Advertisement

Atmospheric Tracers and the Monsoon System: Lessons Learnt from the 1991 Kuwait Oil Well Fires

  • Peter Carl
Conference paper
Part of the Springer Proceedings in Complexity book series (SPCOM)

Abstract

The smoke veil over the Arabian peninsula, the Arabian Sea and adjacent areas, fed by the burning Kuwaiti oil fields from February to October 1991, turned out to provide a key climate sensitivity ‘experiment’ on the global hydrological response to anthropogenic immissions. Though the menace of setting the oil wells alight failed to work as a deterrent and the Earth’s climate did not respond with a “nuclear winter” type ‘retaliatory strike’, the system was hit at a sensitive spot. Inherent climate variability notwithstanding, the 1991 boreal summer took an exceptional turn. Effects of the disturbance were blurred by spectacular evolutions in the atmospheric methane load, the fundamental economic transformation of that time, and the largest volcano eruption of the century, Mt. ;Pinatubo. The challenging mix of political, economic, geophysical and environmental dynamics and events forms the background of the combined data analysis and climate modelling approach presented which aims to rather disentangle the complex issue. The study comprises worldwide oil and gas production as economic proxies, global trace gas loads and growth rates (CH4, CO2), Kuwait fire source strengths and scenario estimation, and related climate model experiments—which lately led to the conception of low-dimensional organization of the monsoon system. Retrospective and prospective conclusions are offered.

Keywords

General Circulation Model Wavelet Transform East Asian Summer Monsoon Match Pursuit Smoke Amount 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

47.1 The Complex Setting

Timely political action and deterrent measures to prevent the invasion were largely missing. On August 2, 1990, Iraqui troops occupied Kuwait. The means to deter liberation were apparently in their hands then: Shortly after the invasion, experiments were started at the oil fields to learn how active wells may be set alight effectively (Husain, 1994a). The international embargo on Iraqui oil export left a clear signal in the worldwide output, induced Saudi Arabia to rush into replacement production—and was ignored by the addressee as a toothless tiger. Saddam’s menacing announcement at September 23 1990, however, of ‘darkening the heaven’ in a “nuclear winter” like scenario should military action be undertaken to liberate Kuwait, failed to work as a deterrent as well. For sure, both sides had learnt their lessons on the concept of deterrence (e.g., Thunborg et al. 1981; Subrahmanyam et al. 1987), but had recourse to it only to a degree of political opportunity, ignoring important aspects of that knowledge at will in a deadly Middle East poker, fanned by the language and measures of escalation (His Majesty King Hussein, 1990). From January 17 until February 23, in the course of the “Desert Storm” air campaign, dozens of wells were deliberately detonated or caught fire due to intense allied bombing. During the short ground battle (February 24–28), retreating Iraqui troops executed their ‘scorched earth’ order to increase the stake by an order of magnitude—leaving some 720 wild wells in addition (Husain, 1994a). The ‘retaliatory strike’ by a smoke-induced climatic disruption, however, failed to appear: The system’s response took another turn. An early, torrential typhoon over the head Bay of Bengal at April 30 took a toll of 125,000 deaths (Philips, 1995). Starting a season this devastating way which elapsed dynamically abnormal throughout, the Asian summer monsoons showed unusual activity in 1991, although the climatic backscene, notably established meridional pressure gradients (cf. Sect. 47.6.1), would have favoured certain inhibition of monsoon dynamics (Carl, 1998). High/low SST anomalies were observed over the Bengal/Arabian Seas (De et al., 1992), the Mei–Yu season over East Asia commenced early by a month (Philips, 1995), and a southwards displaced Western Pacific High steered moist, warm air from the Bay into the rainbelt across China. Start–up over Kerala of the Indian southwest monsoon dated normal but came along without onset vortex (De et al., 1992) and advanced very fast, to end up in extreme spatial stagnation. The East Asian rainy season became locked for as long as two months over the Yangtse–Huaihe basin whilst Northeast China suffered from extended drought (Philips, 1995). Early in July, a broad–scale cyclonic system over the head Bay of Bengal (De et al., 1992) battered the river basin with a 1-week catastrophic rainfall that waterlogged 13 million hectares of farmland (Philips, 1995). Mt. Pinatubo (15. 1 ∘ N, 120. 4 ∘ E) exploded at June 15, at the height of the first activity spell when the southwest monsoon’s advance had just stopped—as if a valve opened to balance the abruptly changing angular momentum tendency of Earth’s rotation (where intraseasonal monsoon dynamics is known to leave its signature (Krishnamurti et al., 1992, e.g.)).
Fig. 47.1

Burning oil wells in Kuwait, photo taken by US Army SGT Perry Heimer (Public domain, source: Wikipedia; uploaded by St. Krekeler, version of Oct. 09, 2007)

Man’s economic activity, as reflected in the oil and gas production figures, ran into a period of stagnation at the beginning of the 1990s. The global atmospheric CH4 and CO2 loads indicated a concurrent regime transition, and “global warming” paused for about a decade. The general setting of crisis and war in the Gulf, at the dawn of a new era after the large systems confrontation of the Cold War, thus bears a challenging mix of diverse impacts at various spatiotemporal scales. This calls for detailed inspection to disentangle the dynamic complexity which man’s activity, notably political decision making of the time, both contributed to and had to cope with.

Source strength estimates of the Kuwait fires are given in Sect. 47.2 and are translated into scenarios. Partially unpublished results of early scenario runs using a General Circulation model (GCM) of the troposphere are sketched in Sect. 47.3. A specific glance is shed on the global hydrological response. Section 47.4 addresses the economic transformation of the time using worldwide oil and gas production as proxies, and Sect. 47.5 provides analyses of global atmospheric trace gas records, with a focus on methane. Section 47.6 offers a synthesis attempt and a range of conclusions. An Appendix provides concise methodological information and refers to the data sources used.

47.2 Kuwait Fires: Sources and Scenarios

Smoke is a perpetual, active component of the global climate system. Little was known about its role in shaping past and present climates, however, when research in a crisis-like mode (MacCracken and Penner, 1987) was triggered by the advent of the “nuclear winter” theory (Crutzen and Birks, 1982; Turco et al., 1983) on climatic and environmental impacts of mass fires and firestorms in a nuclear war (cf. also Carl et al. 2008). Early attention to direct atmospheric and surface effects beneath large-area smoke plumes (Wexler, 1950) refreshed at the beginning of the 1980s (Seiler and Crutzen, 1980) to run into a bulk of studies by the end of the decade (e.g., Cachier et al. 1989; Ferrare et al. 1990; Setzer and Pereira 1991; Veltishchev et al. 1988; Robock 1988; Andrae et al. 1988; Segal et al. 1989; Hulstrom and Stoffel 1990; Pueschel and Livingston 1990). In 1991, when the smoke plumes ascended for more than 8 months from the burning oil fields in Kuwait, sound indication existed that smoke noticeably contributes to the impact on climate, environment and man of hundreds of thousands tropical vegetation fires per year.

Though unsteady and localized in each single case, smoke from natural fires is a systematically varying source of tropospheric heating anomaly due to absorption of shortwave solar radiation. Anthropogenic biomass burning in the tropics, which exerts a more continuous stress, bears amounts of carbonaceous particulate emissions comparable to those in midlatitudes. Seasonality effects differ in their causation, however (wet/dry vs. cold/warm). For the Mediterranean belt interannual fire–climate feedbacks appear to act via ecosystem dynamics (Swetnam and Betancourt, 1990). Rather short tropospheric residence times of smoke notwithstanding (from a couple of days to about 3 weeks; Pruppacher and Klett 1978; Giorgi and Chameides 1986; Ferrare et al. 1990), planetary-scale teleconnections were known for long to emanate from low-latitude surface heating anomalies (e.g., Bjerknes 1969; Lau and Lim 1984; Garcia and Salby 1987; Leathers et al. 1991). This justified the suspicion, the persistent smoke-induced thermal anomaly of the lower troposphere over the Gulf region (in a prevalent anticyclonic atmospheric environment) could influence the 1991 seasons at planetary scale.

Anticyclonic conditions often precede and accompany large-scale forest fires, at that also restricting the smoke injection heights (Golitsyn and Phillips 1986; Setzer and Pereira 1991). Though ascent of substantial smoke amounts into the upper troposphere and their long-range transport did not happen (which otherwise could have induced remote direct impacts), the region carries the strongest equator-crossing jet stream of the lower troposphere in boreal summer, the Somali jet (Findlater, 1969), which is an integral part of the circumglobal summer monsoon system. Depending on the source strengths, indirect remote effects were thus to be expected.

47.2.1 Source Strength Estimation: Trace Gas Loads

Early voices that warned of globally catastrophic results of a large confrontation ‘literally on top of the single richest petroleum reservoir in the world’ (His Majesty King Hussein, 1990) focused on gaseous releases, notably CO, CO2, SO2, and on economic consequences. A substantial promotion of the CO2-induced greenhouse effect, however (and the amount of heat released by the fires all the more), were not among the issues of primary scientific concern on potential environmental consequences of war in the Gulf. Still, first figures became known in this context about the volume of oil that could go up in flames (Aldhous, 1991): 3⋯10 Mio. barrels per day (Mbpd; 1 b = 159 l). To seriously estimate smoke amounts and other emissions, details about the oil fields were needed. Table 47.1 summarizes relevant knowledge that was readily available (Anonymous 198819901991).
Table 47.1

Onshore oil fields of Kuwait and the Kuwaiti administered part of the Neutral Zone at the turn of the 1980s: basic parameters and summary production (Anonymous 198819901991)

Field

Year of

Wells

Conversion

Production (kt/day)

     

name

discovery

Active

Total

b/t

 ∘ APIa

Year

Total

Per well

     
  

Kuwait (total)b

    
  

363

782

7.29c

32.5c

1988

183

0.50

     
      

1989

218

0.60

     
      

1990

220

0.61

     

Greater Burgan

 

292

590

          

Burgan

1938

210

393

7.21

30.8

        

Magwa

1951

71

113

7.31

33.0

        

Ahmadi

1952

11

84

7.25

31.7

        

Raudhatain

1955

41

53

7.37

34.4

        

Sabriyah

1957

9

44

7.46

36.3

        

Minagish

1959

1

21

7.37

34.4

        

Umm Gudair

1962

20

33

7.04

26.9

        
  

Neutral zone (total)b

    
  

343

450

6.83d

22.1d

1988

19.9

0.06

     
      

1989

18

0.05

     

Wafra

1953

322

427

6.68–6.89

18.9–23.5

1988

14.4

0.04

     
    

6.79c

21.2c

        

South Fuwaris

1963

5

9

6.96

25.0

1988

0.3

0.06

     

South Umm Gudair

1966

16

17

6.92

24.3

1988

5.2

0.33

     
  

Kuwait & neutral zone (total)

    
  

706

1,232

7.25e

31.7e

1988

202.9

0.29

     
      

1989

236

0.33

     
      

1990

240

0.34

     

a  ∘ API \(\doteq141.5/S - 131.5\), where S (in kg/l) is the specific gravity of the crude oil at 60  ∘ F (10  ∘ API corresponds to S = 1, the specific gravity of water)

bFields not listed include Ratga (30 wells), Khashman (3), Medina 1 (3), Bahrah (2), Mutriba 3 (2) and Rugei 1 (1) in Kuwait, as well as Arq (1) in the Neutral Zone (number of active wells, conversion factors and production figures unknown for these fields)

cArithmetic mean (for Kuwait, facing lack of more detailed information)

dWeighted mean

eWeighted between Kuwait and NZ

Based on the conversion factors given (the worldwide mean amounted to ∼ 7.33 b/t at that time), Table 47.2 provides ‘back–on–the–envelope’ estimates of the gaseous emissions in question, assuming all the oil released burns: CO, CO2, and CH4 or higher hydrocarbons (collectively labeled C x H y ). A more realistic assessment has to take into account evaporation of unburnt oil. For 90 and 75 % burning (oil scenarios unchanged), Table 47.3 substantiates that it is CH4 which may bear a problem among the gaseous emissions. Similar estimates may be possible for nitrous oxides, but not for photochemical smog, notably tropospheric Ozone, which call for detailed modelling of the chemical system (Carl, 1991a). A blend of Tables 47.2 and 47.3, specified for the GCM scenarios that were lately run, may be found in Carl (1991b).
Table 47.2

Estimated contribution to global atmospheric loads and growth rates (with reference to 1990) of carbonaceous trace gas immissions from the burning oil fieldsa

Oil burnt

COb

CO2

C x H y

CO2 equivalent

 

Mbpd

ppmv

%

ppmv

%

ppmv

%

ppmv

%

 
   

Total load, 1990

 
 

0.23

 

353

 

1.7

 

42.5

12.0

 
   

Annual growth, 1990

 
 

0.004

 

1.6c

 

0.017

 

0.4

26.6

 
   

Annual growth due to Kuwait fires

 

1.8

0.002

44

0.03

1.8

0.0007

4.2

0.02

1.1

 

3.0

0.003

73

0.05

3.1

0.0012

6.9

0.03

1.8

 

6.0

0.006

146

0.10

6.1

0.0024

14

0.06

3.7

 

10.0

0.010

245

0.16

10.2

0.0039

23

0.10

6.1

 

aAssumptions: (i) the total oil volume released burns, (ii) 1 year duration, (iii) elemental carbon (C) contents of oil (by mass) 84.5 %, (iv) mass ratio methane/ethane as 9/1, other hydrocarbons neglected (in this mass balance), (v) CO2 equivalent of CH4: factor 25, (vi) all hydrocarbons taken as CH4 in terms of their CO2 equivalent, (vii) Carbon consumption of CO2/soot/CO/C x H y as 83/10/5/2

bNorthern hemisphere

c(1. 3⋯1. 8)

Table 47.3

Estimated contribution to global atmospheric loads and growth rates of hydrocarbon immissions from evaporating unburnt oila

 

90 % of oil burnt

75 % of oil burnt

 

Oil volume

C x H y

CO2 equiv.

