A Comparative Study of Maximum and Minimum Temperatures over Argentina: NCEP–NCAR Reanalysis versus Station Data

Matilde M. Rusticucci Departamento de Ciencias de la Atmósfera y los Océanos, Universidad de Buenos Aires, Buenos Aires, Argentina

Search for other papers by Matilde M. Rusticucci in
Current site
Google Scholar
PubMed
Close
and
Vernon E. Kousky NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, Maryland

Search for other papers by Vernon E. Kousky in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

This paper compares surface-station temperature observations over Argentina with gridpoint analyses available in the NCEP–NCAR reanalysis dataset. The primary objective is to determine whether the maximum and minimum surface temperatures from the reanalysis can be used to compute statistics on the occurrence of extreme events. The extreme range of topography and geography of Argentina is viewed as a severe test for the reanalysis data. Good agreement, on both the daily and monthly timescales, between the station data and the reanalysis gridpoint data is found over the low-elevation regions in central and eastern Argentina. The agreement is relatively poor for summertime maximum temperatures over northern Argentina. The reanalysis data underestimate the intensity of extreme warm events over northern and southern Argentina and overestimate extreme cold events during winter over central Argentina. High-elevation areas in western Argentina have the poorest correspondence throughout the year. Thus, the NCEP–NCAR reanalysis data have to be used with caution for studies of the magnitude of day-to-day temperature changes. The results of this study indicate that the NCEP–NCAR reanalysis data are sufficient for determining the timing of midlatitude events but are not sufficient for determining the amplitude and frequency in the subtropics and in regions of high relief. The use of anomalies tends to improve the amount of agreement between the reanalysis data and station observations.

Corresponding author address: Matilde Rusticucci, Departamento de Ciencias de la Atmósfera y los Océanos, FCEN, UBA, Ciudad Universitaria Pab II, 1428 Buenos Aires, Argentina. Email: mati@at.fcen.uba.ar

Abstract

This paper compares surface-station temperature observations over Argentina with gridpoint analyses available in the NCEP–NCAR reanalysis dataset. The primary objective is to determine whether the maximum and minimum surface temperatures from the reanalysis can be used to compute statistics on the occurrence of extreme events. The extreme range of topography and geography of Argentina is viewed as a severe test for the reanalysis data. Good agreement, on both the daily and monthly timescales, between the station data and the reanalysis gridpoint data is found over the low-elevation regions in central and eastern Argentina. The agreement is relatively poor for summertime maximum temperatures over northern Argentina. The reanalysis data underestimate the intensity of extreme warm events over northern and southern Argentina and overestimate extreme cold events during winter over central Argentina. High-elevation areas in western Argentina have the poorest correspondence throughout the year. Thus, the NCEP–NCAR reanalysis data have to be used with caution for studies of the magnitude of day-to-day temperature changes. The results of this study indicate that the NCEP–NCAR reanalysis data are sufficient for determining the timing of midlatitude events but are not sufficient for determining the amplitude and frequency in the subtropics and in regions of high relief. The use of anomalies tends to improve the amount of agreement between the reanalysis data and station observations.

Corresponding author address: Matilde Rusticucci, Departamento de Ciencias de la Atmósfera y los Océanos, FCEN, UBA, Ciudad Universitaria Pab II, 1428 Buenos Aires, Argentina. Email: mati@at.fcen.uba.ar

1. Introduction

Tremendous progress has been made in developing consistent long-term gridded datasets for use in climate studies. The efforts to reanalyze historical data using modern data assimilation systems (Schubert et al. 1993, 1995; Kalnay et al. 1996; Kistler et al. 2001) have played an important part in this progress. By using a fixed data assimilation system, jumps in the historical record that resulted from model improvements, such as increases in model resolution and changes in physical parameterizations, have been eliminated. However, jumps in the historical record remain because of nonhomogeneous observational databases and imperfect models into which the data are assimilated. In addition, certain reanalysis variables depend greatly on the physical parameterizations in the model and the procedures used to compute desired quantities. Therefore, it is necessary to validate the reanalysis, whenever possible, using independent observations.

