CORRELATION BETWEEN ATMOSPHERIC COMPOSITION AND VERTICAL STRUCTURE AS MEASURED BY THREE GENERATIONS OF HYPERSPECTRAL SOUNDERS IN SPACE Nadia Smith 1, Elisabeth Weisz 1, and Allen Huang 1 1 Space Science and Engineering Center, University of Wisconsin-Madison, USA Abstract Climate change is broadly understood by how atmospheric composition affects atmospheric structure. The assessment of climate change depends on decadal measurements of atmospheric parameters (e.g., temperature, water vapor and trace gases) that are coincident in and consistent across spacetime. The difficulty in assembling such a data record remains one of the dominant constraints. Hyperspectral sounders have been in polar-orbit since 2002. They have the spectral capability to measure both atmospheric structure and composition at a consistent accuracy, daily, across the globe. In this paper we analyze five years of measurements (2008 2012) from hyperspectral sounders, AIRS (Atmospheric Infrared Sounder) and IASI (Infrared Atmospheric Sounding Interferometer). We aim to investigate global patterns of temperature (T) and ozone (O 3 ) as measured by each instrument. Our focus is on determining the extent to which these instruments agree in their depiction of parameter distribution and correlation across space-time. Once established, we will include measurements from CrIS (Cross-track Infrared Sounder), launched in 2011. We aim to demonstrate the value hyperspectral sounders add to climate change assessment and the continuity that can be established among three generations of hyperspectral sounders in space. 1 INTRODUCTION Unlike ground-based in situ networks, measurements from instruments in low-earth orbit achieve and maintain consistent and unbiased sampling frequencies. Consistency in space-time is a vital attribute of any dataset if it is to be used in the analysis and monitoring of geophysical change, such as weather and climate. A number of different instruments have been designed for application in Earth system science from space. These include, but are not limited to, visible-range imagers, broad-band and hyperspectral infrared sounders, as well as active sensors such as lidar. In this paper our focus is on investigating the patterns and magnitude of large-scale, long-term geophysical change as measured by two different hyperspectral infrared sounders; they are, AIRS (Atmospheric Infrared Sounder) and IASI (Infrared Atmospheric Sounding Interferometer). Hyperspectral sounders measure the infrared spectrum at intervals narrow enough to sustain a high information content about the vertical atmospheric state (e.g. T and humidity profiles, column amounts of trace gas species, as well as cloud and Earth surface parameters). These instruments are unlike any other given the range and quality (accuracy and precision) of measurable parameters they permit. Four hyperspectral infrared sounders are operational in low-earth orbit at present; (i) AIRS, launched in 2002 on Aqua with a ~13h30 local overpass time, (ii) IASI launched in 2006 on Metop-A, (iii) and again in 2012 on Metop-B with a ~09h30 local overpass time, and lastly (iv) the Cross-track Infrared Sounder (CrIS) launched on Suomi-NPP in 2011 on an orbit following close to AIRS in time. Much effort has gone into developing methods for retrieving geophysical information from hyperspectral radiance measurements. The focus has been on maximizing the signal-to-noise ratio (SNR) for quality real-time weather analysis. With more than a decade of these space-borne measurements available, it becomes possible to explore the value they can add to long-term climate analysis. In addition to maintaining a high SNR, retrieval methods need to be fast and stable enough for the vast amounts of data processing required in such applications. 1
The goal of this paper is to study five years (2008 2012) of global, stratospheric T and O 3 anomalies as measured by AIRS and IASI, respectively. We do not aim to quantify instrument differences, per se. Instead, we aim to investigate relative anomalies in T and O 3 in order to measure the strength in agreement between AIRS and IASI measurements. Our focus is on determining similarities in geophysical parameters across space-time, irrespective of instrument type and orbit. Faced with continual technological adjustments, determining the degree to which past, present and future spacebased measurements agree, regardless of changes in instrument design, becomes an important consideration if the use of satellite measurements in climate change assessment is to be considered. 