1. Introduction
Polar wind analyses for numerical weather prediction have long been hampered by a lack of wind observations in polar regions (e.g., Fig. 1). Radiosonde, pilot balloons, or profiler data do not cover the Arctic Ocean, aircraft reports rarely reach latitudes north of 70°N or south of 60°S, and atmospheric motion vectors (AMVs) derived from geostationary satellite data are not available north of 60°N or south of 60°S. Infrared or microwave radiances from polar-orbiting satellites provide some indirect information on the polar wind field, but they remain difficult to use in polar regions because of poor discrimination between clouds and ice or poor representation of the surface emissivity. It is thus not surprising that independent rawinsonde data from Arctic field experiments, for instance, indicate biases in the polar wind fields of reanalyses from the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) or the European Centre for Medium-Range Weather Forecasts (ECMWF; e.g., Francis 2002).
Recently, a new satellite-derived wind product that is capable of providing unprecedented coverage of the polar wind fields has become available (e.g., Key et al. 2003, hereinafter KEY). These winds are derived by tracking structures in three successive 100-min swaths from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument flown on the polar-orbiting Terra and Aqua satellites. The method is based on the established procedures used to derive wind observations from image sequences from geostationary satellites (e.g., Nieman et al. 1997; Velden et al. 1997). The quality of these wind observations is similar to or slightly poorer than that of AMVs from geostationary satellite data, except at lower levels where the MODIS winds appear considerably poorer (KEY).
Initial assimilation studies at ECMWF and the National Aeronautics and Space Administration (NASA) Data Assimilation Office (DAO) indicated substantial gains in forecast skill when these winds are assimilated within three-dimensional variational data assimilation (3DVAR) systems (KEY; Bormann et al. 2002). Assimilation of these winds also had a notable impact on the mean polar wind analysis. In the present study we report on the first four-dimensional variational data assimilation (4DVAR) trials with the new MODIS winds over an extended period.
The structure of the paper is as follows. First we give an overview of the MODIS winds data and the assimilation experiments used. We then provide results from our assimilation trial in terms of the analysis impact, the forecast impact, and an illustrative forecast example. Conclusions and a discussion are provided in the last section.
2. Data and experiments
The derivation of MODIS winds is described in detail in KEY. Feature tracking is possible with MODIS data in an area north of 60°N and south of 60°S. Cloud features are tracked in the infrared (IR) window band at 11 μm and water vapor (WV) features are tracked in the 6.7-μm band. Wind vector heights are assigned by first determining a representative temperature of the feature using either the IR or the WV-intercept method (e.g., Schmetz et al. 1993), using forecast data from the U.S. Navy Operational Global Atmospheric Prediction System (NOGAPS) model at 1.0° spatial resolution and 19 vertical levels. A test dataset of Terra–MODIS winds has been prepared by the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin—Madison for the period 5 March–3 April 2001. Near–real time Terra–MODIS winds have been available since June 2002, and Aqua–MODIS winds have been available in near–real time since December 2002. Because of a combination of occasional late availability of the raw MODIS data and the processing requirements at CIMSS, real-time MODIS winds are currently available only with a highly variable delay (5– 15 h), so that only about 60% of the winds arrive within the operational data cutoff times at ECMWF.1 At the time of writing, the operational cutoff times for ECMWF are 0830 UTC for observations between 1500 and 2100 UTC the previous day, 0900 UTC for 2100– 0300 UTC observations, 1900 UTC for 0300–0900 UTC observations, and 1915 UTC for 0900–1500 UTC observations.
The impact experiments described herein employ ECMWF's 4DVAR system similar to the operational configuration (Rabier et al. 2000; Klinker et al. 2000). The model resolution for these experiments is T511 (≈40 km) with 60 levels in the vertical. Twelve-hourly incremental 4DVAR analyses are performed at T159 (≈125 km). Two study periods are considered: 5 March–3 April 2001 and 13 July–29 August 2002. Ten- day forecasts were run from each 1200 UTC analysis.