C x H y

CO2 equiv.

 

released Mbpd

ppmv

%

ppmv

%

ppmv

%

ppmv

%

 
 

Annual growth due to sabotaged Kuwaiti oil fields

 

1.8

0.0016

9

0.04

2.5

0.0039

23

0.10

6.1

 

3.0

0.0026

15

0.07

4.1

0.0065

38

0.16

10.2

 

6.0

0.0052

31

0.13

8.2

0.0131

77

0.33

20.4

 

10.0

0.0087

51

0.22

13.6

0.0218

128

0.54

34.0

 

aAssumptions: 50 % of the unburnt oil evaporates; otherwise as in Table 47.2

By and large, these initial estimates stood the test of measurement campaigns (taking the factual duration of the fires into account where necessary). Unfortunately, those campaigns in general took too much time to be prepared so as to provide a sound scientific data base for the first weeks to months following the destruction of the Kuwaiti oil fields and industry. Whereas demolitions during the airborne operations prior to February 24 may have released an average of 65 kt per day (Carl, 1991a), i.e. the non-negligible volume of some 0. 45⋯0. 5 Mbpd, the oil loss due to sabotaged oil fields jumped to a figure of 10 % of worldwide annual oil consumption (Seacor, 1994, e.g.), notwithstanding the fact that the pre-occupation Kuwaiti oil production (about 220⋯240 kt/day) covered only 2.5–3 %. A closer look at production conditions of the region may help understanding this fact (which appears to be a generally accepted one today). Together with the scenarios developed in Sect. 47.2.2, Tables 47.147.3 provide a basis as well for estimates of a CH4 scenario (a preliminary one is given in Carl (1998) and Sect. 47.5.2) which may further help coming to grips with global trace gas evolutions at the turn of the 1980s, notably the spectacular methane fluctuation of the early 1990s.

47.2.2 Smoke Scenarios

A detailed foundation of smoke emission scenarios called for information that was hardly available by the end of 1990—and perhaps is not even today. A conceptual approach has thus been adopted as follows: The amount F of combustible (oil) is thought to be composed of fractions according to
$$\displaystyle{ F =\pi \sum _{i}\varphi _{i}\ n_{i}\, }$$
(47.1)
where \(\pi = {P}^{{\ast}}/{N}^{{\ast}}\) is the average productivity, P  ∗  the total daily production, N  ∗  the number of active wells, i the type of well (classes cf. below), \(n_{i} =\nu _{i}\ {N}^{{\ast}}\) the number of sources of type i, and ν i their relative proportion (\(0 \leq \nu _{i} \leq 1\)). Factor \(\varphi _{i} = p_{i}q_{i}f_{i}\) weights the scenario contribution by source type i, p i is the productivity of this source type in relation to the average productivity π, q i is an enhancement factor that takes freely exhausting oil from source type i into account, and f i is the relative proportion of wells of type i that went up in flames (0 ≤ f i  ≤  1). Six types have been distinguished (Table 47.4).
Table 47.4

Source type specification of the conceptual approach to Kuwait fire scenarios

Well type

Source activity

Pressure conditions

Field status

 

1

Eruptive

Natural overpressure

Primary recovery

 

2

Non-eruptive

Natural overpressure

Primary recovery

 

3

Eruptive

Natural overpressure

Secondary recovery

 

4

Non-eruptive

Natural overpressure

Secondary recovery

 

5

Non-eruptive

Unclear

Secondary recovery

 

6

Inactive

Unknown

Unknown

 

At fields under primary recovery, oil extraction exclusively relies on natural overpressure and gas contents, whereas under secondary recovery supporting measures maintain optimal production conditions—which does not mean there is no natural overpressure. Only a few sources of the region ( ∼ 20) were said to have lost overpressure. Parameters have to be chosen in accordance with conditions \(\sum _{i=1}^{5}\nu _{i}\stackrel{!}{=}1\), \(\sum _{i=1}^{5}p_{i}\nu _{i}\stackrel{!}{=}1\) (or \(\sum _{i=1}^{5}p_{i}n_{i}\stackrel{!}{=}{N}^{{\ast}}\)) and \(\sum _{i=1}^{5}f_{i}n_{i}\stackrel{!}{=}{N}^{f}\), where N f is the total number of active wells ablaze (N f  = 600 here, i.e. 85 % of all 706 active wells). Table 47.5 displays conservative choices (cf. also Small 1991).

Of the six scenarios developed in Carl (1991a) three were considered for GCM experiments: “nominal smoke” (KOWF1; daily pre-occupation production without Neutral Zone: 220 kt/day (i.e. ∼ 1.6 Mbpd), a medium case (420 kt/day), and an extreme one (821 kt/day). For the sake of comparability to other studies, though, the latter two were approximated by the twofold (KOWF2) and fourfold nominal case (KOWF4), respectively (Carl 1991b,c). The scenario considered as “extreme” at that time ( ∼ 6.4 Mbpd) uses smoke amounts close to the figure of up to 6 Mbpd officially delivered afterwards (Seacor, 1994, e.g.). The “medium” case settles in the vicinity of another pre-war estimate (Cox, 1991), taken for ‘most probable’ then (when comparing the balances above with Table 47.5, consider roundoff due to integer n and the number of decimals given for p, f, ν).
Table 47.5

Two scenarios of burning oil amountsa according to Carl (1991a) (Kuwait (K) and the Kuwaiti administered Neutral Zone (NZ) are separately balanced)

 

“Medium” scenario

“Extreme” scenario

 

Source

Kuwait

Neutral Zone

Kuwait

Neutral Zone

 

type i

p

q

f

n

ν

p

q

f

n

ν

p

q

f

n

ν

p

q

f

n

ν

 

1

2

2

1

55

0.15

2

2

0.8

34

0.1

2

4

1

55

0.15

2

4

1

34

0.1

 

2

1.5

1.5

1

18

0.05

1.5

1.5

0.8

34

0.1

1.5

3

1

18

0.05

1.5

3

1

34

0.1

 

3

2

2

1

55

0.15

2

2

0.8

34

0.1

2

3

1

55

0.15

2

3

1

34

0.1

 

4

0.7

1.3

1

91

0.25

0.7

1.3

0.8

102

0.3

0.7

2

1

163

0.45

0.7

2

1

172

0.5

 

5

0.4

0.5

0.63

144

0.40

0.6

0.5

0.92

139

0.4

0.05

1

0.25

72

0.20

0.5

1

0.26

69

0.2

 

6

\(\varphi = 0.2\)

419

 

\(\varphi = 0.2\)

117

 

\(\varphi = 0.5\)

419

 

\(\varphi = 0.5\)

117

  

aAssumptions: π K  = 0. 6, π NZ  = 0. 05, \(N_{K}^{{\ast}} = 363\), \(N_{NZ}^{{\ast}} = 343\), N K f  = 309, N NZ f  = 291

Translation into smoke emission scenarios for use in the GCM follows other simple arithmetics (Penner, 1986, e.g.), namely
$$\displaystyle{ r =\epsilon \ s\ \eta \ F\, }$$
(47.2)
where r is the amount of smoke, ε the emission factor, s the proportion that stabilizes in the atmosphere, η (\(= {N}^{f}/{N}^{{\ast}}\)) the part of oil which burns under smoke emission, and F the amount of fuel, for instance according to (47.1). For emissions from uncontrolled oil fires values of ε = 0. 06⋯0. 1, s > 0. 8 and η > 0. 75 had been found in the “nuclear winter” context (Turco et al., 1990). These data were partly obtained for primary storages and tend to yield ‘conservative’ estimates of r in the Kuwait fire case. A cautious choice yields r = 0. 055 F (ε = 0. 075, s = 0. 85, η = 0. 85), a less conservative one r = 0. 08 F (ε = 0. 1, s = 0. 9, η = 0. 9). That is, ∼ 5.5 ⋯8 % of the oil released might have changed into smoke (23. 1⋯33. 6 kt/day of smoke for the “medium” scenario, and 45. 2⋯65. 7 kt/day for the “extreme” one). Figures reported in the literature vary considerably (5⋯100 kt/day; (Seacor, 1994, e.g.)), depending on the time and location of measurement campaigns. In the study (Bakan et al., 1991) conducted at the Max-Planck-Institute for Meteorology (“HH”) r = 0. 085 F has been used.

For the climatic effect of these emissions, smoke ‘absorptivity’ with respect to shortwave solar radiation is another decisive parameter. Depending on the solar zenith angle θ, insolation I 0 is exponentially attenuated according to \(I = I_{0}\exp (-\tau /\cos \theta )\). The turbidity of the atmosphere is expressed as ‘optical depth’ \(\tau = \Phi \ m/A\), where m is the mass of the optically active substance, A is the area over which it is spread, and Φ is the ‘cross section’ by which the substance brings its optical activity into effect. Without going into further detail here, the absorption cross section used in Bakan et al. (1991) amounts to 8 and 10 m2/g is used in the study referred to in Sect. 47.3 (Carl 1991b,c), taking into account that scattering further reduces the direct effects of insolation but has not been considered in this latter assessment.

47.3 GCM Based Analyses

The GCM used is of coarse spatial resolution (Fig. 47.2, left panel), but has a number of (formal) degrees of freedom by two orders of magnitude above that of the Lorenz system (Lorenz, 1982). Origin and properties of the model (Gates et al. 1971; Aleksandrov and Gates 1981) have been sketched in Carl (2013c). Earlier versions were used for studies on the climatic consequences of nuclear war (Thompson et al. 1984; Stenchikov and Carl 1985; Carl and Stenchikov 1988; cf. also Carl et al. 2008). The version referred to here is a regenerated one (Carl, 1988) that unveiled its intriguing monsoon dynamics (Carl, 1992) just in the Kuwait fire study (Carl 1991b,c). GCM projections on the emerging problem—one control and three scenario runs, all started March 1 from annual mean initial conditions and run for one year (simulations done before the fact)—showed sensitive, oscillatory response during boreal summer to the smoke loads (Carl 1991b,c; Fig. 47.2, right panel).