2 METHODS The Dual-Regression (DR) method (Smith et al. 2012, Weisz et al. 2013) was developed to retrieve geophysical parameters from radiance measurements of all four operational hyperspectral sounders (Section 1). Using a single method for parameter retrieval from multiple instruments promotes comparison since it reduces uncertainties due to instrument-specific differences. It is available as a stand-alone open-source software package and is distributed under the University of Wisconsin- Madison Community Software Processing Package, CSPP (available online: http://cimss.ssec.wisc.edu/cspp/). The DR method retrieves profiles of T, humidity and O 3 at 101 vertical layers, cloud and surface properties, as well as column densities of CO, CO 2 and CH 4. This method, therefore, provides information about the vertical atmospheric structure and its composition with consistent accuracy on a global scale. For multi-instrument comparison as well as analysis at a scale suitable for climate research, the measurements had to be projected to an instrument-neutral, regularized grid. For this we used the Space-Time Gridding (STG) method (Smith et al. 2013). Despite using a single retrieval method, a number of instrument differences remain that prevent instantaneous comparisons. These include sampling time (AIRS samples in the PM and IASI in the AM) and spatial distribution (due to differences in orbit and scanning configuration). STG ingests measurements from any source and projects them onto a regularized equal-angle time-aggregate grid, the resolution of which is application-specific (or user-determined). It does this in three steps; (1) filter and bin all measurements per day to a regular grid, (2) calculate a daily statistical aggregate given a minimum sample-size test, and (3) stack daily grids together (for a month, season or year) and calculate a time-aggregate per grid cell. The latter can be any statistical metric (or combination thereof) that helps characterize the parameter at hand. Trends were calculated per grid cell as the slope of the line fit to the data points across five years, 2008 to 2012. 3 DATA Geophysical parameters (Level 2) were retrieved from calibrated geolocated radiance measurements (Level 1B) with the DR method for each instrument, AIRS and IASI. The geophysical parameters of focus were T and O 3 at two different layers in the atmosphere; 10 hpa and 100 hpa. These parameters were projected to a regularized global grid (Level 3) using STG. The spatial resolution was 1-degree, and monthly aggregates were assembled as averages; i.e., a monthly average as the average of daily averages. Level 1B to Level 3 processing was done for four months per year (February, May, August, November) from 2008 to 2012. For each instrument per month per parameter, the result was a 2-dimensional latitude/longitude grid of averages. In addition to monthly averages, a 5-year average for each of the four months was calculated per instrument. E.g., For AIRS (IASI) a 5-year average of February was calculated as the average of all daily averages in the month of February from 2008 to 2012. Relative monthly anomalies (in units, %) were calculated for O 3 by subtracting the 5-year average from the monthly average, divided by the 5-year average. For temperature, on the other hand, the absolute monthly anomalies (in units, K) were calculated. This simply means the subtraction of the 5-year average from the monthly average (no division). The monthly anomalies were calculated for each instrument individually, since the focus was on measurement similarities and not instrument differences. By subtracting the AIRS (IASI) 5-year average from the AIRS (IASI) monthly average, 2
systematic sampling bias due to differences in time of measurement (for example) was removed from the analysis, thus allowing a comparison of measurements from different sources. 4 RESULTS AND DISCUSSION In this section, we present and discuss the global averages and monthly anomalies of geophysical parameters, T and O 3, as retrieved from two different hyperspectral sounders in low-earth orbit; AIRS on a PM orbit and IASI on an AM orbit. Statistical analysis on a regularized grid (Section 3) allows comparison of these results, despite. With this, we aim to contribute (and follow on) to analysis presented by Steinbrecht et al. (2006, 2009). They documented variations in upper stratospheric O 3 and T (and their correlation) on a decadal time scale with the goal to describe agreement among a myriad of data sources, e.g., GOMOS (global ozone monitoring by occultation of the stars), SCIAMACHY (Scanning Imaging Absorption Spectrometer for Atmospheric Chartography), NCEP (National Centers for Environmental Prediction) and ECMWF (European Centre for Mid-range Weather Forecasting). They found that O 3 and T anomalies compared well among the various data sets with differences usually smaller than 5% for O 3 and 3 K for T. Most of the differences they observed could be explained by differences in sampling frequency. The global 5-year average of O 3 as measured by AIRS is presented in Figure 1 below. Seasonal patterns are depicted at two atmospheric layers, 10 hpa (left) and 100 hpa (right). A clear difference is observed between the month of February (top) and August (bottom). Figure 1: Seasonality of Ozone (O 3) at two atmospheric pressure layers as depicted by the hyperspectral sounder, AIRS, on a 1 x 1 degree equal-angle grid. (a) February average at 10 hpa, (b) February average at 100 hpa, (c) August average at 10 hpa and, (d) August average at 100 hpa. Following this, in Figure 2 global averages of T (top) and O 3 (bottom) at 100 hpa is presented. This comparison helps shed light on what is broadly understood in climate change as a strong correlation between atmospheric composition and atmospheric structure. Here (Figure 2) the correlation between O 3 and T is observed, such that a decrease in stratospheric O 3 leads to a cooler stratosphere (and warmer troposphere not shown). Despite instrument differences, the 5-year averages of these parameters compare well between the two instruments, AIRS and IASI. 3