Two experiments were run for each period. In the control (CTL) experiment, the operational set of observations was assimilated, comprising a variety of conventional and satellite data. This included radiosonde, pilot balloon, profiler, and aircraft data, observations from synoptic stations, ships and buoys, scatterometer wind estimates, AMVs from geostationary satellites, Special Sensor Microwave Imager (SSM/I) data, and selected radiances (microwave and infrared) from two National Oceanic and Atmospheric Administration (NOAA) satellites as well as clear-sky water vapor radiances from Meteosat-7. In the MODIS experiment, Terra–MODIS winds were assimilated in addition to the operational set of observations. Over land we used MODIS IR and WV winds above 400 hPa only. Over sea, we used IR winds above 700 hPa and WV winds above 550 hPa. These restrictions for lower-level winds were as in KEY, and they were chosen after earlier trial experiments indicated poorer quality of the lower-level winds, most likely a result of height assignment problems over high orography and ice. All other settings for the MODIS winds (including observation errors) were as for operational AMVs from geostationary satellites (Rohn et al. 2001; Bormann et al. 2003): the winds were thinned to 140-km resolution, and quality control was based on an asymmetric check against the first guess (FG; Järvinen and Undén 1997). While all winds of the offline test dataset were considered in the assimilation for the March–April 2001 period, ECMWF's operational cutoff times applied during the July–August 2002 period for which the real-time Terra–MODIS winds were used. This resulted in a loss of available data as described above.
3. 4DVAR results
We now discuss the assimilation results for the 4DVAR experiments for the two periods 5 March–3 April 2001 and 13 July–29 August 2003. We first discuss the impact of the MODIS winds on the analysis, before illustrating the forecast impact.
a. Analysis impact
The assimilation of the MODIS winds has a notable impact on the mean polar wind analysis. For the March– April 2001 study period the differences for the Arctic are largest over the Arctic Ocean, with differences up to 2.5–3 m s−1 at most levels. Here, the MODIS winds act to strengthen the cyclonic circulation (e.g., Fig. 2). The changes to the mean wind analysis appear to be supported by the rest of the observing network: the u- component bias between the FG and Canadian pilot reports is marginally improved in the MODIS experiment (not shown). The differences in the mean polar wind analysis also agree well with the ones reported for the 3DVAR experiments over the same study period (Bormann et al. 2002), with the peak differences shifted slightly eastward. Over the Antarctic region, where fewer winds are used because of the blacklisting over land, differences in the mean wind analysis are somewhat weaker. For the July–August 2002 experiment, differences in the mean wind analyses are much smaller (less than 1.5 m s−1 vector difference at all levels; not shown). The reason for smaller differences may be seasonal variations in the bias pattern between the MODIS winds and the model data. Also, changes in the ECMWF model or the data usage between the two periods may have reduced the discrepancies. Note that the above discussion applies to differences in the mean wind analysis only; instantaneous differences can of course be much larger. In addition to the changes in the mean wind analysis, the variability of the analyses of the upper- level geopotential tends to be slightly decreased in the MODIS experiment.
Globally, the fit of other observations against the FG or the analysis is not significantly altered when MODIS winds are assimilated (not shown). This indicates a good overall consistency between the assimilation of MODIS winds and the assimilation of other observations. Locally, we could identify some changes, most notably for the U.S. profiler network for which the analysis fit is considerably improved, and more observations are used in the MODIS experiment (Fig. 3). This suggests a better model wind field in this region. As MODIS winds do not cover this area, the improvement is likely due to downstream developments, possibly as a result of a better positioning of the jet stream in the MODIS winds analyses thanks to the nature of the 4DVAR system. A marginal degradation can be reported locally for some conventional wind observations in the Arctic region covered by the MODIS winds (north of 60°N). Here, the standard deviation of the FG departures for radiosonde or pilot observations is very slightly increased around 300 hPa, whereas the fit to aircraft wind observations is mainly unaltered (not shown). The marginal degradation may be due to altered variability in the FG as a result of the MODIS winds assimilation, and it may well be compensated by better FG quality over the Arctic Ocean where no other wind observations are available. Nevertheless, the differences suggest some small local discrepancy in the use of the MODIS winds compared to other observations, possibly as a result of misspecified error or bias characteristics. For instance, MODIS winds may be prone to similar spatial error correlations as AMVs from geostationary satellites [nonzero error correlation on scales of about 800 km and thus larger than the thinning scales used (Bormann et al. 2003)], and these error correlations are not accounted for in our assimilation.