Fig. 47.2

Left panel: 12 ∘ lat × 15 ∘ lon land–sea mask of the coarse resolution GCM; right panel: GCM response to lower tropospheric smoke immission scenarios of the Kuwait oil well fire study (Carl (1991a,b); constant immission rates start March 1 each): Control run (no smoke; thick line), lower edge scenario KOWF1, reference case (KOWF2; dashed), upper edge scenario (KOWF4; dotted); shown are zonally averaged evolutions of surface pressure at latitudes 30 ∘ N and 54 ∘ N (Source of figure: Carl (2013c))

Though the initial immissions were assumed to be confined to the lower troposphere, and an efficient smoke removal scheme due to dry deposition and ‘washout’ was used (Stenchikov and Carl, 1985), the GCM’s intraseasonal oscillations became pronounced with increasing smoke amounts in the higher scenarios (Fig. 47.3; alias “HH”
Fig. 47.3

Evolving smoke loads and their (hemispherically averaged) optical effect for scenarios KOWF1 (alias ‘HH’ here), KOWF2 (‘2 × HH’) and KOWF4 (‘4 × HH’), assuming 1 year of burning oil; left panel: Lower (full line) and upper (dashed line) tropospheric smoke loads (in μg/kg); right panel: absorption optical depth (Source of figures: Carl (1991b))

refers to the precursor manuscript of study Bakan et al. 1991). The response of the model’s water cycle, and notably of its Asian monsoon system(s), to the smoke emission scenarios is shown in Figs. 47.4 and 47.5. The water vapour contents of the northern troposphere increases substantially with the smoke amount, but not symmetrically to the seasonal march of the control run (Fig. 47.4, left panel). Build-up of the extra water budget proceeds earlier, whereas its ‘mining’ in autumn remains largely unchanged. The precipitation response (center panel) clearly hints at ever higher excited intraseasonal dynamics of an interhemispherically organized, planetary system—the global monsoon. The right panel of Fig. 47.4 shows corresponding changes in the Indian summer monsoon (ISM), including somewhat earlier onset and higher intraseasonal activity with more pronounced break(s). Whereas the GCM’s Southeast Asian summer monsoon (SEASM; Fig. 47.5) appears to be largely controlled by the South Asian branch (similarity with ISM response, Fig. 47.4), the East Asian summer monsoon (EASM) shows substantially different behaviour in accordance with its (observed and modelled) undisturbed dynamics. There is no premature GCM response here of the pre-monsoon (“Mei-Yu”) phase, but it becomes increasingly shortened with the amount of smoke injected. For Northern China this bears substantial loss of water during the season indeed, whilst the South experiences increasing EASM activity. The earlier termination there of the Mei-Yu season gives way to heavy rainspells, with ever shorter breaks as the smoke amount increases. This raises the risk of flooding—as was the case in summer 1991, when the Yangtse valley suffered from a series of severe monsoon spells and catastrophic inundation.
Fig. 47.4

Hydrological response to Kuwait oil fire scenarios (Carl 1991a,b; alias as in Fig. 47.3)—control run (thick line), KOWF1 (thin), KOWF2 (dashed), KOWF4 (dotted); left panel: Changing lower tropospheric water vapour contents over the northern hemisphere (NH; mixing ratio in g/kg); center panel: hemispherically averaged precipitations (in mm/d); right panel: Indian summer monsoon precipitations (in mm/d) (Source: Carl (1991b))

Fig. 47.5

SEASM and EASM response to Kuwait oil well fire scenarios: control (thick lines), KOWF1 (thin), KOWF2 (dashed) and KOWF4 (dotted) (Source of figure: Carl (1991c))

That the remote hydrological impact described of smoke immissions over Kuwait is part in this GCM of circumglobal dynamical change, demonstrates Fig. 47.6. Given the high intraseasonal variability
Fig. 47.6

Zonally averaged GCM response to lower tropospheric smoke immission scenarios of the Kuwait oil well fire study (Carl 1991a,b); left panel: Change of circulations in the ‘nominal smoke’ case (KOWF1, thin line) as compared with the control run (thick line) for (top to bottom) (a) mid-tropospheric (600 hPa) mass convergence (in hPa m2/s), (b) lower tropospheric (800 hPa) equator crossing meridional winds, (cg) upper tropospheric (400 hPa) zonal winds at the equator and at 12  ∘ N, 24  ∘ N, 36  ∘ N, 48  ∘ N, respectively (all winds in m/s); center panel: 48  ∘ N 400 hPa zonal winds (in 10 m/s) vs. cross-equatorial 800 hPa meridional winds (in m/s) as time series (left column) and phase plots (right column) for (top-to-bottom) (a) control run, (bd) scenarios KOWF1, KOWF2, and KOWF4, respectively, (e) ‘calendar average’ of cases (ad), (f ) ‘tuned composite’ of cases (ad), adjusted to the date of the GCM’s monsoon onset each; right panel: ‘tuned composite’ circulation phase plots as before, but for four northern latitudes (top-to-bottom): 48  ∘ N, 36  ∘ N, 24  ∘ N, 12  ∘ N (Source of figures: Carl (1991c))

known from the real system (and the somewhat lower than observed excitation of the control run’s monsoon dynamics), the four simulation results have been used to form a preliminary composite. It clearly displays fundamental features of observed intraseasonal monsoon dynamics (Krishnamurti and Subrahmanyam 1982; Krishnamurti et al. 1985; Chen et al. 1988). A ‘civilian’ study using the same GCM led to a hypothesis on the dynamic constitution of the global monsoon system (Carl 1992; Carl et al. 1998) which has been further substantiated and worked out as sketched in Carl (2013a) and roughly outlined in Sect. 47.6. A comprehensive summary presentation is given in Carl (2013c). As a ‘natural’ (not astronomically driven) oscillator in the 40–60 days band, with interacting regional branches, this GCM’s boreal summer monsoon rules a distinct global climate regime of the season (Carl 1994; Tschentscher et al. 1994); cf. also top panels of Fig. 47.7. The GCM’s undisturbed seasonal march of rainfall, for grid cells representing the South and East Asian summer monsoon branches respectively (Carl, 1994), is shown in comparison with observations (Krishnamurti and Bhalme 1976; Lau et al. 1988) in the bottom part of Fig. 47.7. Though the simulated East Asian rainfall amounts to about half the climatological level there, whereas the Indian monsoon rains are stronger than observed (a model version was used in this case with three activity spells per season, as observed (Webster et al., 1998)), the structural features of both are striking. This lends credence to the suspicion that worldwide hydrological extremes of the boreal summer season 1991, in the first instance the early typhoon over the head Bay of Bengal and the catastrophic course the East Asian monsoon has taken, were remote results of the occurrences on the Kuwaiti oil fields.
Fig. 47.7

“Summer monsoon solution” of the coarse resolution GCM under climatological forcing (Carl, 1992): Seasonal march of global rainfall and upper troposphere zonal winds (top) as well as regional rainfall over the Indian subcontinent (bottom left) and Southeast Asia (bottom right) Carl (1994) vs. observation (Krishnamurti and Bhalme 1976; Lau et al. 1988) (Source of figure: Carl (2013c))

47.4 Man’s Economic Activity: Oil & Gas Production

In a fossil energy–driven world the oil and gas production reflects man’s economic activity. Figure 47.8 shows monthly outputs from May 1982 to August 1997 (OGJ 1982–1992; OGJ Data book 1984–1998), centered around the period of interest here. Against the backscene since the turn of the 1980s of a major economical transition due to disintegration of an empire, the exceptional impacts the early 1990s witnessed also stamped their marks on the world economy. The intended “doomsday” type retaliation ended in flat environmental, economic and political crime, but the two perturbations of different scale should have went down into the climate record as detectable and separable signals during the study period.
Fig. 47.8

Monthly oil (in Mbpd; left panel) and gas output (in bcf/d, i.e. 109 cubic feet per day, 1 cf = 0. 028317 m3; right panel) worldwide, and of the major producers: ME—Middle East, FSU—(former) Soviet Union, SA—Saudi Arabia, USC—United States and Canada, WE—Western Europe; the shaded area beneath the SA oil production curve marks the crisis mode replacement production after Iraq’s Kuwait invasion (world production curves shifted as indicated in the insets); data period 05/1982–08/1997

Features of the production curves which are visible to the naked eye include: (i) stagnation in the worldwide oil and gas outputs during the first half of the 1990s, largely caused by decline of the (former) Soviet Union (FSU) production and parallel recovery in Kuwait’s oil sector, (ii) sharp setback in the Middle East (ME) oil production after Iraq’s Kuwait invasion, followed by Saudi Arabian replacement production—which stayed on stream after liberation, (iii) very smooth ME oil stream after about 1993 as compared with the 1980s, and (iv) increasing (decreasing) seasonality in FSU (USC) gas production, also since about 1993.

Stabilization of ME oil production mainly resulted from the fact that Saudia Arabia (SA) could seize the chance of reaching outputs that were not negotiable before under OPEC regime. An environmental effect of this quick response to the situation (within less than a month) might result from activating wells in a crisis mode which were already closed or held for later production. At least initially, part of them should have been run using incomplete or outdated equipment and/or insufficiently trained personnel. Earlier peaks in SA production took months to reach their maxima (cf. Fig. 47.8). For the (F)SU, the loss of embargoed Iraqui oil (as a balance of debts) fed back negatively on the economic transformation, when COMECON trade rules gave space to the market, leading to new balances in the FSU oil & gas sector. Intensifying seasonality of FSU gas production toward the end of the record shown was certainly caused by increased export to Western Europe. Concurrent loss of seasonality in worldwide oil production, where it was most pronounced around 1988–1993 and almost completely ceased to exist, probably reflects a lasting change in northern midlatitude winter heating systems.

Quantitative studies are required to clarify potential impacts of oil and gas production on atmospheric trace gas loads and climate. Time series of Fig. 47.8 have been analyzed to this end by means of (i) iterative Singular-System Analysis (SSA (Vautard et al., 1992)), (ii) the Wavelet Transform (WT Torrence and Compo 1998), and (iii) the Matching Pursuit approach (MP (Mallat and Zhang, 1993)) to sparse data approximation (cf. Appendix). In the latter case a frequency-modulated (FM) ‘Gabor atom’ has been used as analysing waveform, extending the signal space (and thus the adaptive capability of the method) by three dimensions this way. For the world production series of Figs. 47.847.10 show slow and seasonal components, respectively, as obtained by SSA and MP-FM (Fig. 47.9) or MP-FM alone (Fig. 47.10).
Fig. 47.9

Slow components in worldwide monthly oil (in 0.2 Mbpd) and gas production (in bcf/d) as obtained by iterative SSA (left panel (Carl, 1998)) and MP-FM (right panel), respectively; linear SSA trends before stagnation are shift-copied to demonstrate their match after the episode, and are mirrored in the MP-FM figure for comparison (05/1982–08/1997)

Iterative SSA has a filtering effect which is dispensed with in the MP-FM approach. The right panel of Fig. 47.9 shows composites of the two leading slow MP-FM modes each; i.e. an approximation of the slow part of oil & gas production by these two unmixed modes. To demonstrate reasonably good capture, the SSA trends are retained at their original positions.

Fig. 47.10

Seasonal composites and their contributing MP-FM modes in monthly worldwide gas (in bcf/d; left panel) and oil production (in 0.2 Mbpd; right panel); 05/1982–08/1997

Figure 47.11 shows the leading five modes each of monthly worldwide gas and oil production of Fig. 47.8, and Table 47.6 presents MP-FM “structure books” providing concise quantification of all the modes displayed. This illustrates MP-FM approximations and demonstrates the high flexibility of the “Gaussian logon”. MP-FM spectrograms, i.e. time–frequency (TF) representations, of monthly worldwide gas and oil production are presented in Fig. 47.12 (due to an ‘uncertainty relationship’ in signal processing, the thickness of traces representing individual modes is inversely related to their scale s; further, sequences of spots merely result from plot resolution and are to be read as the continuous traces they mimic). An eye-catching quasi-quadriennial (QQ) modulation and a slower yet deeper mode of ‘subharmonic’ FM (ca. 8 years) in the gas production series call for confirmation and explanation. Toward the end of the record shown, a ‘gas rush’ (mode #5) is clearly represented as an isolated feature. Fading volatility in the oil production during the same period is nicely represented by the loss of higher-frequency components.
Table 47.6

MP-FM structure books of worldwide gas and oil production (05/1982–08/1997) centered analyses, performed in monthly steps (start 1, step 1); to correctly interpret the u k values given, transform the time axis from monthly to yearly (as done with the figures) by \(t_{y}:= (t_{m} + 3.5)/12 + 82\); f, \(\tilde{f}\) in cycles per month (cpm)

k

α k

s k

u k

f k a

ϕ k

\(\tilde{f}_{k}\)

β k

\(\tilde{\phi }_{k}^\prime\)

 

Worldwide gas production (in bcf per day)

 

1

293.43

1,024

24

0.0023

-2.450

0.00049

0.28

-0.393

 

2

168.03

512

105

0.0923

-0.051

0.00107

8.54

3.142

 

3

78.29

1,024

156

0.0274

1.024

0.00116

18.26

-2.553

 

4

26.01

1,024

180

0.0599

-3.118

0.00288

14.10

-0.589

 

5

17.83

32

177

0.1424

-1.814

0.01105

8.76

1.963

 

6

16.02

512

108

0.0503

1.224

0.00343

8.77

3.142

 

7

14.63

128

119

0.2102

2.091

0.02026

2.86

-0.589

 

8

12.91

256

79

0.0549

-1.302

0.00408

9.14

1.374

 

9

11.84

256

91

0.2500

0.815

0.00748

25.40

0.785

 

10

10.35

256

93

0.2611

3.000

0.01058

17.01

-1.571

 

Worldwide oil production (in Mb per day)

 

1

-44.88

2,048

11

0.0024

0.570

0.00056

0.23

-1.963

 

2

14.07

128

114

0.0089

-2.075

0.00127

2.85

-1.374

 

3

8.55

128

61

0.0549

0.497

0.00374

8.77

-1.374

 

4

6.74

64

31

0.1098

1.192

0.01372

3.69

-2.356

 

5

5.52

512

120

0.0776

1.371

0.00232

25.83

2.749

 

6

4.44

128

53

0.1693

0.815

0.00507

29.38

3.142

 

7

3.79

128

61

0.1693

-2.520

0.00781

10.01

1.178

 

8

3.22

128

60

0.2195

-0.676

0.00781

5.99

-1.178

 

9

2.94

512

85

0.0599

-2.620

0.00253

18.26

0.982

 

10

2.77

128

75

0.2611

-1.438

0.00852

15.26

1.767

 

aIn the structure books heretoforth component k of carrier frequency f c is written f k

Fig. 47.11

Leading MP-FM components, and their signal envelopes, of worldwide monthly gas (in bcf/d; left column) and oil production (in 0.2 Mbpd; right); 05/1982–08/1997

Fig. 47.12

MP-FM spectrograms (modes #1–10 each) of monthly worldwide gas (left panel) and oil production (right); 05/1982–08/1997; frequency unit: cycles per month (cpm)

As proxies of man’s economic activity, leading MP-FM modes of both time series might have left their signatures in other data of the system. Though the climate study (Carl, 2013b) uses annual aggregates and cannot provide clearcut connections, the leading MP-FM mode of the North Atlantic Oscillation (NAO) and the second one of the Southern Oscillation (SO) are both drifting across the period range of ∼ 8 years during the timespan under scrutiny (Carl 20112013b). Anyway, the data first to look at are atmospheric trace gases.

