle and as latitude‐time series to show the interannual and longer‐term variability.

Global precipitation estimations using Defense Meteorological Satellite Program F10 and F11 special sensor microwave imager data

F10 and F11 satellites from the Defense Meteorological Satellite Program currently observe the earth‐atmosphere system four times a day, at 0530, 1030, 1730, and 2230 local solar time. This study uses the special sensor microwave imager data from both satellites to retrieve precipitation over land and ocean and presents some preliminary results of the spatial and temporal variations of rainfall during a month. Diurnal variations of precipitation over the globe are analyzed for August 1993 using measurements from both satellites. Over most oceans, precipitation displays the well‐known morning maximum. However, over land, precipitation is highly variable for the study period. Error analyses indicate that over ocean the monthly precipitation estimates from the F11 satellite alone may result in 20–25% errors due to insufficient temporal sampling. This error increases over some land regions to as much as 50–70%. This study also analyses the difference between the monthly rainfall estimates using a sample‐averaging method and that using a lognormal probability density function. It is found that an overall difference over the globe is small with less than 5%.

ECMWF and SSM/I Global Surface Wind Speeds

Monthly mean, 2.5 deg - x 2.5 deg-resolution, 10-m height wind speeds from the Special Sensor Microwave Imager (SSMI) instrument and the European Center for Medium-Range Weather Forecasts (ECMWF) forecast-analysis system are compared between 60 deg S and 60 deg N during 1988-1991. The SSMI data were uniformly processed while numerous changes were made to the ECMWF forecast-analysis system. The SSMI measurements, which were compared with moored-buoy wind observations, were considered to be a reference data set to evaluate the influence of the changes made to the ECMWF system upon the ECMWF surface wind speed over the ocean. A demonstrable yearly decrease of the difference between SSMI and ECMWF wind speeds occurred in the 10 deg S - 10 deg N region, including the 5 deg S - 5 deg N zone of the Pacific Ocean, where nearly all of the variations occurred in the 160 deg E - 160 deg W region. The apparent improvement of the ECMWF wind speed occurred at the same time as the yearly decrease of the equatorial Pacific SSMI wind speed, which was associated with the natural transition from La Nina to El Nino conditions. In the 10 deg S - 10 deg N tropical Atlantic, the ECMWF wind speed had a 4-year trend, which was not expected nor was it duplicated with the SSMI data. No yearly trend was found in the difference between SSMI and ECMWF surface wind speeds in middle latitudes of the northern and southern hemispheres. The magnitude of the differences between SSMI and ECMWF was 0.4 m s^(-1) or 100 percent larger in the northern than in the southern hemisphere extratropics. In two areas (Arabian Sea and North Atlantic Ocean) where ECMWF and SSMI wind speeds were compared to ship measurements, the ship data had much better agreement with the ECMWF analyses compared to SSMI data. In the 10 deg S - 10 deg N area the difference between monthly standard deviations of the daily wind speeds dropped significantly from 1988 to 1989, but remained constant at about 30 percent for the remaining years. y.

On the problem of measuring interannual wind speed variations using SSMI data

The first Special Sensor Microwave Imager (SSMI) was launched on the Defence Meteorological Satellite Program (DMSP) F8 spacecraft in July 1987, and wind speed was no longer retrieved after December 1991. A second SSMI was launched on DMSP F10 in December 1990. Interpretation of the 1987–1993 (or longer) SSMI wind speed time series is dependent upon the space and time characteristics of the differences between F8 and F10 SSMI measurements. The 10°‐zonal averaged monthly mean F8–F10 wind speed difference was negative (positive) for wind speeds less (greater) than 7.9 m s−1, reaching −0.43 (0.32) m s−1 at 5 (10) m s−1. Between 60°S and 60°N the 10°‐zonal averaged monthly mean F8‐F10 wind speed bias was greater than ±0.5 m s−1 on several occasions. From 60°S–60°N the 1991 average value of the monthly mean root‐mean‐square difference between daily F8 and F10 wind speeds in 10°‐longitudinal bands was 2.0 m s−1. In the 60°S–60°N region, about 50% of the daily F8 and F10 wind speed differences was caused by measurement non‐simultaneity and about 50% of the difference was attributed to other factors, such as instrument noise and the different azimuthal orientations of each SSMI. Caution is advised in the interpretation of large‐scale monthly mean analyses obtained from a combined F8 and F10 wind speed time series.

Ocean surface wind retrievals from special sensor microwave imager data with neural networks

Several fully connected, feed forward neural networks trained on a set of special sensor microwave imager examples matched with buoy winds have yielded retrieval accuracies considerably better than those achieved by the current operational method. Equations and coefficients for using two of these networks, each with four input brightness temperatures and a hidden layer containing two neurodes, are given for implementation in wind retrieval codes. The first demonstrated an rms retrieval error of 1.41 m/s at a reference height of 19.5 m using an independent data set representing clear sky conditions. The second network yielded rms retrieval accuracies of 2.39 m/s under adverse weather conditions. This represents a factor of more than 2 improvement over the alternate algorithms that were examined for nonclear conditions.