(c) Slope coefficients of simple lnI-Td, lnI-lnCAPE and lnCAPE-Td regressions (first row), and for multiple regression (Eq. 2), and gamma (second row; see text for details).

lnCAPE - T

75 8590 98 99 99.90.020.040.060.080.10.120.14

%75 8590 98 99 99.9

0.1

0.2

0.3

0.4

0.5

%75 8590 98 99 99.9

0.020.040.060.080.10.120.14

%758590 989999.999.99

ï1

0

1

2

%

Regression coefficientslnIp -Td lnIp - lnCAPE

p%

p%

p%

Sim

ple

75 90 990

0.05

0.1

75 90 990

0.2

0.4

75 90 990

0.05

0.1

0.15

75 90 990

0.05

0.1

75 90 990

0.2

0.4

75 90 990

0.05

0.1

0.15

75 90 990.02

0.04

0.06

0.08

0.1

75 90 990

0.5

1

!γ758590 9899 99.999.99

0

0.02

0.04

0.06

0.08

0.1

0.12

758590 9899 99.999.99 758590 9899 99.999.99 758590 9899 99.999.99

NorthCentralEastSouthCïC

lnIp -Td lnIp - lnCAPE

75 90 990

0.05

0.1

75 90 990

0.2

0.4

75 90 990

0.1

0.2

0.3

75 90 990

0.05

0.1

75 90 990

0.2

0.4

75 90 990

0.1

0.2

0.3

p%

p%

p%

75 90 990

0.05

0.1

75 90 990

0.5

1

Mul

tiple

re

anal

ysis

M

ultip

le

soun

ding

10 20 305

6

7

8

9

75 8590 9899 99.9 99.990

0.05

0.1

0.15

%

10 20 305

6

7

8

9

75 8590 9899 99.9 99.990

0.05

0.1

0.15

%

a)

10 20 30

10

20

30

10 20 300.20.40.60.81

10 20 30

10

20

30

10 20 300.20.40.60.81

Td

T Td T Td

DJF+JJA

lnCAPElnCAPE

4 6 8

0

2

4

4 6 8

0

2

44 6 8

0

2

4

4 6 8

0

2

4

lnIp

lnIpall year DJF, JJA lnCAPE

Td Td

b)

CEN

TRA

L EA

ST

CEN

TRA

L EA

ST

10 20 300

2

4

10 20 300

2

410 20 30

0

2

4

10 20 300

2

4

10 20 300

2

4

10 20 300

2

4

10 20 30

10

20

30

10 20 300.20.40.60.81

10 20 30

10

20

30

10 20 300.20.40.60.81

all yearTd lnIp

lnIp |CAPE

AbstractWe analyze how extreme rainfall intensities in the Eastern United States depend on temperature T, dew point temperature Td and convective available potential energy CAPE, in addition to geographic sub-region, season, and averaging duration. When using data for the entire year, rainfall intensity has a quasi Clausius-Clapeyron (CC) dependence on T, with super-CC slope in a limited temperature range and a maximum around 25. While general, these features vary with averaging duration, season, the quantile of rainfall intensity, and to some extent geographic sub-region. By using Td and CAPE as regressors, we separate the effects of temperature on rainfall extremes via increased atmospheric water content and via enhanced atmospheric convection. The two contributions have comparable magnitudes, pointing at the need to consider both T and atmospheric stability parameters when assessing the impact of climate change on rainfall extremes.

Data

GC51A-0378 Friday, December 19, 2014 08:00 AM - 12:20 PM Moscone West Poster Hall

Temperature and CAPE Dependence of Rainfall Extremes in the Eastern United States Chiara Lepore (1), Daniele Veneziano (2), Annalisa Molini (3,4)

!