Figure 2: Monthyly average of Temperature (top, a & b) and Ozone (bottom, c & d) at 100 hpa as measured by AIRS (left, a & c) and IASI (right, b & d) Zonal averages of O 3 anomalies are compared between AIRS and IASI per month across five years in the upper (Figure 3a) and lower (Figure 3b) stratosphere. A zone, in this instance, is considered to be all grid cells within the latitude range of 5 North and 5 South. Figure 3: Zonal averages (5 North to 5 South) of relative Ozone (O 3) anomalies at (a) 10 hpa and (b) 100 hpa as measured by AIRS (orange) and IASI (blue). In the upper stratosphere (10 hpa), zonal anomalies for O 3 (figure 3a) vary less than 5% from month to month. Instrument differences are on average less than 0.3%. In the lower stratosphere (100 hpa) 4
O 3 monthly anomalies show greater variation per month. This raises questions as to the contribution from seasonal pollution events. On average, instrument differences are less than 3% at this height. This is well within the range reported by Steinbrecht et al. (2009). The differences in instrument agreement with height could be attributed to variations in instrument sensitivity to O 3 with height. The peak of the O 3 weighting function is at ~20 hpa. To investigate the correlation between atmospheric structure and composition, zonal averages of T and O 3 anomalies are compared for each instrument at 100hPa in Figure 4 below (IASI in blue and AIRS in orange). These results confirm the strong correlation observed in Figure 2 between T and O 3. Figure 4: Comparing zonal averages (5 North to 5 South) of absolute T [K] and relative O 3 [%] anomalies at 100 hpa as measured by (a) IASI and (b) AIRS. Figure 5: Five-year trend (2008 2012) in O 3 mixing ratio [ppmv] for February at 10 hpa (top, a & b) and 100 hpa (bottom, c & d) as measured by AIRS (left, a & c) and IASI (right, b & d) 5
The global five-year trend in O 3 mixing ratio (ppmv) for the month of February is depicted at two different levels, 10 hpa and 100 hpa, in Figure 5. A comparison is made between AIRS and IASI. Note the strong agreement between the trends given measurements from two different instruments. A clear height dependence is evident. Further studies will focus on characterizing these patterns and establish the basis for similarities and differences observed. 5 CONCLUSIONS AND FUTURE RESEARCH Change assessment depends on decadal measurements of atmospheric parameters that are coincident and consistent in space-time. As the availability of and confidence in space-borne measurements continue to grow, it becomes important to understand how they can contribute to existing efforts in climate research. With their unprecedented consistency in quality and sampling frequency, they offer great advances over geophysical models and in situ measurement networks. While still preliminary, these results demonstrate the close agreement that can be achieved when two different hyperspectral datasets are compared. Hyperspectral sounders have been in low-earth orbit since 2002. Presently there are four in operation. Together they promise to span decades of measurements about atmospheric composition and structure, alike. We are working to repeat this study across the full hyperspectral record, using all four instruments in an attempt to determine continuity among different generations of instruments and demonstrate their value in long-term largescale studies. REFERENCES Smith, N., W. P. Menzel, E. Weisz, A. Heidinger and Baum, B. A. (2013) A uniform space-time gridding algorithm comparison of satellite data products: Characterization and sensitivity studies. J. Appl. Meteor. Clim., 52, pp 255 268. Smith, W. L., E. Weisz, S. Kirev, D. K. Zhou, Z. Li, and Borbas, E.E. (2012) Dual-Regression Retrieval Algorithm for Real-Time Processing of Satellite Ultraspectral Radiances. J. Appl. Meteor. Clim., 51(8), pp 1455 1476. Steinbrecht, W., et al. (2006) Long-term evolution of upper stratospheric ozone at selected stations of the Network for the Detection of Stratospheric Change (NDSC), J.G.R. 111, doi:10.1029/2005jd006454. Steinbrecht, W., et al. (2009) Ozone and temperature trends in the upper stratosphere at five stations of the Network for the Detection of Stratospheric Change (NDSC), Inter. J. Rem. Sen., 30(15 16), pp 3875 3886. Weisz, E., W. L. Smith and Smith, N. (2013), Advances in simultaneous atmospheric profile and cloud parameter regression based retrieval from high-spectral resolution radiance measurements, J.G.R.-Atmospheres, 118, pp 6433 6443. 6