In this paper, we focus on the near-surface values of maximum and minimum temperature in the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996). These are model-derived quantities that are computed from 6-hourly integrations of the model. The NCEP–NCAR reanalysis data are available on a 2.5° latitude–longitude grid. We address the issues of 1) how well the 2-m temperatures in the reanalysis data archive compare to station observations and 2) whether these reanalysis variables can be used to determine a climatological description of extreme temperature events.

2. Methodology

Our validation dataset consists of maximum temperatures (max T) and minimum temperatures (min T) for selected stations in Argentina for the 40-yr period of 1959–98. The location and altitude of the selected stations are shown in Fig. 1 and are listed in Table 1. The stations used were selected outside the main cities to avoid the possible urban heat island effect. This effect produces mean differences of about 3°C between temperatures within and outside Buenos Aires (Rusticucci and Vargas 1995).

Some potential problem areas that we will address in subsequent sections include 1) the use of gridded analyses in comparison with point observations at selected stations and 2) the effect of model versus real topography in producing temperature biases.

We first investigate the accuracy of the reanalysis data by comparing time series of the observed maximum and minimum temperatures at the stations with the corresponding time series at the reanalysis grid point nearest to the station locations. The daily, seasonal, and decadal variability in the differences are investigated for each station.

We also investigate the accuracy of the maximum (minimum) temperature anomalies in the reanalysis data. Anomalies for both the station data and the reanalysis data are computed for the periods of 1959–78 and 1979–98 by removing the respective 20-yr mean daily maximum (minimum) temperature from each of the datasets. We chose to break the entire 40-yr record into two 20-yr periods, one for the presatellite period of 1959–78 and the other for the satellite period of 1979–98. This allows us to assess the impacts of the change in the observational database on our results.

For the analysis of extremes, threshold values were selected to define the extreme events and the duration of warm and cold spells that occurred during the 40-yr period. All days during the year were used to calculate the percentiles. Maximum (minimum) temperature anomalies that ranked in the upper (lower) 25% of the distribution were used as thresholds for being included as warm (cold) spells. The spells start with a 1-day-long duration, but we used the longest duration. Extreme warm (cold) spells were defined as those whose length, based on spell duration, ranked in the upper 10% of the distribution. The ranking and determination of extreme events were done for each month separately. For each month, there were approximately 10 extreme spells over the 40-yr period. We applied the same criteria for determining warm and cold spells, and extreme events, to the reanalysis data. We then compared the results for extreme events by counting the number of the events that had at least one day in common between the two datasets.

In addition, the agreement between the two datasets for extreme-event cases was further analyzed by comparing the 24-h temperature changes in both datasets for every day of the year. Max T was used to calculate increments, and min T was used to calculate drops in temperature. It is assumed that the largest temperature changes are due to frontal passages, which is most likely the case for large 24-h temperatures decreases (cold fronts).

Differences and correlation coefficients have been tested using a Student's t test at a 95% confidence level.

3. Results

Comparisons of the reanalysis maximum and minimum temperature with two selected stations are shown in Figs. 2a,b. Each station represents a region with the same characteristics and biases. These years were selected for a better representation of the differences. The day-to-day variability associated with weather systems appears to be represented well in the reanalysis data, with good agreement in the timing of the maxima and minima in the two curves. At Pergamino (Fig. 2a), there is also good agreement in the magnitude of the day-to-day changes and in the actual values of the extremes for both the maximum and minimum temperatures throughout the annual cycle. In contrast, the time series for Iguazú (Fig. 2b) shows that there is good correspondence between the reanalysis and station data only for the minimum temperatures. The maximum temperatures at this station show substantial offsets between the two curves, especially during the summer, autumn, and early winter months, with the reanalysis data being cooler than the station data. However, there is some agreement in the timing of day-to-day weather events between the two datasets, and the large difference between summer and winter variability is well represented. The differences (reanalysis minus station) between simultaneous temperatures show large values, even as monthly averages. Reanalysis data are relatively cold over and near the Andes (Salta in the north, Comodoro Rivadavia in the south), warm in central Argentina (Pergamino), and cold for max T over northern Argentina (Iguazú; see these examples in Fig. 3).