47.5 Atmospheric Trace Gas Loads: CH4, CO2

Of the gaseous emissions blamed for man-made climate change, notably CO2 and CH4 (Fig. 47.13), methane is directly linked to oil and gas outputs via venting at the well heads (of oil associated gas) and leakage rates in production, transport and consumption. Other major anthropogenic contributors include livestock and agriculture, coal mining, landfills etc., whereas temperature or water table fluctuations and wildfire feed natural emissions. To the extent that signal structures are discernible and separable, production related trace gas loads might be identified by means of sparse approximation. There is reasonable confidence that MP-FM bears this power (Carl 20112013b), but alternative methods may broaden the perspective.
Fig. 47.13

Left panel: Time series, centered over their respective data periods, of CH4 (07/1983–12/1998; in 5 ppbv) and CO2 (01/1979–12/1998; in ppmv); right panel: phase plot CH4 vs. CO2 (07/1983–12/1998; all slow MP-FM modes each)

47.5.1 The Methane Fluctuation at the Early 1990s

Figure 47.13 shows for 1991 a CH4 load increase that reflects an exceptional growth rate fluctuation which challenged conceptual insights into atmospheric trace gas loads and their relation to climate variability and change (Dlugokencky et al. 1994a,c; Hogan and Harriss 1994; Dlugokencky et al. 1994b; Lowe et al. 1997; Lelieveld et al. 1998; Dlugokencky et al. 1998). Concurrent anomalies appeared to extend over the atmospheric carbon cycle (Dettinger and Ghil, 1998, e.g.). The SSA based study (Carl, 1998) confirms a quasi-biennial pulse in the de-seasonalized CH4 growth rate (Dlugokencky et al., 1998, e.g.) which suggests a general role in the CH4 balance of the Tropospheric (quasi-) Biennial Oscillation (TBO Meehl 1997)—and its amplification into the 1991–1992 event. Figure 47.14 presents the detailed growth rate analysis (Carl, 1998) which sheds a specific (iterative SSA based) glance at climate dynamics in the back.
Fig. 47.14

Iterative SSA–based analysis of the signal at the early 1990s in the CH4 growth rate (period of analysis 07/1983–12/1997 here; from Carl (1998)): (a) slow approximation of the deseasonalized time series (violet) in comparison with the NOAA analysis (grey (Dlugokencky et al., 1998)); arrows mark the upward/downward swings, both departing from ‘normal’ up to ∼ 5 ppbv/a; removal of the nonlinear trend (pink) gives the dotted violet curve which may be further decomposed into a quasibiennial mode (TBO, red) and a slower, ‘quasi-quadriennial’ one (QQO, blue); (b) seasonality analysis in terms of envelopes of the seasonal cycle (SC) mode (dot-dashed green) and of the semiannual one (dot-dashed orange; both vertically displaced); removal of the nonlinear SC envelope trend gives the dotted green curve; (c) normalized view of phase relationships between SC (full green line), detrended SC envelope, TBO and QQO (as in (a) and (b)); phase coincidences emphasized; (d) further TBO decomposition into two modes which develop into a more strictly biennial (black, showing QQO phase synchrony during the event) and a quasi-triennial one (turquoise)

Figure 47.14a decomposes the de-seasonalized CH4 growth rate approximation into quasi-biennial (‘TBO’) and quasi-quadriennial (‘QQO’) modes. Seasonality is addressed in Fig. 47.14b, where the envelope of the basic seasonal cycle (SC) mode, i.e. the signal energy in the SC, is shown to synchronize with the CH4 event of Fig. 47.14a, reflecting internal phase–amplitude synchrony of the time series. Remarkably, the envelope of the semiannual cycle (and similarly of higher harmonics, 1/3 a, 1/4 a; not shown) exhibits a decrease in (inverse) analogy to the slow evolution of worldwide oil & gas output (Fig. 47.9). This might hint at a simplifying SC structure of the CH4 growth rate with the increasing anthropogenic impact of the data period.

Figure 47.14c combines SC, TBO and QQO dynamics in a normalized view that unveils temporary phase synchronies toward the end of 1991 (SC & TBO) and at mid-1992 (TBO & QQO). The (detrended) SC envelope reflects takeover by the QQO during the CH4 event, i.e. an apparent dynamic regime transition at the beginning of the 1990s. Figure 47.14d shows a further TBO decomposition that points to splitting toward the turn of the 1980s into a more strictly biennial mode and a quasi-triennial one (‘QTO’). The latter starts dominating with the event in question, whereas the former directly shapes the signal for almost one year via QQO phase synchrony. The combined result describes the obvious slowing down of climate dynamics in the back as a TBO/QTO–QTO/QQO transition.

CH4 load spectrograms of the leading 10 MP-FM modes and of the Wavelet Transform (WT) are shown in Fig. 47.15. The MP-FM modes are displayed en detail in Fig. 47.16 and Table 47.7. This CH4 load view does not directly confirm the SSA based (growth rate) findings, but a TBO may be borne in MP-FM modes #5, 6 and 8 altogether; its signature might thus emerge from drifting modes—which also cooperate in generating the 1991/1992 event. Indeed, modes #6 and 8 are crossing each other in-phase just at the height of the (load) pulse around the beginning of 1992 (Fig. 47.15, left panel), and mode #5 plus the basic SC mode #3 contribute to the apparent resonance that makes up the event (another such cooperative effect at the beginning of the record shown generates a similar but negative excursion).
Fig. 47.15

MP-FM (modes #1–10; left panel) and WT spectrograms (Morlet; right) of monthly global CH4 load; 07/1983–12/1998; frequency unit: cycles per month (cpm)

Table 47.7

MP-FM structure books of modes #1–10 of monthly global CH4 load and growth rate (07/1983–12/1998), centered analyses (transform time from monthly to yearly according to \(t_{y}:= (t_{m} + 6)/12 + 83\)); f, \(\tilde{f}\) in cycles per month (cpm)

k

α k

s k

u k

f k

ϕ k

\(\tilde{f}_{k}\)

β k

\(\tilde{\phi }_{k}^\prime\)

 

Global CH4 load (in ppbv)

 

1

516.06

512

39

0.0072

0.256

0.00051

9.55

2.749

 

2

80.37

1,024

37

0.0405

0.099

0.00121

29.38

2.749

 

3

49.88

1,024

110

0.0884

-1.937

0.00204

2.90

-2.945

 

4

31.23

2,048

92

0.1768

-2.476

0.00079

13.23

3.142

 

5

23.49

1,024

23

0.0221

-1.465

0.00253

3.54

3.142

 

6

18.17

256

59

0.0682

1.163

0.00329

16.04

-2.945

 

7

16.35

128

142

0.0356

0.389

0.00107

29.38

2.945

 

8

11.66

256

91

0.0549

-2.049

0.00686

3.69

1.374

 

9

10.05

64

23

0.2102

-0.967

0.03559

1.85

0.982

 

10

8.12

256

74

0.1693

1.601

0.00577

8.10

3.142

 

Global CH4 growth rate (in ppbv/a)

 

1

381.30

2,048

7

0.1768

-0.214

0.00086

12.13

2.749

 

2

305.31

512

86

0.0653

-0.708

0.00061

29.70

0.000

 

3

172.36

64

24

0.2726

1.859

0.03263

3.06

2.356

 

4

141.15

512

19

0.2500

1.064

0.01205

0.83

0.785

 

5

87.79

256

97

0.3242

-0.470

0.00391

2.56

-0.785

 

6

82.06

256

96

0.1928

2.760

0.02026

4.39

2.945

 

7

71.40

256

135

0.2293

1.477

0.01154

15.36

0.785

 

8

65.85

4

3

0.2102

0.060

0.21953

0.84

0.982

 

9

59.62

128

61

0.2847

-1.615

0.00716

19.54

2.945

 

10

57.26

256

87

0.0846

2.092

0.01013

3.86

1.374

 
Fig. 47.16

Leading 10 MP-FM modes and their signal envelopes of the global monthly CH4 load (in ppbv); 07/1983–12/1998

As confirmed by the WT magnitude spectrogram (Fig. 47.15, right panel), for the 1991–1992 event there is no sign of a corresponding singularity in the signal energy. A TBO signature during the 1980s in the CH4 load is found in the frequency modulation of MP-FM mode #9 (Fig. 47.15, left panel). The WT phase spectrogram (Fig. 47.17, left panel) underlines the dynamic causation of the event: Showing only a slight drift from the shortest to the longer periods, there is a rare, deep phase coincidence across the whole frequency domain down to, and including, the biennial to quadriennial range. In the analogous growth rate phase spectrogram (Fig. 47.17, right panel), the coincidence appears a couple of months earlier, toward the end of 1991 (recall Fig. 47.14c), and extends even deeper into the longer-period range.
Fig. 47.17

WT based phase spectrograms of monthly global CH4 load (left panel) and growth rate (right), 07/1983–12/1998; frequency unit: cycles per month (cpm)

For more direct comparison with the SSA study (Carl, 1998), MP-FM and WT magnitude spectrograms of the CH4 growth rate are given in Fig. 47.18. Among MP-FM modes #1–10 (not individually displayed) there is no slower time domain motion; the spectrogram is dominated by the seasonal, semiannual and higher SC-harmonic modes. This means that the adaptive capacity and precision of deseasonalization may be crucial to correct detection of the TBO and slower modes. The WT spectrogram (right panel) exhibits a weak, more localized signal that culminates toward the second half of 1991 and extends from the seasonal cycle mode toward the slower quasibiennial range—where it arises in early 1992. However, time localization of longer-period signals is uncertain in the WT which bears unfavourable ‘uncertainty’ properties in the TF plane (cf. Appendix). Note that the resonance effect near the left-hand boundary is represented in the MP-FM spectrogram (left panel) as an isolated, localized signal—but it is the vicinity to the boundary of the data period here that makes its representation uncertain as well.
Fig. 47.18

MP-FM (#1–10; left panel) and WT spectrograms (Morlet; right) of monthly global CH4 growth rate; 07/1983–12/1998; frequency unit: cycles per month (cpm)

Anyway, the MP-FM spectrogram bears a real surprise, namely frequency modulations which resemble those found in the gas production time series (Fig. 47.12, left panel). The question thus arises whether these MP-FM and SSA based growth rate approximations might not provide hints indeed at the nature of anthropogenic impacts on climate dynamics. A preliminary answer is tried in Sect. 47.6. Beforehand, a glance has to be shed on potential effects of crisis and war in the Gulf on the global methane load and growth rate.

47.5.2 CH4 Scenario of the Gulf Crisis and War

Figure 47.19 provides a simple “Middle East CH4 scenario” (Carl, 1998) that roughly takes the success of fire extinction and control over wild wells into account. No doubt disputable, it may illustrate a potential gross effect. Oil translates into atmospheric CH4 via its associated gas, which contains about 80 % CH4 (90 % CH4 are a good estimate for natural ‘dry’ gas). The “gas to oil ratio” (GOR) of a reservoir is available in some cases. Kuwait’s Burgan field has a GOR of 464, i.e. 464 cf gas are associated with each barrel of oil (Husain, 1994a). With a specific weight of 0.7157 kg/m3 for CH4 and a conversion factor 0.3614 ppbv/Tg for its global mole fraction (cited in Dlugokencky et al. (1998)), complete loss—as assumed here for gushing wells—would rise the globally averaged CH4 mixing ratio by ∼ 2.719 pptv per Mb of Burgan oil (pptv—parts per trillion by volume). Five percentage escape at the wellheads may be assumed for those ablaze, adding to further two percentage that may have survived incomplete combustion.
Fig. 47.19

Left panel: Middle East CH4 growth rate scenario (white; dotted line), fed by burning (black; long-dashed) and gushing oil wells (black; full line) as well as SA replacement production (grey; dash-dotted); right panel: MP-FM modes #3 of world gas production (full line) and #2 of the global CH4 load (dash-dotted) & growth rate scenario of the left panel.