The assimilation of MODIS winds significantly increases the upper-level analysis increments in the polar areas covered by the MODIS winds (e.g., Fig. 4), with similar increases over the northern and the southern polar region. Increments are the adjustments made to the FG in the assimilation to produce an analysis that incorporates new observational information. The increased increments in the polar region are not surprising, as they occur in an area with few other observations (e.g., Fig. 1). Without the MODIS winds, the wind analyses in this region will be largely determined by the forecast model, whereas with MODIS winds previously undetected FG errors can be corrected. Supporting this, the increased increments over the Arctic Ocean occur in a region where the increments in the CTL experiment tend to be much smaller than over the well-observed surrounding areas (Fig. 4). In contrast, in the MODIS experiment the analysis increments have a much more consistent size over the entire higher midlatitudes. We therefore consider the increased increments over the poorly observed polar region a positive sign of the MODIS assimilation. It should be noted, however, that in well-observed regions smaller or unaltered increments are usually preferred, as they indicate smaller FG errors and a good consistency of the assimilation. Indeed, such consistency can also be reported for our experiments over better-observed regions such as northern Europe, the North Atlantic, or North America where the analysis increments are overall not similarly increased.
b. Forecast impact
There is a significant positive impact on forecasts of the geopotential height when MODIS winds are assimilated, particularly over Europe and also over the entire Northern Hemisphere extratropics. Figure 5 shows the improvement in forecasts of the 500-hPa geopotential height when the MODIS winds are assimilated. The figure shows the correlation between the geopotential height anomalies of the forecasts and the verifying analyses, with the forecasts each validated against their own verifying analyses (resulting in a total of 58 cases). The forecast improvements over Europe and the North Atlantic are significant at the 90% confidence level or better (t test) for a forecast range beyond 5 days. Over other parts of the Northern Hemisphere the impact on the geopotential forecast is more neutral, with no statistically significant changes in the anomaly correlations except for an improvement over the Arctic region and a slight degradation at lower levels over the North Pacific region. Wind forecast errors over the Northern Hemisphere are also reduced, including over the Arctic region (Fig. 6).
For the Southern Hemisphere, wind forecast errors over Antarctica are also improved considerably (Fig. 6), whereas the impact over the entire Southern Hemisphere extratropics is neutral overall, with a marginal improvement for the forecasts of the geopotential in the day 5 to 6 range (Fig. 5). The interpretation of forecast scores against analyses over the Southern Hemisphere is more difficult, as fewer available observations lead to smoother verifying analyses, whereas the addition of MODIS winds alters the variability of these analyses. However, the more neutral impact is also confirmed by a verification against observations that avoids the above problems, although it is meaningful only over limited areas. The more neutral impact is in agreement with findings in KEY, and there are a number of possible reasons for the somewhat weaker impact of the MODIS winds over the Southern Hemisphere. Fewer MODIS winds are used because of the strict blacklisting of lower-level winds over land. Also, height assignment for the MODIS winds is more difficult over the high and steep orography of the Antarctic Continent, possibly limiting the quality of the MODIS winds (e.g., KEY). Moreover, the more zonal flow over the Southern Hemisphere may limit the interaction between polar regions and the midlatitudes.