Some extreme seasonal cases were selected to show the differences and their statistical significance. For the winter season, 1988 (extremely cold) and 1990 (extremely warm) are compared. For summer, a comparison is made between 1975 (cold) and 1972 (warm). The differences, reanalysis minus station data, and their 95% significance level are drawn in Fig. 4. It is evident that there are no appreciable differences in the reanalysis behavior for different extreme years. In winter (Fig. 4a), reanalysis data are significantly warmer over the east and in the south, especially in min T. For max T, the reanalysis data are closer to station data, and they correctly represent the warmest winter with very few significant differences. Summers (Fig. 4b) show different patterns depending on which variable is examined. Min T shows the same behavior in winter as in summer. Max T is significantly colder almost everywhere, except for central Argentina.

To examine the long-term daily correspondence between the reanalysis and station data, we first calculated the daily correlations between the two datasets for the entire 40-yr record, for summer [December–January–February (DJF)] and winter [June–July–August (JJA)]. These maps (Fig. 5) show that the best correspondence (locally) of daily temperature variability between the two datasets is during winter for minimum temperatures over the eastern part of Argentina. Overall, winter maximum temperature is just as good as, if not better than, minimum temperature with the exception of in the northeast. The correlations generally decrease toward the west (Andes) at all latitudes, although all coefficients are significant. Therefore, the reanalysis data represent the daily variability very well, especially over eastern Argentina (La Plata River basin). Summer max T, which has the greatest interannual variability in its extremes (Rusticucci and Barrucand 2001), is not represented as well.

The correlation patterns for monthly averaged temperatures between the two datasets (Figs. 6 and 7) for the entire 40-yr record show similar patterns as those for the daily data (Fig. 5). For maximum temperatures (Fig. 6), there is considerable seasonality in the correlations over the northern portion of the region. For this area, correlations are relatively low (less than 0.5, regions unshaded) during the late spring, summer, and early autumn months (October–April) and are uniformly high (greater than 0.7, medium gray shading) throughout most of the domain during May–September. For minimum temperatures (Fig. 7), the correlation pattern is similar throughout the year. However, there is a marked west-to-east gradient in the correlations, with the highest values (greater than 0.7) over the eastern sections and the lowest values (less than 0.5) in the vicinity of the Andes. In this case, the 95% significance limit for the correlation coefficients is 0.3, so the majority of stations have significant correlation.

These correlations between the two datasets do not show any significant decadal variability. Yearly correlation values were computed and were found to oscillate around the long-term mean values without any sign of a trend that would indicate an improvement or worsening of the correspondence between the reanalysis and station data (results not shown).

Seasonal mean temperature anomalies were analyzed over moving 10-yr periods, running from 1959–68 to 1989–98, to investigate their long-term variability. The station and reanalysis decadal mean anomalies at three points located near 60°W and spanning 15° of latitude are shown for summer (DJF) and winter (JJA) for maximum (Fig. 8) and minimum (Fig. 9) temperatures. The best long-term correspondence is for min T in winter, especially during the last decades at Pergamino and Las Lomitas. The min T time series for Bahia Blanca shows no trend and has good correspondence with the reanalysis data. In general, the correspondence for decadal maximum temperature anomalies is poorer than for minimum temperatures, with results being particularly poor in northern Argentina during the summer. In contrast to the very good correspondence observed at Bahia Blanca for minimum temperature, the correspondence for maximum temperatures is relatively poor throughout the series.

a. Extreme events

The criterion for the coincidence of extreme daily events that we used (at least one day in common between the two datasets) resulted in a coincidence of nearly 100% every month for a majority of stations in central Argentina. The coincidence percentage averaged over the country is from 78% (January) up to 92% (August) for warm events and from 90% up to 96% for cold events. The extreme cold events are better represented in the reanalysis data than are the extreme warm events, at least from the point of view of the days involved in the extreme event.