North

East

South

Central

S0.99

75 8590 9899 99.9 99.990

0.02

0.04

0.06

0.08

0.1

0.12

75 8590 9899 99.9 99.99 75 8590 9899 99.9 99.99 75 8590 9899 99.9 99.99

!"#"$% $#$$ $$&$$$&$$%

%&%'

%&%(

%&%)

%&%#

%&*

%&*'

!"#"$% $#$$ $$&$$$&$$ !"#"$% $#$$ $$&$$$&$$ !"#"$% $#$$ $$&$$$&$$

+,-./012.-34536.7,8./0�0

p%

a)

b)

EUS

Sp,[9-22]

EUS

Figure 2 Slope values for the lnIp-T regressions, for various percentiles for the 4 macro regions presented in Fig. 1. These slopes are calculated for a temperature range 9-22 C. In this range we observe no inflection (see Fig. 3). The overall regional slopes increase with the percentile levels and they reach values that are 150% CC-slope.

p

all year DJF+JJA

NORTH

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

EAST

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

CENTRAL

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

SOUTH

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

!" #" $""

#

%

!" #" $"�"&"'

"

"&"'

"&!

T T T T

lnIp Sp lnIp Sp

p

National Climatic Data Center (NCDC) Hourly precipitation time series - 270 station from the 6000 the National Climatic Data Center (NCDC), with not long data gaps, few missing values, and free of hourly estimates from disaggregated daily totals. They have data since 1979 and overlap the ERA records. 182 are in the EUS region we study. [270 Colored dots in Figure 1] ERA Interim - ECMWF Data span the period 1979-2008 and were bi-linearly interpolated to a spatial resolution of 0.125 from the original grid (ECMWF, 2013). Daily time series of T and CAPE for the NCDC stations were taken from the closest interpolated reanalysis cell. Mean daily air temperatures T - Data were extracted from the sub-daily screen level (2 m) temperatures of ERA-Interim. Mean daily convective available potential energy CAPE were derived from the ERA 12-hour predicted fields. NCDC Integrated Global Radiosonde Archive (IGRA) The radiosonde stations are matched to the nearest NCDC gauge station and grouped by sub-region in analogy with the other variables. [49 black circles in Figure 1] Mean daily Dewpoint Temperature Td - From the NCDC IGRA (Durre et al., 2006) from bi-daily radio-soundings (00 and 12UTC) at the m.s.l. Bi-daily CAPE from soundings - For completeness, we consider CAPE values from soundings (data and documentation available at http://www.ncdc.noaa.gov/oa/climate/igra/).

North

East

South

Central

S0.99

75 8590 9899 99.9 99.990

0.02

0.04

0.06

0.08

0.1

0.12

75 8590 9899 99.9 99.99 75 8590 9899 99.9 99.99 75 8590 9899 99.9 99.99

!"#"$% $#$$ $$&$$$&$$%

%&%'

%&%(

%&%)

%&%#

%&*

%&*'

!"#"$% $#$$ $$&$$$&$$ !"#"$% $#$$ $$&$$$&$$ !"#"$% $#$$ $$&$$$&$$

+,-./012.-34536.7,8./0�0

p%

a)

b)

EUS

Sp,[9-22]

EUS

CC slope

Figure 1 Slope map for the lnI-T 99th

percentile for the 270 CONUS NCDC stations (colored circles); the black circles are the location of the sounding stations. The EUS values are mostly CC or slightly higher than a CC slope. The rest of the US is very variable, with slopes mostly below CC rate. The rest of the analysis are focused on the EUS of the CONUS, divided in 4 regions: North, Central, South, East.