It is worth mentioning that the impact of the MODIS winds appears to vary over time, suggesting that the impact depends on the synoptic situation. Our experiments indicate the strongest and most significant positive impact over the Northern Hemisphere for the first half of the July–August 2002 experiment, while for other times the impact is more neutral. It appears that during July and the beginning of August 2002 the polar analysis played a more important role for the subsequent forecast. Daily examinations of ECMWF's numerical forecasts indeed tend to indicate greater sensitivity of the forecast for Europe to the polar regions in May–July compared to other times of the year (F. Grazzini 2003, personal communication). Typically, this greater sensitivity occurs during periods with northerly or northwesterly flow over the North Atlantic, so that analysis errors over northern Canada or Greenland propagate toward Europe during the forecast.
c. Forecast example
We will now discuss a sample forecast initialized on 1200 UTC 4 August 2002 to illustrate how the assimilation of MODIS winds can improve subsequent forecasts. The coverage of the MODIS winds for the corresponding 12-h assimilation cycle is illustrated in Fig. 7; for comparison, the coverage of conventional wind observations over the same period is indicated in Fig. 1. Figure 8 shows the initial analysis for this forecast together with the analysis differences between the MODIS and CTL experiments. Differences between the two analyses are noticeable north of the Bering Strait and Siberia, Russia, regions well covered by MODIS winds over this assimilation cycle. Other differences are also present, for instance, southeast of Greenland and near Newfoundland, Canada. Peak differences are of the order of 1–3 gpm for the 500-hPa geopotential. These differences propagate downstream in the subsequent forecast, grow, and interact, leading to large forecast differences over North America and the northern Atlantic already at day 4 (Fig. 9). A comparison with the verifying analysis reveals, for instance, a better location and orientation of the trough north of Alaska and a much improved forecast of the ridge over the North Atlantic.
To investigate which regions in the analysis played a particularly important role for the subsequent forecast we estimated so-called key analysis errors highlighting the sensitivity to initial conditions for the sample forecast (e.g., Klinker et al. 1998). These sensitivity calculations estimate analysis perturbations required to minimize the subsequent 2-day Northern Hemisphere (north of 20°N) forecast error under the total energy norm, using the adjoint method. In other words, these calculations estimate “key analysis errors” by using the analysis 2 days after the initial time and mapping the forecast errors against this verifying analysis back to the initial analysis. The model used to evaluate the error propagation is based on the linear approximation and has a somewhat lower resolution (T63, approximately 320 km). Such sensitivity calculations provide a powerful tool to identify particularly sensitive regions in the initial analysis.
The calculations indicate considerable sensitivity of the 2-day forecast to the polar regions. Figure 10 shows the absolute value of the sensitivity perturbations for the streamfunction on model level 39 (approximately 500 hPa) for the CTL experiment, with maxima in the polar region around the Bering Strait, north of Siberia, Russia, over Scandinavia, and over Québec, Canada. For the MODIS experiment, the sensitivity perturbations over most of these very sensitive areas are noticeably reduced by about 0.1–0.2 m2 s−1 (10%–30%), except north of Québec and over the northern Atlantic (Fig. 10). The reduction of the sensitivity perturbations suggests a reduction of analysis errors in these areas, introduced either through the MODIS winds assimilated in this assimilation cycle or through earlier improvements already present in the FG. The improvements north of Siberia and northwest of the Bering Strait coincide with areas well covered by the MODIS winds used in the 1200 UTC analysis on 4 August 2002 (Fig. 8), whereas no MODIS winds were present in the region around Québec where a slight degradation is present. While it is difficult to directly relate the changes in the sensitivity pattern to the MODIS winds for only one cycle, the findings highlight the ability of MODIS winds to reduce key analysis errors over the polar regions.
Forecast improvements linked with a reduction of key analysis errors in the polar regions have been observed in a number of other forecast cases in the MODIS experiment. While the reduction is not necessarily systematic, the assimilation of MODIS winds tends to reduce extreme values in the sensitivity perturbations, as highlighted in the above example.
4. Discussion and conclusions
This paper reported on the first impact experiments with polar winds from MODIS within a 4DVAR framework. The main findings are the following.