For each event, the average departure from normal was determined (which we refer to as intensity of the event). The mean intensities averaged over the coincident events in the station and reanalysis datasets were compared. In Fig. 10, the mean intensity differences (calculated as station minus reanalysis) between extreme warm events were plotted for summer (December–February) and winter (June–August), and significant differences are shaded. It is evident that the observed extreme warm events have higher mean intensity (are stronger) than reanalysis-based extreme events, and there is no marked seasonality. The cause for the weaker warm events may be the amount of clouds produced in the reanalysis data (the same reason why our results were not as good in the subtropics). Too much cloudiness would tend to keep temperatures down during the day.

The same analysis was done for cold events. As shown in Fig. 11, the reanalysis-based extreme cold events generally have higher mean intensity (i.e., they are colder) than station-based extreme events, especially in winter months. Always, the southern region is poorly represented in the reanalysis data, as is the northwest portion of the domain.

b. 24-h temperature changes

To assess the ability of the reanalysis to capture frontal passages, the 24-h tendencies in maximum and minimum temperature were calculated in both datasets, as Tday(i)Tday(i−1). An analysis of the agreement was performed, based on both the sign and the magnitude of the differences. In the analysis of sign agreement, all cases of negative and positive tendencies were identified in each dataset. An analysis was made of the percentage of the number of cases in agreement. For the analysis of the agreement in magnitude of the tendencies, cases were selected for which the 24-h temperature changes were either in the upper or lower 10% (largest 10% of positive and negative tendencies) of the total distribution. The threshold limit for negative changes was found to be generally near −5°C, with some variability depending on the time of year and on station location. Changes of this magnitude or greater are assumed to be due to cold-frontal passages. In a similar manner, the upper 10% limit was used to define possible warm-frontal passages. The percentage of cases for the extreme negative and positive 24-h temperature changes that is common to the two datasets was determined for each station. These percentages are presented for the following regions: (a) over the Andes, and (b) northeastern, (c) southern, and (d) central Argentina.

The mean monthly percentages for agreement of the sign of the temperature tendencies for each region (Fig. 12) show that over 60% of the cases are common to both datasets, with the highest values found over the east and northeast. The agreement for the extreme events (magnitude of the tendency in the upper or lower 10% of the distribution) in these regions is not high, with 40% or less correspondence between the two datasets. The poorest agreement is found over the northeastern stations for maximum-temperature tendencies, where some stations agree in less than 20% of the cases.

4. Discussion

There is good agreement between the station data and the reanalysis gridpoint data for low-elevation regions in central and eastern Argentina. The poorest correspondence is in the vicinity of the Andes and, in general, at all low latitudes during summer. The poor correspondence over the Andes is probably due to the differences between model topography and real topography. The poor results during summer (daily values in the reanalysis are colder than the station data) over northern Argentina imply that there may be differences between model-predicted cloudiness, possibly resulting from convection, and observed cloudiness. Because summertime cloudiness is an important factor in determining temperature, weaknesses in the physical parameterizations of cloudiness and convection could very well be the cause of the poor summertime agreement in that region. This is an important result, because it gives the user of the reanalysis data information on where and when it is appropriate to use the surface temperatures.

The analysis of extreme-duration events revealed that the reanalysis data underestimate the intensity of extreme warm events over northern and southern Argentina and overestimate winter extreme cold events over central Argentina. These results indicate a negative temperature bias in the reanalysis data for extreme events. Otherwise there are no differences in extreme-event intensity, with the exception of southernmost and higher-elevation locations.

The NCEP–NCAR reanalysis data correctly indicate the sign of the 24-h temperature changes in about 60% of the cases, with results being best over eastern and northeastern Argentina. However, when a comparison was made for cases having the largest negative and positive changes, there was less than 40% agreement and in some cases less than 20% agreement. Thus, the NCEP–NCAR reanalysis data have to be used with caution for studies of the magnitude of day-to-day temperature changes.