Results

lnIp Sp

d=2 and 8hr - Within storm

lnIp Sp

d=2 and 8hr - block averaging daily values

T T T T T T

5 10 15 20 25 30

2

4

6

8

10

12

14

5 10 15 20 25 30 5 10 15 20 25 30 5 10 15 20 25 30

E[D]

T T T T

i) ii) iii) iv)

758590 9899 99.999.990

0.02

0.04

0.06

0.08

0.1

0.12

758590 9899 99.999.99 758590 9899 99.999.99 758590 9899 99.999.99

NorthCentralEastSouthCïC

lnIp Sp

NO

RTH

10 20 300

2

4

10 20 30ï0.05

0

0.05

0.1

10 20 300

2

4

10 20 30ï0.05

0

0.05

0.1

10 20 30ï0.05

0

0.05

0.1

10 20 300

2

4

Figure 4 (top row) Variation of average storm duration D (hr) with temperature T for (i) all year, (ii) DJF, (iii) JJA , (iv) SON (solid) and MAM (dashed); (bottom row) Quantile and slope plots for 2 averaging durations: 2hr (colored) and 8 hr (black) within storms (Columns 1 and 2), using block averaging (Columns 3 and 4), and for daily average values (last two columns). The grey shading reproduces Figure 3. The storms are identified using a fractal method (Veneziano and Lepore, 2012). Through this method, we reduce dry periods, especially inside longer storms and the averaging duration doesn’t affect the shape of lnIp. When using block averaging the inclusion of dry periods reduces the intensities values, which are much lower than those for 1hr (grey lines). This is especially evident for d = 8 e daily averages.

Figure 3 Quantile (lnIp) and slope (Sp) plots for the 4 sub-regions for different probabilities levels, p = 0.75, 0.9, 0.99, and 0.999. The dotted vertical lines mark the 9-22C range, the black-dashed lines (1st and 3rd column) and red dashed lines (2nd and 4th column) show the CC rate (0.068). The first two columns are for the all-year analysis. The last two columns are for winter (DJF, colored lines) and summer (JJA, black lines). Results for the spring and fall seasons are close to the all-year results and are omitted. Most of the super-CC slopes are within the 9-22 range, but even the maximum local slopes remain below 1.5CC except for isolated cases. The slopes for the whole year have a local minimum around 8-10oC. The Sp plots for the winter (after the initial local minimum) and the summer have a “parabolic” shape. The “zeros” of the parabola, i.e. at about T=10 and 25 oC in the North during summer, correspond to local minima and maxima of the lnIp plots. The parabolic shape of the Sp curves corresponds to an “S” shape of the lnIp plots. The peak of the parabola tends to decrease as p increases.

Figure 5 (a) Td-T and lnI-Td quantile plots for all-year (first and second columns) and DJF and JJA superposed (colored and black lines, respectively, third and fourth columns). (b) lnI-lnCAPE plots for all-year (first column) and DJF+JJA (second column). The third column shows lnCAPE-Td quantiles and last column shows lnI-Td relationships for different CAPE classes. Td-T is a 1:1 relationship up to 22-25C. lnI-Td regressions result in a less peak-shaped relationships compared to Figure 3, but some curvature persists. lnI-lnCAPE is a very regular regression, with slope values all below 0.5. lnI-Td for different CAPE classes results in a decreasing slope value as CAPE increases, suggesting a different role of the potential energy for different CAPE ranges.

Figure 6 We considered the following equations: lnIp = c +aTd + blnCAPE (1) T=dlnCAPE (2) g = a/(a+db) (3)!The first row shows the slope coefficients of simple lnI-Td, lnI-lnCAPE and lnCAPE-T regressions; in the second row we show the coefficients for multiple regression (Eq. 1), and g (Eq. 3) The third row repeats the analysis using sounding data. In the second row The values are highly consistent across regions and vary from about 0.5 for p=0.75 (equal contribution from water vapor content and the intensity of convection) to about 0.8 for p = 0.99 (predominance of water vapor variation). When we consider sounding data, the main effect is a decrease in the lnIp-lnCAPE slope and an increase of g towards 1, with a weaker dependence of rainfall on CAPE.