The MODIS polar winds have a positive forecast impact, particularly over the polar regions and Europe, but also over the Northern Hemisphere as a whole. The Southern Hemisphere forecast impact is neutral overall, partly because of a more restricted use of the MODIS winds over land.
The assimilation of the MODIS winds can considerably alter the mean polar wind analysis for some periods, suggesting that the MODIS winds can correct systematic deficiencies in model analyses.
A forecast example highlights how Northern Hemisphere forecasts are sensitive to analysis perturbations over the polar regions and how the assimilation of MODIS winds can reduce key analysis errors in these areas, subsequently leading to improved forecasts.
The above findings confirm earlier results of positive forecast impact with the MODIS winds in a coarser- resolution 3DVAR configuration over a shorter time period (KEY; Bormann et al. 2002). Within the current observing network, the MODIS winds are capable of adding wind information in an otherwise poorly observed and sensitive region also in a 4DVAR framework. As a result of these findings, the Terra–MODIS winds have been assimilated operationally at ECMWF since 14 January 2003.
The study highlights considerable sensitivity of mid- and high-latitude forecasts to polar regions. Further investigations are required to characterize the typical regions or weather systems with particular sensitivity, or to relate the sensitivity to specific observations, for instance using methods of Hello et al. (2000). This would also help the targeting of observations.
The positive forecast impact demonstrated with MODIS winds in 3DVAR and 4DVAR experiments is particularly encouraging, as it has been obtained by merely adopting the assimilation approach from geostationary AMVs with few specific adjustments for the MODIS winds. As a consequence there is scope for refinements in the assimilation of the MODIS winds. Quality control decisions, for instance, have been fairly cautious, with no assimilation of low-level winds and highly restricted use of winds over land. Revised quality control may allow a better use of the MODIS winds, for instance over the Antarctic, which could aid the forecast impact over the Southern Hemisphere. In addition, quality indicators for the derived winds are provided with the data (e.g., Holmlund et al. 2001), and this information has not been used in our experiments. Use of such information in the quality control or the thinning step of the assimilation can be beneficial (e.g., Rohn et al. 2001). More generally, fine-tuning of the observation error characteristics remains to be done, since at present values corresponding to geostationary AMVs have been used. This fine-tuning includes a revision of the thinning scales to suppress spatially correlated errors in light of recent findings on spatial error correlations for winds derived from geostationary satellite data (Bormann et al. 2003). Also, bias characteristics for the MODIS winds should receive further attention to assure that they are treated appropriately in the assimilation (e.g., Bormann et al. 2001). Progress in these areas is expected to lead to improvements over our present “day 1” system.
In addition, further improvements are also possible on the data side. For instance, this study investigated only the impact of the Terra–MODIS winds. In the meantime, MODIS winds have also become available from the Aqua satellite, providing even better coverage of the polar wind field. On top of this, combined use of Terra and Aqua data in the winds processing allows the shortening of the time interval between swaths (currently 100 min) and could therefore lead to improved tracking. It has also been suggested that the use of relatively coarse-resolution forecast data in the winds processing negatively affects the height assignment for lower-level MODIS winds, and this aspect will be investigated further by using higher-resolution ECMWF fields. A further challenge would be to provide and assimilate a clear-sky radiance product from MODIS data, similar to that used from some geostationary satellites (e.g., Köpken et al. 2002; Munro et al. 2000).
Acknowledgments
This study was done under the EUMETSAT/ECMWF Fellowship Programme. The MODIS winds have been provided by David Santek, Jeffrey Key, and Christopher Velden at CIMSS/NOAA/ NESDIS, and Paul Menzel (NOAA/NESDIS) has been very supportive in initiating this project. The ECMWF system is the product of many staff members and consultants at ECMWF and Météo France, and we are especially grateful to Ioannis Mallas, Milan Dragosavac, Federico Grazzini, and Ernst Klinker for their help during this study. Graeme Kelly provided many comments and enthusiasm for the project. Rob Hine is gratefully acknowledged for his help with the preparation of the figures. Two anonymous reviewers provided valuable comments and suggestions.