Our results indicate that the NCEP–NCAR reanalysis data are sufficient for determining the timing of midlatitude events but are not sufficient for determining the amplitude and frequency in the subtropics and in regions of high relief. The use of anomalies tends to improve the amount of agreement between the reanalysis data and station observations.

Further diagnostic studies are necessary to verify the true causes for the differences we found. The topography is too high in the reanalysis, which might lead to the conclusion that temperatures should be too cold. For minimum temperatures, however, radiational cooling is an important process, which usually results in an inversion in the lowest layers. Perhaps the reanalysis does not allow for enough radiational cooling, because of too much cloudiness or too much wind. These are speculations and need to be proven, but reanalysis can be used in this region with the knowledge of differences that has been established.

Acknowledgments

The visit of Matilde Rusticucci to NCEP was supported by UBA Grants JX29, TW06 and by the Departamento de Ciencias de la Atmósfera FOMEC.

REFERENCES

  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kistler, R., and Coauthors. 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82 , 247267.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and W. Vargas, 1995: Seasonal and diurnal patterns of dry- and wet-bulb temperatures over Argentina. Int. J. Climatol., 15 , 12731283.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and M. Barrucand, 2001: Variabilidad interanual de temperaturas extremas en la República Argentina [Interannual variability of extreme temperatures in the Republic of Argentina]. CD VIII Congreso Latinoamericano e Ibérico de Meteorología, Buenos Aires, Argentina, Centro Argentino e Meteorólogos, N°169, 8pp.

    • Search Google Scholar
    • Export Citation
  • Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc., 74 , 23312342.

    • Search Google Scholar
    • Export Citation
  • Schubert, S. D., and Coauthors. 1995: A multiyear assimilation with the GEOS-1 system: Overview and results. NASA Tech. Memo. 104606, 207 pp.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Location of stations used. Smoothed altitude is in meters

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 2.
Fig. 2.

Daily temperatures (upper panel: max T, lower panel: min T) for (a) Pergamino for 1969 (dashed line) and nearest grid point (solid line), and (b) Iguazú for 1993 (dashed line) and nearest grid point (solid line).

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 3.
Fig. 3.

Daily differences (reanalysis minus station data) averaged monthly, 1959–98, for max T (white bars) and min T (black bars) in four locations

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 4.
Fig. 4.

Daily differences (reanalysis minus station data) averaged over the indicated season for min T and max T. Significant 95% differences evaluated by a Student's t test are marked with crosses. Contour interval is 2°C, shaded areas are negative values (reanalysis too cold). (a) Winter 1988 (1990) was the coldest (warmest) in record, and (b) summer 1975 (1972) was the coldest (warmest) in record

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 5.
Fig. 5.

Correlations between station and reanalysis daily values for the whole period 1959–98 for (top) winter (JJA) and (bottom) summer (DJF) for (right) maximum and (left) minimum temperatures. Correlations are significant at 95% level over 0.3

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 6.
Fig. 6.

Monthly maximum temperature correlations between reanalysis and station data over period 1959–98. Contour interval is 0.1. Shades: 0.5–0.7, light gray, 0.7–0.9, medium gray; over 0.9, dark gray

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 7.
Fig. 7.

Same as Fig. 6, but for monthly minimum temperature

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 8.
Fig. 8.

Maximum temperature anomalies, moving 10-yr averages starting in 1959–68, for (left) summer and (right) winter for station data (bold solid line) and reanalysis (dashed line)

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 9.
Fig. 9.

Same as Fig. 8, but for minimum temperature anomalies

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 10.
Fig. 10.

Mean intensity differences for extreme warm events. Positive values indicate station events stronger than reanalysis events for summer (DJF) and winter (JJA). Contour interval is 1°C. Shaded areas are significant at 95% level

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 11.
Fig. 11.