Conclusions• We have evaluated the rainfall intensity-temperature (lnI-T) relation in the EUS using ground observations, soundings, and reanalysis products. • Our focus is on rainfall intensity quantiles with p ≥ 0.75. Within the EUS, the variation in the lnI-T relationships is relative small, except in winter. The differences between the EUS and

the western part of the country are much sharper. • We compared global and local slopes of the lnI-T relationships to the CC rate. The overall range of these slopes is in line with prior literature for the EUS (Shaw et al., 2011; Utsumi et

al., 2011). This includes the “peak-like structure” with maximum rainfall intensity at about 22oC. We rarely see super-CC rates. Such rates occur almost exclusively in an intermediate range of temperatures and only for the highest quantiles and rarely exceed 150% of the CC value.

• We separated the two main pathways through which temperature affects precipitation (increased water vapor and enhanced atmospheric convection) by jointly regressing lnI against Td and CAPE;

• Between rainfall intensity quantiles and CAPE, there is a power-law relationship with coefficients depending on region. These exponents are below the value 0.5 (from w ~ CAPE1/2 ), hence the energy-conversion mechanism is less efficient when CAPE is higher.

• We investigated the peak-like structure of the lnI-T curves to determine whether the temperature T at the peak varies seasonally and regionally (it does) and whether the peak is due to a limit in the amount of moisture in the atmosphere or to the fact that at higher temperatures a larger fraction of rainstorms have sub-hourly duration (the primary cause is the limit on Td).

• To quantify extreme rainfall under possible future warmer climates it is important to assess the seasonal and annual limits of Td . It is also important to quantify how CAPE will change, although CAPE is more influential on the lower quantiles of rainfall intensity.

• Understanding the root causes of the inefficiencies of the moisture-rainfall and potential-to-kinetic energy conversions and consideration of frontal versus convective precipitation could help explain the regional and seasonal patterns of rainfall intensity with temperature;

• Differences were found when using CAPE values from soundings instead of ERA Interim estimates, suggesting the need to make a more detailed analysis of CAPE-related indices.

MethodsRainfall Intensity, I, and Temperature, T, data are matched and binned into 2-deg temperature intervals 0-2, 2-4,..oC. We extracted precipitation quantiles for several non-exceedance probabilities p, resulting in a matrix of intensities Ip,T. The dependence of rainfall intensity on T is displayed through quantile plots, which are plots of Ip,T against T for different p. Through regression of lnIp,T vs. T, we infer, for each p, different measures of sensitivity of Ip to T: the overall slope Sp based on all temperature bins, the average slope Sp,[T1-T2] within a selected range of temperatures [T1-T2] , and the local slope Sp,T.

This material has been published: C. Lepore, D. Veneziano, A. Molini (2014), Temperature and CAPE Dependence of Rainfall Extremes in the Eastern United States, GRL doi: 10.1002/2014GL062247 Funding: C. Lepore acknowledges fundings from the DEES and LDEO at Columbia University. D. Veneziano and C. Lepore acknowledge NSF National Science Foundation under Grant # EAR-0910721. A. Molini acknowledges fundings from the Masdar Institute (One-to-One MIT-MI, #12WAMA1 and Flagship #12WAMC1). We are grateful to Dr. Tanvir Ahmed and Seonkyoo Yoon for the processing and quality control of the NOAA data. References Durre, I., Vose, R.S., Wuertz, D.B. (2006), Overview of the Integrated Global Radiosonde Archive, J. Clim., 19, 53– 68. ECMWF (2013), IFS DOCUMENTATION, 40 ed., ECMWF, Shinfield Park, Reading, RG2 9AX, England Shaw, S. B., A. A. Royem and S. J. Riha, (2011), The Relationship between Extreme Hourly Precipitation and Surface Temperature in Different Hydroclimatic Regions of the United States. J. of Hydr., 12(2), 319–325. Utsumi, N., Seto, S., Kanae, S., Maeda, E.E., and T. Oki (2011), Does higher surface temperature intensify extreme precipitation?, Geophysical Research Letters, doi:10.1029/2011GL048426.

g