REFERENCES
Bormann, N., G. Kelly, and J-N. Thépaut, 2001: Characterising and correcting speed biases in atmospheric motion vectors within the ECMWF system. Proc. 2001 EUMETSAT Meteorological Satellite Data User's Conf., Antalya, Turkey, EUMETSAT, 596– 603.
Bormann, N., J-N. Thépaut, J. R. Key, D. Santek, and C. S. Velden, 2002: Impact of polar cloud track winds from MODIS on ECMWF analyses and forecasts. Preprints, 15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., 15– 18.
Bormann, N., S. Saarinen, G. Kelly, and J-N. Thépaut, 2003: The spatial structure of observation errors in atmospheric motion vectors from geostationary satellite data. Mon. Wea. Rev, 131 , 706–718.
Francis, J. A., 2002: Validation of reanalysis upper-level winds in the Arctic with independent rawinsonde data. Geophys. Res. Lett.,29, 1315, doi:10.1029/2001GL014578.
Hello, G., F. Lalaurette, and J-N. Thépaut, 2000: Combined use of sensitivity information and observations to improve meteorological forecasts: A feasibility study applied to the “Christmas storm” case. Quart. J. Roy. Meteor. Soc, 126 , 621–647.
Holmlund, K., C. Velden, and M. Rohn, 2001: Enhanced automated quality control applied to high-density satellite-derived winds. Mon. Wea. Rev, 129 , 517–529.
Järvinen, H., and P. Undén, 1997: Observation screening and background quality control in the ECMWF 3DVAR data assimilation system. Tech. Memo. 236, ECMWF, Reading, United Kingdom, 33 pp.
Key, J. R., D. Santek, C. S. Velden, N. Bormann, J-N. Thépaut, L. P. Riishøjgaard, Y. Zhu, and W. P. Menzel, 2003: Cloud-drift and water vapor winds in the polar regions from MODIS. IEEE Trans. Geosci. Remote Sens, 41 , 482–492.
Klinker, E., F. Rabier, and R. Gelaro, 1998: Estimation of key analysis errors using the adjoint technique. Quart. J. Roy. Meteor. Soc, 124 , 1909–1933.
Klinker, E., F. Rabier, G. Kelly, and J-F. Mahfouf, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part III: Experimental results and diagnostics with operational configuration. Quart. J. Roy. Meteor. Soc, 126 , 1191–1215.
Köpken, C., G. Kelly, and J-N. Thépaut, 2002: Monitoring and assimilation of Meteosat radiances within the 4DVAR system at ECMWF. ECMWF/EUMETSAT Fellowship Rep. 9, ECMWF, Reading, United Kingdom, 31 pp.
Munro, R., G. Kelly, and R. Saunders, 2000: Assimilation of Meteosat radiance data within the 4DVAR system at ECMWF. ECMWF/ EUMETSAT Fellowship Rep. 8, ECMWF, Reading, United Kingdom, 41 pp.
Nieman, S. J., W. P. Menzel, C. M. Hayden, D. Gray, S. T. Wanzong, C. S. Velden, and J. Daniels, 1997: Fully automated cloud-drift winds in NESDIS operations. Bull. Amer. Meteor. Soc, 78 , 1121–1133.
Rabier, F., H. Järvinen, E. Klinker, J-F. Mahfouf, and A. Simmons, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part I: Experimental results with simplified physics. Quart. J. Roy. Meteor. Soc, 126 , 1143–1170.
Rohn, M., G. Kelly, and R. W. Saunders, 2001: Impact of a new cloud motion wind product from Meteosat on NWP analyses. Mon. Wea. Rev, 129 , 2392–2403.
Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gaertner, A. Koch, and L. Van De Berg, 1993: Operational cloud- motion winds from Meteosat infrared images. J. Appl. Meteor, 32 , 1206–1225.