For the Saudi Arabian crisis production a doubled venting/flaring rate of associated gas is taken, even though the normal rate is already high in the Middle East ( ∼ 30 %). Assuming 20 % unburnt escape, i.e. 6 % of the gas associated with that excess production, the instantaneous CH4 growth rate scenario completes as shown in the left panel of Fig. 47.19. GOR 500 was taken for Kuwaiti and GOR 400 for Saudi Arabian oil, the replacement production was adopted from Fig. 47.8 as indicated there, and peak values of 0. 5 (5. 0) Mbpd for gushing (burning) oil were blended with a monthly [Feb.–May] filter of [0.5/1/1/0.5] for gushing wells, and a [Feb.–Oct.] filter of [0.5/1/1/0.9/0.8/0.7/0.5/0.3/0.1] for burning ones.

Though one should not overstrain imperfect coincidences, the right panel of Fig. 47.19 shows a normalized view of two conspicuously similar MP-FM modes, namely #3 of worldwide gas production (Fig. 47.11) and #2 of the global CH4 load (Fig. 47.16), together with the growth rate scenario of the left panel. To say the least, it does not contradict the hypothesis of a discernible impact on the global CH4 load from the ME occurrences. At the apparent end of the emission anomaly, early in 1993, CH4 lags behind gas production by ∼ 9.5 months, as compared with ∼ 2 months at its beginning.

47.5.3 Carbon Dioxide in Perspective

Turning to CO2, Fig. 47.20 shows MP-FM spectrograms of the global mixing ratio (left panel; cf. also Fig. 47.21 and Table 47.8) and its growth rate (right), again of the leading 10 modes. Whereas a technical boundary effect might partially be blamed for the leftmost slow-mode patch in the mixing ratio (mode #4 in Fig. 47.21), the other patches (modes #2, 6, 7) are traced back in the growth rate figure (right panel) to a change between modulated modes, carrying the MP-FM view on a regime transition just at the beginning of the 1990s—as also seen in the iterative SSA study (Fig. 47.22) (Carl, 1998).
Fig. 47.20

MP-FM spectrograms of CO2 mixing ratio (in ppmv; left panel) and growth rate (in ppmv/a; right); 01/1979–12/1998; frequency unit: cycles per month (cpm)

Fig. 47.21

Leading 10 MP-FM modes and their signal envelopes of the global monthly CO2 load (in ppmv); 01/1979–12/1998

Fig. 47.22

QQO/QTO and TBO growth rate variability of CO2 (in 0.2 ppmv/a; dash-dotted) and CH4 (in ppbv/a; full line) as obtained via iterative SSA (Carl, 1998)

WT of the CO2 growth rate (Carl, 1998) shows branching at the turn of the 1980th of the low-frequency range into 3–4 and 5–6 year variability. Enforced frequency localization (by iterative SSA) generates the drifting and weakening QQ/QT mode of Fig. 47.22 to which the QQO in CH4 temporarily synchronizes. A TBO in the CO2 growth rate, roughly in quadrature with its CH4 counterpart, changes phase during 1988/1989 to take the lead from the latter. The QTO in the CO2 record has been attributed to Eastern Pacific sea surface temperatures, the QQO to the terrestrial biosphere (Dettinger and Ghil, 1998).
Table 47.8

MP-FM structure books of modes #1–10 of monthly global CO2 load and growth rate (01/1979–12/1998), centered analyses (transform time from monthly to yearly according to \(t_{y}:= t_{m}/12 + 79\)); f, \(\tilde{f}\) in cycles per month (cpm)

k

α k

s k

u k

f k

ϕ k

\(\tilde{f}_{k}\)

β k

\(\tilde{\phi }_{k}^\prime\)

 

global CO2 load (in ppmv)

 

1

128.72

2,048

21

0.0021

-2.782

0.00049

0.33

-0.393

 

2

19.78

128

202

0.0405

3.023

0.00121

29.38

-2.749

 

3

19.81

512

122

0.1051

-0.613

0.00082

27.30

3.142

 

4

9.52

16

4

0.0058

2.689

0.00000

0.00

0.000

 

5

7.75

512

155

0.1768

2.592

0.00063

16.43

-2.945

 

6

6.99

32

232

0.0884

0.671

0.00465

14.71

-2.553

 

7

3.28

64

89

0.0964

2.461

0.00288

29.38

3.142

 

8

2.68

512

111

0.0776

3.078

0.01154

0.86

2.160

 

9

1.71

512

110

0.0372

-0.103

0.00151

19.07

-2.553

 

10

1.56

256

81

0.0923

1.602

0.00465

13.50

-2.945

 

global CO2 growth rate (in ppmv/a)

 

1

131.34

1,024

123

0.0923

1.539

0.00063

14.36

3.142

 

2

94.69

512

120

0.1768

2.252

0.00066

15.73

3.142

 

3

28.39

1,024

15

0.2500

2.834

0.03881

0.26

-2.945

 

4

13.00

64

75

0.1928

2.512

0.01632

7.06

0.785

 

5

10.98

64

214

0.2726

-2.877

0.00929

18.30

3.142

 

6

10.83

256

127

0.1621

1.791

0.00552

3.74

-2.356

 

7

8.73

512

115

0.3105

1.422

0.00288

27.74

-0.393

 

8

8.19

512

135

0.0884

-1.912

0.00301

17.53

1.767

 

9

7.72

128

107

0.2611

0.624

0.03559

2.81

-1.767

 

10

7.54

2,048

226

0.1552

0.675

0.00748

16.04

-0.785

 

However, the discernible QQO–QTO transition in CO2 since the 1980s (Fig. 47.22) parallels the QTO emergence in the CH4 growth rate after the 1991–1992 event, flanked by the weaker biennial lobe (Fig. 47.14d). These CH4 evolutions should relate to the transition found in Dlugokencky et al. (1994c) from counter- to co-evolution between the hemispheric growth rates. Both the CO2 and CH4 records appear to hint at a change in the dynamic interaction between boreal and austral monsoons, in concert with a jump at the turn of the 1980th (signalled by the strong 1988/1989 La Niña event) in the interplay between ocean–atmosphere and atmosphere–land dynamics (Carl, 1998). Although of great interest in the present context, in-depth scrutiny of the climatic backscene would lead away from the path and scope of this paper (cf. Sect. 47.6.1).

47.6 Synthesis and Conclusions

It is an established method in physics to draw conclusions on a dynamic system from its response to perturbations. Tailored experiments may help identifying the system’s dynamical status and mechanisms that govern its motion in the vicinity of the disturbed state. As cynical as it might sound, since such an ‘experiment’ with the climate system has been suffered by all involved parties, including the worldwide public (although under substantial protest), drawing climatic insights from the drama is the least to be done after facing the failure of efforts to prevent it. A concluding endeavour is indicated to synthesize the results obtained, with a focus on lessons administered by, and learned from, the 1990/1991 Gulf crisis and war and its disastrous results. Apart from methodological issues, which are addressed in the Appendix, these lessons and conclusions belong to three major areas at least:
  1. (i)

    Background climate dynamics of the occurrences,

     
  2. (ii)

    Man’s economic activity and its climatic signature,

     
  3. (iii)

    Policy implications with respect to anthropogenic climate change.

     

More general political aspects (e.g., His Majesty King Hussein 1990; Deutscher Bundestag 1991), economic, legal and security lessons (Seacor, 1994, e.g.), as well as the apparent ‘failure of science’ in countenance of catastrophe–bearing developments of this type, are entangled with the environmental theme but were largely omitted from this study.

47.6.1 Climate Dynamics: The Backscene

The most important physical result of the research presented concerns the low-dimensional dynamic organization found at this occasion of the GCM’s boreal summer monsoon system (Carl, 1991b,c19921994; Tschentscher et al., 1994; Carl et al., 1995; Carl, 2013a). From these archetypal dynamics, a broad range of conclusions may be drawn on those of the real system (Carl, 2013c). Considering the seasonal cycle (SC) of the climate system as a dynamic object of limit cycle type, driven by the annual external forcing, the system undergoes topological change(s) in summer which temporarily inflate this object into a torus segment, where the minor circumference is made up of the intraseasonal 40–60 days oscillation that carries the major monsoon activity cycle. The sensitive model response to smoke load scenarios, based on an undisturbed GCM climate that shows essential features of observed monsoon dynamics, is to be taken seriously. Simplified physical conditions under which the model has been run, including prescribed climatological lower boundary conditions, fixed insolation (solar constant), and invariable, globally averaged trace gas loads (taken as CO2 equivalent), call for advanced GCM studies. Nevertheless, the conceptual progress achieved allows conclusions about the climatic response at issue.

Monsoon retreat in autumn from the toroidal phase space object, i.e. re-adjustment of intraseasonal dynamics to the annual pace, is such a key aspect. It may proceed along three pathways, in essence: (i) from dormant monsoon, running smoothly into the externally forced SC; (ii) from active monsoon, bearing an excess load of atmospheric mass over the northern hemisphere (NH) and running into an “Indian Summer” type autumn climate, with anomalous autumn trajectories in both tropics and NH extratropics; (iii) from monsoon revival, leaving a largely symmetric forcing—with circulation anomalies in autumn confined to the tropics/subtropics—that may trigger a latent warm event (El Niño) in the tropical Pacific.

Real world analogues to cases (ii) and (iii) may have set the stage in autumn 1990 and 1991 for the CH4 anomaly. Though the Southern Oscillation (SO), i.e. the atmospheric component of the El Niño–Southern Oscillation (ENSO) system, was prepared for El Niño to occur, a decidedly ‘uncooperative’ monsoon withdrew from its active phase into a pronounced Indian Summer type autumn in 1990, with northwards displaced tropical convection and tropospheric zonal jets, and a strong Pacific blocking, all enduring until November (Mo, 1993). This type (ii) retreat met the actual state of the TBO thus enhancing its asymmetric SC anomaly that inhibited evolution of the (symmetric, largely equatorial) SO into El Niño. The frustrated SO, in turn, retarded the normal mining of the excessive NH mass load left in autumn 1990. Consistently, the winter monsoon developed below normal in NH precipitation, poleward advance over Australia and intraseasonal activity.

In contrast to 1990, a monsoon retreat of type (iii) matches the 1991 NH autumn patterns. A band of enhanced convection stretched from Southeast Asia into the Central Pacific (Janowiak, 1993) and a related lower troposphere westerly wind event (Philips, 1995) fanned the evolving El Niño for weeks. As outlined in Carl (1998), an enhanced TBO, a type (iii) monsoon retreat and a typical western/central Pacific dipole structure evolved, just as in the prelude to the strong 1982/1983 El Niño (the moderate 1986/1987 event was preceded by a type (i) retreat). The climatic background thus favoured an enhanced TBO signal as reflected in the CH4 record (Carl, 1998). The methane burst from the Middle East should have strengthened the upward swing in the growth rate, moreover, whereas the cooling effect attributed to Mt. Pinatubo’s eruption may be blamed in part for the strength of the subsequent downward swing (cf. Fig. 47.14a). The CH4 growth rate scenario of Fig. 47.19 perfectly matches the rising 1991 TBO/QTO flank of Fig. 47.14d and makes up about two third of this biennial mode flank (or one fourth of the steep QTO rise).