Same as Fig. 10, but for extreme cold events. Dashed contours are negative differences, reanalysis extreme events are stronger than station extreme events. Contour interval is 1°C. Shaded areas are significant at 95% level

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Fig. 12.
Fig. 12.

Monthly percentages of coincidence in sign (solid lines) and in the top/bottom decile (dashed lines) for minimum (open circles) and maximum 24-h temperature change, for each of the four regions shown on the map below the graphs

Citation: Journal of Climate 15, 15; 10.1175/1520-0442(2002)015<2089:ACSOMA>2.0.CO;2

Table 1.

List of stations, station abbreviations (shown in Fig. 1), their locations, and height (m) above sea level (MSL), plus the coordinates of the nearest Gaussian grid point in the NCEP–NCAR reanalysis archive

Table 1.
Save
  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kistler, R., and Coauthors. 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82 , 247267.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and W. Vargas, 1995: Seasonal and diurnal patterns of dry- and wet-bulb temperatures over Argentina. Int. J. Climatol., 15 , 12731283.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and M. Barrucand, 2001: Variabilidad interanual de temperaturas extremas en la República Argentina [Interannual variability of extreme temperatures in the Republic of Argentina]. CD VIII Congreso Latinoamericano e Ibérico de Meteorología, Buenos Aires, Argentina, Centro Argentino e Meteorólogos, N°169, 8pp.

    • Search Google Scholar
    • Export Citation
  • Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc., 74 , 23312342.

    • Search Google Scholar
    • Export Citation
  • Schubert, S. D., and Coauthors. 1995: A multiyear assimilation with the GEOS-1 system: Overview and results. NASA Tech. Memo. 104606, 207 pp.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Location of stations used. Smoothed altitude is in meters

  • Fig. 2.

    Daily temperatures (upper panel: max T, lower panel: min T) for (a) Pergamino for 1969 (dashed line) and nearest grid point (solid line), and (b) Iguazú for 1993 (dashed line) and nearest grid point (solid line).

  • Fig. 3.

    Daily differences (reanalysis minus station data) averaged monthly, 1959–98, for max T (white bars) and min T (black bars) in four locations

  • Fig. 4.

    Daily differences (reanalysis minus station data) averaged over the indicated season for min T and max T. Significant 95% differences evaluated by a Student's t test are marked with crosses. Contour interval is 2°C, shaded areas are negative values (reanalysis too cold). (a) Winter 1988 (1990) was the coldest (warmest) in record, and (b) summer 1975 (1972) was the coldest (warmest) in record

  • Fig. 5.

    Correlations between station and reanalysis daily values for the whole period 1959–98 for (top) winter (JJA) and (bottom) summer (DJF) for (right) maximum and (left) minimum temperatures. Correlations are significant at 95% level over 0.3

  • Fig. 6.

    Monthly maximum temperature correlations between reanalysis and station data over period 1959–98. Contour interval is 0.1. Shades: 0.5–0.7, light gray, 0.7–0.9, medium gray; over 0.9, dark gray

  • Fig. 7.

    Same as Fig. 6, but for monthly minimum temperature

  • Fig. 8.

    Maximum temperature anomalies, moving 10-yr averages starting in 1959–68, for (left) summer and (right) winter for station data (bold solid line) and reanalysis (dashed line)

  • Fig. 9.

    Same as Fig. 8, but for minimum temperature anomalies

  • Fig. 10.

    Mean intensity differences for extreme warm events. Positive values indicate station events stronger than reanalysis events for summer (DJF) and winter (JJA). Contour interval is 1°C. Shaded areas are significant at 95% level

  • Fig. 11.

    Same as Fig. 10, but for extreme cold events. Dashed contours are negative differences, reanalysis extreme events are stronger than station extreme events. Contour interval is 1°C. Shaded areas are significant at 95% level

  • Fig. 12.

    Monthly percentages of coincidence in sign (solid lines) and in the top/bottom decile (dashed lines) for minimum (open circles) and maximum 24-h temperature change, for each of the four regions shown on the map below the graphs

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 429 99 3
PDF Downloads 203 42 5