Velden, C. S., C. M. Hayden, S. J. Nieman, W. P. Menzel, S. Wanzong, and J. S. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc, 78 , 173–195.
Example of the locations of conventional polar wind observations used in the operational ECMWF assimilation for the 12-h period 0300–1500 UTC 4 Aug 2002. See the legend for an explanation of the symbols
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
(top) Mean polar wind analyses at 400 hPa for the CTL experiment for 1200 UTC 5 Mar–3 Apr 2001, and (bottom) the difference between the mean wind analysis for the MODIS and CTL experiments for the same period. Shading indicates the length of the difference vector (m s−1)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Statistics for the first guess (solid) and the analysis departures (dashed) for used U.S. profiler data for the MODIS (black) and CTL (gray) experiments. (left) Standard deviations (m s−1) vs pressure and (right) biases (m s−1) vs pressure level, with statistics for the (top) u component and (bottom) υ component. The number of winds used is also shown between the columns, with the difference MODIS−CTL given in gray. Statistics for the two study periods have been pooled together
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Root-mean-square of the analysis increments of the 500-hPa geopotential height (gpdm) for the (left) CTL and (right) MODIS experiments for the Jul–Aug 2002 period
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Anomaly correlations for the 500-hPa geopotential height forecast against the verifying analysis as a function of forecast range. The forecasts for the MODIS (solid black) and CTL (dashed gray) experiments have each been verified against their own analyses, and scores for both periods have been pooled together (58 forecasts). The four plots show anomaly correlations over the Northern Hemisphere extratropics, the Southern Hemisphere extratropics, the North Atlantic, and Europe, respectively. Over the Northern Hemisphere, the North Atlantic, and Europe the day 8 differences between the MODIS and CTL experiments are statistically significant at the 90% level or better (t test)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Root-mean-square errors for the 300- and 500-hPa wind forecast for the MODIS (solid black) and CTL (gray dashed) experiments, with each experiment verified against their own analyses. Scores for the two study periods have been pooled together (58 forecasts). The panels show values for the Arctic (north of 65°N) and Antarctic region (south of 65°S), respectively. At 500 hPa and over the Antarctic region, the differences between the MODIS and CTL experiments are statistically significant at the 95% level or better (t test) in the day 4–6 forecast range, whereas for the other regions shown, 90% significance or better is reached later in the forecast range
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Coverage of the used MODIS winds for the 0300–1500 UTC assimilation cycle on 4 Aug 2002, stratified by pressure level as indicated. The winds have been slightly thinned for display purposes
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
(left) Analysis of the 500-hPa geopotential height (gpm) for the MODIS experiment for 1200 UTC 4 Aug 2002. (right) Differences MODIS-CTL between the 500-hPa geopotential height analyses for the same day. Solid contours indicate positive differences, and dashed contours indicate negative differences (0.4-gpm contour interval)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
Four-day forecasts of the 500-hPa geopotential height for the (a) MODIS and (b) CTL experiments valid on 1200 UTC 8 Aug 2002. (c) Verifying analysis for 1200 UTC 8 Aug 2002 from the CTL experiment. (d) Differences between the MODIS and CTL forecasts shown in (a) and (b). Solid contours indicate positive differences, and dashed contours indicate negative differences (2.0-gpm contour interval)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
(left) Shading shows absolute values of the sensitivity perturbations for the streamfunction (m2 s−1) at model level 39 (approximately 500 hPa) for the analysis valid 1200 UTC 4 Aug 2002. Contours show the 500-hPa geopotential (gpm). (right) Shading shows the differences between the MODIS and CTL experiments in terms of the absolute value of the sensitivity perturbations for the streamfunction shown on the left. Contours indicate the 500-hPa geopotential (gpdm)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0929:IOMPWI>2.0.CO;2
This situation has been improved since our experiments have been performed. Now, about 80% of the MODIS winds arrive within the operational data cutoff times at ECMWF.