Figure 47.23 (Carl, 2013b) shows another climatic evolution of interest in the present context: A phase plot of annual hemispheric surface air temperatures (SAT; 1870–1997, left panel) exhibits thermal stagnation of the global climate system from the turn of the 1980s until 1997, the end of the record analysed there. It was accompanied by a longer-term decline of insolation (at top of the atmosphere, i.e. uninfluenced by the volcano eruption; right panel). This turning point in the climate system’s long-term evolution just around the beginning of the 1990s renders it even more difficult to separate the anthropogenic trace gas signals. Moreover, sunspot cycle #22 started terminating, with a steep flank, just early in 1992 when the CH4 growth rate showed its maximum decline—ironically coinciding with the terminating tail of the growth rate scenario of Fig. 47.19 (left panel). Note also in passing that the 1982/1983 El Niño followed a large volcano eruption (El Chichon), as with Mt. Pinatubo that preceded the 1992/1993 Pacific warm event.
Fig. 47.23

MP-FM slow mode perspective on global warming (1870–1997); left: NH vs. SH SAT (in K); right: global SAT (in K) vs. insolation (SRAD; in W/m2), anomaly analyses in annual resolution; straight lines: linear regressions (Source of figure: Carl (2013b))

Obviously, in this complex setting qualitative reasoning does not much help to disentangle man-made from natural effects in the global trace gas budget. As one of the opportunities, a set of detailed and specifically initialized GCM experiments would be desirable, using an advanced trace gas scheme. To substantiate this idea, Fig. 47.24 (left panel) shows the GCM’s typical monsoon retreat variability which bears various interannual scales, not to the least the biennial one (Carl 2013a,c). A new set of experiments might be started from carefully chosen GCM states which approximate the 1990/1991 situation—to replace the annual mean state chosen in the early experiments (Sect. 47.3). These may be taken from a longer control run as displayed in Fig. 47.24 (right panel; also comprising the example of the left panel).
Fig. 47.24

Left panel: Monsoon retreat behaviour of the GCM, displayed as NH tropospheric mass anomaly (in 1016 kg), showing the three different types of retreat of which the “Indian summer” type (IS) exerts asymmetric pressure gradient forcing; right panel: 100 year segment of a control run showing substantial interannual variability at various scales (365 days average) (Sources of figure: Carl (2013a,c))

47.6.2 Signatures of Man’s Economic Activity

Quantitative comparison of modal structures is a second promising approach. It may be used to estimate sensitivities and transfer rates between anthropogenic emissions and atmospheric budgets of CH4, and should be supported by time series analyses using scenario estimates like that of Fig. 47.19. A preliminary example of such an assessment of direct and indirect ‘translations’ between emissions and atmospheric methane growth rates is provided in Carl (1998) for man’s economic activity: Stalling oil and gas production may have slowed down the instantaneous CH4 growth rate during 1991–1993 by ∼ 2.5 ppbv/a2, of which the direct contribution was certainly about one third. Counterbalanced by climate effects and Middle East emissions, these evolutions could not prevail in the global CH4 record before 1992, however.

The estimate goes as follows: Adopting a mean venting/flaring loss of 5.7 % and a consumption loss of 2.5 %, worldwide gas production would change the global CH4 growth rate by ∼ 0.1125 ppbv/a2 per bcf/d production change. Comparing the seasonal cycles (SCs) in both gas production and the CH4 growth rate yields ∼ 0.3 ppbv/a per bcf/d. The annual rise in production before the 1990–1994 stagnation amounted to ∼ 7.75 bcf/d for gas and ∼ 1.785 Mbpd for oil (Fig. 47.9). Assuming GOR 300, the growth rate would thus have directly increased by ∼ 0.932 ppbv/a2, gas accounting for ∼ 90 %. Indirect estimation using QTO cycles confirms the ‘climatic amplification’ found in the SC of the CH4 growth rate, resulting in a ∼ 2.5 ppbv/a2 decline due to stagnation of worldwide oil and gas production. The anthropogenic effect may thus be roughly confined between the two limiting figures given. The uncertainty range can certainly be narrowed by modal structure analyses of accordingly modified time series. As with the Middle East CH4 scenario, it goes without saying that this calls for series of carefully tailored data experiments, however, which have to be left for another study.

A range of further tasks and opportunities can be identified to reduce uncertainties in the estimation of man-made impacts on climate dynamics at the turn of the 1980s and the early 1990s. These include (i) a time series based comprehensive climate study in monthly resolution (to fit the oil and gas production data), in analogy to the annually resolved study (Carl, 2013b) aimed to uncover the active modal structure network of the period and the status of internal synchronization at this timescale; (ii) incorporation of river runoff into such a climate study, as a result of natural spatial aggregations within the terrestrial hydrological cycle; (iii) in-depth and broad scrutiny of measurements at the oil fields concerning their external conditions, objectives, methods etc., with a view on sound CH4 scenario development and related data experimentation; (iv) detailed study of seasonality effects ((sub-)structure and timing of the SC); (v) methodological advancement, notably extension of the MP-FM dictionary by an analysing waveform that directly captures impulse–response situations in order to better grasp perturbations.

47.6.3 Minutes on Climate Policy in a Nonlinear World

A fundamental lesson of relevance to climate policy reads: The world climate has sensitive spots where man’s activity may influence the system’s pathway more effectively than elsewhere; it does therefore matter, where and when emissions take place. As a political conclusion, the rules of emission trade (as ineffective as they might work at the time being) must become adjusted to this knowledge, which in turn has to be further substantiated and broadened. Only worldwide oil and gas productions have been considered here. The same analysis should be conducted for each major producer, and data bases on other anthropogenic emissions should enter a new round of investigations.

At the other side of this medal an opportunity emerges which also might enter negotiations on the mitigation of climate change: Having identified ‘hot spots’ of anthropogenic impacts on climate dynamics, notably the Middle East with its exposed location in relation to the boreal summer monsoon system, technical and technological measures to reduce emissions may focus on the regional low-pressure systems of production and transport. Specifically targeted investments may easier come into effect than those required to solve the global emission problem. They are an alternative and supplement, at least, and may help gaining the time needed to reach the latter goal. It would be a matter of negotiation to arrive at a fair cost sharing scheme.

Another fundamental but more disputable theme is the way how to deal with risks. It reflects a strange self-consciousness of politics (and of interested circles in the back) if pressure is exerted to withheld data, for example (Seacor, 1994, e.g.)—and likewise of representatives of the scientific community, if risks are either minimized or exaggerated (where there are simply gaps of knowledge). The Scientific Committee on Problems of the Environment (SCOPE) of the International Council of Scientific Unions (ICSU) launched a remarkable initiative in the early 1980s when the environmental risks of nuclear war had been discovered: A completely open, worldwide process of research to which every serious contribution was welcome. It resulted in an authoritative statement on the problem by the world scientific community (Pittock et al. 1986; Harwell et al. 1985) which reflected the state of knowledge and its uncertainties but did not refrain from drawing conclusions—a fateful hour of science en face of torrential global risks (even though it came after a bidecadal period of silence (Carl et al., 2008, e.g.)).

47.7 Appendix: Methods and Data

Three general remarks are in order before briefly drawing attention to the methods used in this study: (i) A methodical challenge is borne in the relatively small contributions that may be assigned to slow CH4 growth rate processes; that is, the ‘art of de-seasonalization’ appears to be crucial with a view on the search for causal effects; (ii) in following the idea of sparse approximation, which the MP-FM approach stands for here, one implicitly addresses the dimensionality of the problem at hand; minimization of the \({\ell}^{0}\) norm, i.e. of the number of modes that may suffice to approximate a time series with pregiven accuracy, thus touches a fundamental issue of the analysis of complex systems; (iii) ‘post mortem’ data analyses as usually conducted (also here) should be supplemented by impulse-response analyses to address the instantaneous, causal response of the system; as a compromise, the MP-FM dictionary could be supplemented in a future study by a corresponding analysing wavelet.

Three methods of time series analysis have been used in the present study: Singular–system Analysis (SSA; originally termed “Singular Spectrum Analysis” (Vautard et al., 1992)), the Wavelet Transform (WT (Torrence and Compo, 1998)) and the Matching Pursuit technique (MP (Mallat and Zhang, 1993)). Just like measuring instruments in experimental and field research, these methods have their own, individual features each and provide different views of the study object. Speaking in terms of linear operator theory, they transform the data according to their eigen- (or more generally, singular) systems which consist of fundamental modes and their respective contributions to the data object as obtained by means of projection. The very nature of this task is inversion of a signal transduction system
$$\displaystyle{ \chi (t) = \mathcal{T}\{\breve{\chi }\}(t) \equiv \int _{T}\kappa (t,t^\prime)\breve{\chi }(t^\prime)\ dt^\prime\, }$$
(47.3)
where χ(t) is the time series as obtained by a measuring operation \(\mathcal{T}\) (written as integral operator with a certain kernel κ) from the ‘true’ data object \(\breve{\chi }(t)\). T is the time support, i.e. the length of the time series. A concise yet fairly complete background discussion may be found in Carl and Behrendt (2008).

47.7.1 Singular–System Analysis (SSA)

SSA belongs to the family of Principal Component Analyses and provides thus a statistical decomposition of the data χ(t) (Carl and Behrendt, 2008, e.g.). It emulates the delay coordinate approach to dynamic systems analysis from time series and has been invented just to extract dynamic modes from short, noisy data (Vautard and Ghil, 1989). The expansion exploits dependence in second–order statistics and is complete and optimal in the sense of captured variance. The SSA eigenvalue problem is an integral equation of convolution type,
$$\displaystyle{ {\mathcal{T}}^{SSA}\{\phi _{ i}\}(t):= \frac{1} {\theta } \int _{\theta }C_{\chi }(t-\tau )\phi _{i}(\tau )\ d\tau =\lambda _{i}\phi _{i}(t)\, }$$
(47.4)
where κ in (47.3) is replaced with the normalized autocovariance function of χ at lag τ, \(C_{\chi }(t-\tau )/\theta\). Each λ i has the physical meaning of a variance; θ is a time window. SSA decomposition; that is, the attempt to reconstruct individual modes that reflect the diverse impacts which might have generated the data object \(\breve{\chi }\), proceeds via \(\chi (t,\tau ) =\sum _{ i=1}^{N}\chi _{i}(t+\tau ) =\sum _{ i=1}^{N}\xi _{i}(t)\phi _{i}(\tau )\), where the projection coefficients (“principal components”) ξ i become time-dependent due to the windowed approach. SSA modes are finally reconstructed to full time series length via \(\chi _{i}(t) = \frac{1} {\vartheta } \int _{\vartheta }\xi _{i}(t-\tau )\phi _{i}(\tau )\ d\tau\) (\(\vartheta =\theta\), except at the boundaries). Certain freedom in the choice of N (or θ) relates to embedding.

A problem with the application of SSA in the analysis of complex systems results from its variance criterion which tends to mix dynamically discernible modes. To overcome this shortage, the SSA modal structure has been re-sorted into two components of which one carries the ‘mode’ of interest (identified by its Fourier peak), and the other one the residue. Both are treated by SSA again and re-sorted until the Fourier Transform (FT) provides a sufficiently clean mode. This mode is extracted from the time series then, and the procedure starts anew in order to extract the next mode, and so forth. This stepwise (“greedy”) decomposition is common practice today and has a predecessor in the ‘pre-whitening’ approach—which in general focused on a ‘clean’ residual statistic, though.

Proceeding this way, it turned out that certain Fourier peaks could not be isolated as pure modes, leading to the conclusion that frequency modulation (FM) must play a role in the data at hand. One of the FT traps is just its misleading FM representation which may pretend a signal energy distribution that has nothing to do with reality (generation of spurious spectral peaks or suppression of real ones, e.g.). Note also that the recommended Toeplitz structure of the covariance matrix C bears the presumption of stationarity and is not useful here since it produces, after lengthy iteration, nothing but the result of the FT (after all, it confirmed correct coding at least). A lagged form of C has therefore been used in the iterative SSA study of Carl (1998).

47.7.2 Wavelet Transform (WT)

As a ‘fractional’ time–to–scale transform, the WT yields a 2D result. The time–scale plane is subdivided to this end into tiles of equal area, a choice that bears an ‘uncertainty relationship’ of signal processing. In the larger scale–range the tiles cover longer time segments and thus have a bad temporal (but a good scale) resolution. For shorter scales, in contrast, the time resolution is higher at the expense of a worse scale resolution. Singularities, which settle in the shortest scale–range, are precisely localized in time by the WT.

Unlike the SSA, both FT and WT have recourse to a data model. Whereas the FT maps a time series χ(t) onto its basis of infinitely extended harmonic signals (\(\cos (\omega t)\), \(\sin (\omega t)\)), the WT provides analysing signals of finite scale (“wavelets”), which all result from a “mother” wavelet via compression or stretching and translation along the time axis. A natural extension of the harmonic FT basis is the “Morlet”, a harmonic with Gaussian envelope, \(\psi _{M}(t) = g(t)\ exp(i\omega t)\). Its “daughter” wavelets formally equal the “Gabor atom” of signal processing (Gabor, 1946) (“Gabor wavelet”):
$$\displaystyle{ \psi _{\gamma }(t) = \frac{1} {\sqrt{s}}\ g\left (\frac{t - u} {s} \right )\ {e}^{i\omega t}. }$$
(47.5)
Signal space index γ = (s, u, ω) comprises the signal attributes scale s, translation u (i.e. position of maximum signal energy), and frequency ω = 2π f, the Gaussian time window normalizes the signal energy of these basis functions: \(g(\tau ) = \root{4}\of{2}\exp ({-\pi \tau }^{2})\).
Like the FT, the WT is defined as inner product between χ(t) and a basis function ψ(t) (e iωt with the FT), written here as daughter wavelet:
$$\displaystyle{ \mathcal{T}_{\psi }^{WT}\{\chi (t)\}(u,s) = \frac{1} {\sqrt{s}}\ \int _{-\infty }^{\infty }\chi (t)\ \bar{\psi }\left (\frac{t - u} {s} \right )dt }$$
(47.6)
(\(\bar{\psi }\) is the complex conjugate of ψ). This yields for the Morlet ψ M , in complete analogy to the FT:
$$\displaystyle{ \mathcal{T}_{\psi _{M}}^{WT}\{\chi (t)\}(u,s) = \frac{1} {\sqrt{s}}\ \int \chi (t)\ g\left (\frac{t - u} {s} \right ){e}^{-i\omega t}dt. }$$
(47.7)
As problem specific as various wavelets might be defined, they have a common restriction (aside of other ‘admissibility conditions’) in their fixed relationship between scale and frequency—which is the basis, however, of any WT application as a TF method. It removes the argument ω in (47.6) which becomes redundant also in γ with the transition \(\psi _{\gamma } \Rightarrow \psi _{M}\). This makes the difference between Morlet and Gabor atom. Throughout the present paper, the Morlet has been taken as analysing wavelet.

47.7.3 Matching Pursuit (MP)

The MP technique combines a “greedy” procedure (as also used in the iterative SSA application) with the data model approach. To cope with frequency modulation (FM), a specific dictionary of analysing wavelets is used: An MP-FM “mode” as defined here is the projection of the actual best-fitting “Gaussian logon” (47.8) on the time series (or its respective residual after the initial step of extraction). This new elementary signal,
$$\displaystyle{ \psi _{r\tilde{\gamma }}(t) = \frac{1} {\sqrt{s}}\ g\left (\frac{t - u} {s} \right )\cos \left (\omega _{c}(t - u) +\phi ^\prime +\tilde{\varphi } (t - u)\right )\, }$$
(47.8)
spans a 7D signal space \(\tilde{\gamma }= (s,u,\omega _{c},\phi,\beta,\tilde{\omega },\tilde{\phi })\): \(\omega _{c} = 2\pi f_{c}\) is the angular carrier frequency, ϕ the phase constant of the carrier mode, \(\beta =\delta \omega /\tilde{\omega }\) the modulation index (δ ω is peak deviation from ω c ), \(\tilde{\omega }= 2\pi \tilde{f}\) the modulation angular frequency, and \(\tilde{\phi }\) the phase constant of harmonic phase modulation, \(\tilde{\varphi }(\tau ) =\beta \sin (\tilde{\omega }\tau +\tilde{\phi }^\prime)\). The explicit use of phase constants unties the signal’s phase(s) from its energy localization u (phase constants \(\phi ^\prime =\phi +\omega _{c}u\) and \(\tilde{\phi }^\prime =\tilde{\phi } +\tilde{\omega }u\) earmark this independence on location u). The ‘double harmonic Gabor atom’ thus raises all the restrictions this way that are imposed by construction to the WT—at the expense of a much higher demand for computational resources, notably runtime (due to the high signal space resolution required, MP-FM bears a generic massively parallel task).

Though the decomposition is linear, the Gaussian logon may match highly nonlinear waveforms for each individual mode. Convergence has been proved for long (Mallat and Zhang, 1993), i.e. summing up the modal structure approximates the time series to any desired accuracy. In contrast to the SSA or customary TF representations, like windowed FT or WT, the MP-FM “structure book” precisely describes the modal structure of a time series in terms of eight parameter values per mode. Besides the 7D signal space location, it contains a projection coefficient α k that quantifies the part of signal energy captured by mode k. Seven-dimensionality of the MP-FM signal notwithstanding, only amplitude (the Gaussian envelope) and phase (or its time-derivative, frequency) bear time-dependence in the Gaussian logon. The traces that all modes of a time series χ(t) describe in the signal space may thus be projected on the TF plane if the (Gaussian) amplitude modulation is displayed as intensity (isolines); the result is an MP-FM spectrogram of χ(t). MP-FM performance has been discussed in a broad study into synchronous motions hidden in customary climate data (Carl, 2013b), which the interested reader is referred to.

47.7.4 Data Sources

Monthly oil and gas production data have exclusively been taken from the Oil & Gas Journal (OGJ, 1982–1992) and its Data Books (OGJ Data book, 1984–1998). Details used for the scenarios of Kuwait fire GCM experiments have been found in diverse sources as cited, scattered over the literature. CH4 data of the Carbon Cycle Group (CCG) of the National Oceanic and Atmospheric Administration, Climate Monitoring and Diagnostics Laboratory (NOAA-CMDL), were initially used as available at June 12, 1998, in postscript format via ftp (update of August 12, 1997; Dlugokencky et al. 1998). They could partially be replaced later by original time series kindly provided by the authors (cf. Acknowledgements). The climate data used for illustration may be found in Jones (1994) (SAT; unfortunately provided in deseasonalized form) and Lean et al. (1995) (SRAD).

Notes

Acknowledgements

This report refers to studies conducted (from time to time) over more than two decades and has recourse to results obtained with the indirect help of many authors. Detailed descriptions of the methods used, namely SSA (Vautard and Ghil, 1989), WT (Torrence and Compo, 1998) and MP (Mallat and Zhang, 1993), were especially important for own code developments. Graphics tools XvGr (Turner, 1992) and GraDS (Doty, 1992) have extensively been used and are referred to with due respect and gratitude. Invaluable data sources as cited are gratefully acknowledged, but a generous support has to be mentioned with special thanks: To enable direct comparison and to avoid mistakes due to their reconstruction from published PostScript figures, Ed Dlugokencky once provided the time series of deseasonalized CH4 and CO2 growth rates as obtained by the Carbon Cycle Group of the NOAA Climate Monitoring and Diagnostics Laboratory.

References

  1. Aldhous P (1991) Oil-well climate catastrophe? Nature 349(6305):96ADSGoogle Scholar
  2. Aleksandrov VV, Gates WL (1981) The performance of a coarse-grid version of the OSU two-level atmospheric GCM. Report No. 24, Climatic Research Institute, Oregon State University, CorvallisGoogle Scholar
  3. Andrae MO, Browell EV, Garstang M, Gregory GL, Harriss RC, Hill GF, Jacob DJ, Pereira MG, Sachse GW, Setzer AW, Dias PLS, Talbot RW, Torres AL (1988) Biomass-burning emissions and associated haze over Amazonia. J Geophys Res D 93(2):1509–1527CrossRefGoogle Scholar
  4. Anonymous (1988) Greater Burgan field record, issued April 1988. Petroconsultants S.A., GenevaGoogle Scholar
  5. Anonymous (1991) Kuwait’s oil sector at a glance. Oil Gas J 89(4):21Google Scholar
  6. Bakan S, Chlond A, Cubasch U, Feichter J, Graf H, Grassl H, Hasselmann K, Kirchner I, Latif M, Roeckner E, Sausen R, Schlese U, Schriver D, Schult I, Schumann U, Sielmann F, Welke W (1991) Climate response to smoke from the burning oil wells in Kuwait. Nature 351(6325):367–371ADSCrossRefGoogle Scholar
  7. Bjerkness J (1969) Atmospheric teleconnections from the equatorial Pacific. Mon Weather Rev 97(3):163–172CrossRefADSGoogle Scholar
  8. Cachier H, Brèmond M-P, Buat-Mènard P (1989) Carbonaceous aerosols from different tropical biomass burning sources. Nature 340(6232):371–373ADSCrossRefGoogle Scholar
  9. Carl P (1988) Software engineering aspects of computational systems analysis in physics. In: Sydow A, Tzafestas SG, Vichnevetsky R (eds) Systems analysis and simulation 1988, vol II. Math Res 47:375–378Google Scholar
  10. Carl P (1991a) Dämmerung über dem Morgenland. Klimarisiken der Ölbrände am Golf (in German). Unpublished manuscriptGoogle Scholar
  11. Carl P (1991b) Notes on the climatic response in the aftermath of Gulf War II. Z Meteorol 41(6):476–480Google Scholar
  12. Carl P (1991c) Persistent localized lower troposphere smoke within the planetary monsoon system. Available from the author; submitted to Nature Nov. 21, Ms.-No. C11408Google Scholar
  13. Carl P (1992) Zur dynamischen Struktur des planetaren Monsuns (in German). Wiss. Z. Humboldt-Univ. Berlin. R Math Nat wiss 41(2):29–35MathSciNetGoogle Scholar
  14. Carl P (1994) Monsoon dynamics in a low-dimensional GCM. WCRP-84, WMO/TD-No. 619(II), WMO, Geneva, pp 773–780Google Scholar
  15. Carl P (1998) The 1991–1992 fluctuation in atmospheric methane – a tale on global climate dynamics and man’s interference. Manuscript submitted to Nature Sept. 12, Ms.-No. C09253 (1998). Report No. 8, Climate Dynamics Group, Berlin, GermanyGoogle Scholar
  16. Carl P (2011) MP based detection of synchronized motions across the instrumental climate record. In: Proceedings of the IEEE statistical signal processing workshop. IEEE, pp 557–560. 978-1-4577-0568-7/11Google Scholar
  17. Carl P (2013a) On the dynamical status of the climate system – I: a general circulation model en route to chaos. In: Stavrinides SG, Banerjee S, Caglar H, Ozer M (eds) Chaos and complex systems. Springer, Berlin/Heidelberg, pp 521–528. doi:10.1007/978-3-642-33914-1_73CrossRefGoogle Scholar
  18. Carl P (2013b) On the dynamical status of the climate system – II: synchronous motions galore across the records. In: Stavrinides SG, Banerjee S, Caglar H, Ozer M (eds) Chaos and complex systems. Springer, Berlin/Heidelberg, pp 529–539. doi:10.1007/978-3-642-33914-1_74CrossRefGoogle Scholar
  19. Carl P (2013c) A general circulation model en route to intraseasonal monsoon chaos. In: Banerjee S, Rondini L (eds) Applications of chaos and nonlinear dynamics in science and engineering, vol 3. Understanding complex systems, chap 3. Springer, Berlin/Heidelberg. doi:10.1007/978-3-642-34017-8_3 (in print)Google Scholar
  20. Carl P, Behrendt H (2008) Regularity based functional streamflow disaggregation. I. Comprehensive foundation. Water Resour Res 44(W02420). doi:10.1029/2004WR003724Google Scholar
  21. Carl P, Stenchikov GL (1988) Structural analysis of the climatic response to a nuclear war. In: Sydow A, Tzafestas SG, Vichnevetsky R (eds) Systems analysis and simulation 1988, vol II. Math Res 47:33–36Google Scholar
  22. Carl P, Worbs KD, Tschentscher I (1995) On a dynamic systems approach to atmospheric model intercomparison. WCRP-92, WMO/TD-No. 732. WMO, Geneva, pp 445–450Google Scholar
  23. Carl P, Eichler T, Kroschk S, Lönhardt H, Schimmel E, Tschentscher I, Worbs KD (1998) Konzeptionelle Studien zur Dynamik des Atmosphäre–Land–Systems (in German). Report No. 7, Climate Dynamics Group, Berlin, GermanyGoogle Scholar
  24. Carl P, Svirezhev Y, Stenchikov G (2008) Environmental and biospheric impacts of nuclear war. In: Jørgensen SE, Fath BD (eds) Global ecology, vol 2 of encyclopedia of ecology. Elsevier, Oxford, pp 1314–1321CrossRefGoogle Scholar
  25. Chen T-C, Tzeng R-Y, Yen M-C (1988) Development and life cycle of the Indian monsoon: effect of the 30–50 day oscillation. Mon Weather Rev 116(11):2183–2199ADSCrossRefGoogle Scholar
  26. Cox J (1991) Environmental consequences of a Gulf war. Manuscript prepared for Environmental Protection Bulletin. Institution of Chemical EngineersGoogle Scholar
  27. Crutzen P, Birks JW (1982) The atmosphere after a nuclear war: twilight at noon. AMBIO 11(2/3):114–125Google Scholar
  28. De US, Desai DS, Bias SCS (1992) Monsoon season (June–September 1991). Mausam 43(3):333–346Google Scholar
  29. Dettinger MD, Ghil M (1998) Seasonal and interannual variations of atmospheric CO2 and climate. Tellus B 50(1):1–24ADSCrossRefGoogle Scholar
  30. Deutscher Bundestag, Stenographischer Bericht, 2. Sitzung (14. Januar 1991) (in German; German Parliament, 12th Election Period, 2nd Session, January 14). Plenarprotokoll 12/2, BonnGoogle Scholar
  31. Dlugokencky EJ, Masarie KA, Lang PM, Tans PP (1998) Continuing decline in the growth rate of the atmospheric methane burden. Nature 393(6684):447–450ADSCrossRefGoogle Scholar
  32. Dlugokencky EJ, Masarie KA, Lang PM, Tans PP, Steele LP, Nisbet EG (1994a) A dramatic decrease in the growth rate of atmospheric methane in the northern hemisphere during 1992. Geophys Res Lett 21(1):45–48ADSCrossRefGoogle Scholar
  33. Dlugokencky EJ, Masarie KA, Lang PM, Tans PP, Steele LP, Nisbet EG (1994b) Reply to “Comments on ‘A dramatic decrease in the growth rate of atmospheric methane in the northern hemisphere during 1992”’. Geophys Res Lett 21(22):2447–2448ADSCrossRefGoogle Scholar
  34. Dlugokencky EJ, Steele LP, Lang PM, Masarie KA (1994c) The growth rate and distribution of atmospheric methane. J Geophys Res D 99(8):17,021–17,043CrossRefGoogle Scholar
  35. Doty BE (1992) Using the Grid Analysis and Display System (GrADS). Manual, Version 1.3.1., University of MarylandGoogle Scholar
  36. Ferrare RA, Fraser RS, Kaufman YJ (1990) Satellite remote sensing of large scale air pollution—Measurement of forest fire smoke. J Geophys Res D 95(7):9911–9925CrossRefGoogle Scholar
  37. Findlater J (1969) A major low-level air current near the Indian Ocean during the northern summer. Quart J R Meteorol Soc 95(404):362–380CrossRefADSGoogle Scholar
  38. Gabor D (1946) Theory of communication, part III. J IEE 93:429–457Google Scholar
  39. Garcia RR, Salby ML (1987) Transient response to localized episodic heating in the tropics, II: Far-field behavior. J Atmos Sci 44(2):499–530ADSCrossRefGoogle Scholar
  40. Gates WL, Batten ES, Kahle AB, Nelson AB (1971) A documentation of the Mintz–Arakawa two-level atmospheric General Circulation Model. R-877-ARPA. Rand Corporation, Santa MonicaGoogle Scholar
  41. Giorgi F, Chameides WL (1986) Rainout lifetimes of highly soluble aerosols and gases as inferred from simulations of a General Circulation Model. J Geophys Res D 91(13):14,367–14,376CrossRefGoogle Scholar
  42. Golitsyn GS, Phillips NA (1986) Possible climatic consequences of a major nuclear war. Report of the world climate programme (WCP–113), WMO/TD No. 99, Geneva, p 21Google Scholar
  43. Harwell MA, Hutchinson TC, Cropper WP, Jr. Harwell CC, Grover HD (1985) SCOPE 28: environmental consequences of nuclear war. Vol II. Ecological and agricultural effects. Wiley, Chichester, p 523. (2nd edn with an updating preface (1989))Google Scholar
  44. His Majesty King Hussein Bin Talal of the Royal Hashemite Kingdom of Jordan (1990) Address to the second world climate conference, Nov. 6 1990. In: Jäger J, Ferguson HL (eds) Climate change: science, impacts and policy. In: Proceedings of the second world climate conference, Cambridge, pp 511–513Google Scholar
  45. Hogan KB, Harriss RC (1994) Comment on ‘A dramatic decrease in the growth rate of atmospheric methane in the northern hemisphere during 1992’ by E. J. Dlugokencky et al. Geophys Res Lett 21(22):2445–2446ADSCrossRefGoogle Scholar
  46. Hulstrom RL, Stoffel TL (1990) Some effect of the Yellowstone fire smoke cloud on incident solar irradiance. J Climate 3(12):1485–1490ADSCrossRefGoogle Scholar
  47. Husain T (1994a) Extinguishing of Kuwait oil fires—challenges, technology, and success. Atmos Environ 28(13):2139–2147ADSCrossRefGoogle Scholar
  48. Husain T (1994b) Kuwait oil fires—modeling revisited. Atmos Environ 28(13):2211–2226ADSCrossRefGoogle Scholar
  49. Janowiak JE (1993) The global climate for September–November 1991: Warm (ENSO) episode conditions strengthen. J Climate 6(8):1616–1638ADSCrossRefGoogle Scholar
  50. Jones PD (1994) Hemispheric surface air temperature variations: a reanalysis and an update to 1993. J Climate 7(11):1794–1802ADSCrossRefGoogle Scholar
  51. Krishnamurti TN, Bhalme HN (1976) Oscillations of a monsoon system. Part I. Observational aspects. J Atmos Sci 33(10):1937–1954Google Scholar
  52. Krishnamurti TN, Subrahmanyam D (1982) The 30–50 day mode at 850 mb during MONEX. J Atmos Sci 39(9):2088–2095ADSCrossRefGoogle Scholar
  53. Krishnamurti TN, Jayakumar PK, Sheng J, Surgi N, Kumar A (1985) Divergent circulations on the 30 to 50 day time scale. J Atmos Sci 42(4):364–375ADSCrossRefGoogle Scholar
  54. Krishnamurti TN, Sinha NC, Krishnamurti R, Osterhof D, Comeaux J (1992) Angular momentum, LOD and monsoonal LF mode. J Meteorol Soc Jpn 70(1):131–165Google Scholar
  55. Lau KM, Lim HJ (1984) On the dynamics of equatorial forcing of climate teleconnections. J Atmos Sci 41(2):161–176ADSCrossRefGoogle Scholar
  56. Lau KM, Yang GJ, Shen SH (1988) Seasonal and intraseasonal climatology of summer monsoon rainfall over East Asia. Mon Weather Rev 116(1):18–37ADSCrossRefGoogle Scholar
  57. Lean J, Beer J, Bradley R (1995) Reconstruction of solar irradiance since 1610: implications for climate change. Geophys Res Lett 22(23):3195–3198ADSCrossRefGoogle Scholar
  58. Leathers DJ, Yarnal B, Palecki MA (1991) The Pacific/North American teleconnection pattern and United States climate. Part I: regional temperature and precipitation associations. J Climate 4(5):517–528Google Scholar
  59. Lelieveld J, Crutzen PJ, Dentener FJ (1998) Changing concentration, lifetime and climate forcing of atmospheric methane. Tellus B 50(2):128–150ADSCrossRefGoogle Scholar
  60. Lorenz EN (1982) Low-order models of atmospheric circulation. J Meteorol Soc Jpn 60(1):255–267Google Scholar
  61. Lowe DC, Manning MR, Brailsford, GW, Bromley AM (1997) The 1991–1992 atmospheric methane anomaly: southern hemisphere 13C decrease and growth rate fluctuations. Geophys Res Lett 24(8):857–860ADSCrossRefGoogle Scholar
  62. MacCracken MC, Penner JE (1987) Under-examined aspects of the potential environmental effects of nuclear war. UCID-21111, Lawrence Livermore National LaboratoryGoogle Scholar
  63. Mallat SG, Zhang Z (1993) Matching pursuits with time–frequency dictionaries. IEEE Trans Signal Process 41(12):3397–3415ADSCrossRefzbMATHGoogle Scholar
  64. Meehl GA (1997) The South Asian monsoon and the tropospheric biennial oscillation. J Climate 10(8):1921–1943ADSCrossRefGoogle Scholar
  65. Mo KC (1993) The global climate of September–November 1990: ENSO–like warming in the western pacific and strong ozone depletion over Antarctica. J Climate 6(7):1375–1391ADSCrossRefGoogle Scholar
  66. Oil Gas Journal 80–90 (1982–1992)Google Scholar
  67. Oil gas journal data book (1984–1998 edn). PennWells, TulsaGoogle Scholar
  68. Oil Gas Journal Data Book (1990) 1990 edn. PennWells, TulsaGoogle Scholar
  69. Penner JE (1986) Uncertainties in the smoke source term for ‘nuclear winter’ studies. Nature 29(6094):222–226ADSCrossRefGoogle Scholar
  70. Philips D (ed) (1995) The global climate system review (climate system monitoring June 1991–November 1993). WMO–IP No. 819, World Meteorological Organization, Geneva. ISBN 92-63-10819-6Google Scholar
  71. Pittock AB, Ackerman TP, Crutzen PJ, MacCracken MC, Shapiro CS, Turco RP (1986) SCOPE 28: environmental consequences of nuclear war. Vol I. Physical and atmospheric effects. Wiley, Chichester, p 359. (2nd edn with an updating preface (1989))Google Scholar
  72. Pruppacher HR, Klett JD (1978) Microphysics of clouds and precipitation. Reidel, HinghamCrossRefGoogle Scholar
  73. Pueschel RF, Livingston JM (1990) Aerosol spectral optical depths: jet fuel and forest fire smoke. J Geophys Res D 95(13):22,417–22,422CrossRefGoogle Scholar
  74. Robock A (1988) Enhancement of surface cooling due to forest fire smoke. Science 242(4850):911–913ADSCrossRefGoogle Scholar
  75. Seacor JE (1994) Environmental terrorism: lessons from the oil fires of Kuwait. Am Univ Int Law Rev 10(1):481–523Google Scholar
  76. Segal M, Weaver J, Purdom JFW (1989) Some effects of the Yellowstone fire plume on northeast Colorado at the end of summer 1988. Mon Weather Rev 117(10):2278–2284ADSCrossRefGoogle Scholar
  77. Seiler W, Crutzen PJ (1980) Estimates of gross and net fluxes of Carbon between the biosphere and the atmosphere from biomass burning. Clim Chang 2(3):207–247CrossRefGoogle Scholar
  78. Setzer AW, Pereira MC (1991) Amazon biomass burning in 1987 and an estimate of their tropospheric emissions. AMBIO 20(1):19–22Google Scholar
  79. Small RD (1991) Environmental impact of fires in Kuwait. Nature 350(6313):111–112ADSCrossRefGoogle Scholar
  80. Stenchikov GL, Carl P (1985) Climatic consequences of nuclear war: sensitivity against large–scale inhomogeneities in the initial atmospheric pollutions. GDR Academy of Science and Physical Society, BerlinGoogle Scholar
  81. Subrahmanyam K, Carasales JC, Fakhr A, Müller M, Osgood RE, Svensson U, Wegener H, Zhurkin VV (1987) Study on deterrence. United Nations Publication A/41/432. United Nations, New York. ISBN 92-1-142127-6Google Scholar
  82. Swetnam TW, Betancourt JL (1990) Fire–southern oscillation relations in the southwestern United States. Science 249(4972):1017–1020ADSCrossRefGoogle Scholar
  83. Thompson SL, Aleksandrov VV, Stenchikov GL, Schneider SH, Covey C, Chervin RM (1984) Global climatic consequences of nuclear war: simulations with three dimensional models. AMBIO 13(4):236–243Google Scholar
  84. Thunborg AI et al (1981) Comprehensive study on nuclear weapons – report of the secretary-general. United Nations Publication A/35/392, UN Study Series Disarmament, No. 1. United Nations, New YorkGoogle Scholar
  85. Torrence C, Compo GP (1998) A practical guide to wavelet analysis. Bull Am Meteorol Soc 79(1):61–78CrossRefGoogle Scholar
  86. Tschentscher I, Worbs KD, Carl P (1994) Frequency drift and retreat variability of a GCM’s monsoon oscillator. WCRP-84, WMO/TD-No. 619(II), WMO, Geneva, pp 781–788Google Scholar
  87. Turco RP, Toon OB, Ackerman TP, Pollack JB, Sagan C (1983) Nuclear winter: global consequences of multiple nuclear explosions. Science 222(4630):1283–1292ADSCrossRefGoogle Scholar
  88. Turco RP, Toon OB, Ackerman TP, Pollack JB, Sagan C (1990) Climate and smoke: an appraisal of nuclear winter. Science 247(4939):166–176ADSCrossRefGoogle Scholar
  89. Turner PJ (1992) ACE/gr user’s manual. Graphics for exploratory data analysis. Software documentation series, SDS3, 91–11Google Scholar
  90. Vautard R, Ghil M (1989) Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series. Physica D 35(3):395–424MathSciNetADSCrossRefzbMATHGoogle Scholar
  91. Vautard R, Yiou P, Ghil M (1992) Singular–spectrum analysis: a toolkit for short, noisy chaotic signals. Physica D 58(1–4):95–126ADSCrossRefGoogle Scholar
  92. Veltishchev NN, Ginzburg AS, Golitsyn GS (1988) Climatic effects of mass fire (in Russian). Izvestiya – Fiz Atmos Okeana 24:296–304Google Scholar
  93. Webster PJ, Magaña VO, Palmer TN, Shukla J, Tomas RA, Yanai M, Yasunari T (1998) Monsoon: processes, predictability, and the prospects for prediction. J Geophys Res 103(C7):14,451–14,510ADSCrossRefGoogle Scholar
  94. Wexler H (1950) The great smoke pall—September 24–30, 1950. Weatherwise 3(6):129–134, 142Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2014

Authors and Affiliations

  1. 1.ASWEX – Applied Water ResearchBerlinGermany

Personalised recommendations