The Smoking Code, part 2

Climategate Code Analysis Part 2

atomThere are three common issues that have been raised in my previous post that I would like to officially address concerning the CRU’s source code.

If you only get one thing from this post, please get this. I am only making a statement about the research methods of the CRU and trying to show proof that they had the means and intent to falsify data. And, until the CRU’s research results can be verified by a 3rd party, they cannot be trusted.

Here are the four most frequent concerns dealing with the CRU’s source code:

  1. The source code that actually printed the graph was commented out and, therefore, is not valid proof.
  2. No proof exists that shows this code was used in publishing results.
  3. Interpolation is a normal part of dealing with large data sets, this is no different.
  4. You need the raw climate data to prove that foul play occurred.

If anyone can think of something I missed, please let me know.

The source code that actually printed the graph was commented out and, therefore, is not valid proof.

Had I done a better job with my source analysis, I would have found a later revision of the briffa_sep98_d.pro source file (linked to in my previous post) contained in a different working tree which shows the fudge-factor array playing a direct result in the (uncommented) plotting of the data.

Snippit from: harris-tree/briffa_sep98_e.pro (see the end of the post for the full source listing)

;

; APPLY ARTIFICIAL CORRECTION

;

yearlyadj=interpol(valadj,yrloc,x)

densall=densall+yearlyadj

  ;

  ; Now plot them

  ;

  filter_cru,20,tsin=densall,tslow=tslow,/nan

  cpl_barts,x,densall,title='Age-banded MXD from all sites',$

    xrange=[1399.5,1994.5],xtitle='Year',/xstyle,$

    zeroline=tslow,yrange=[-7,3]

  oplot,x,tslow,thick=3

  oplot,!x.crange,[0.,0.],linestyle=1

  ;

Now, we can finally put this concern to rest.

Interpolation is a normal part of dealing with large data sets, this is no different.

This is partially true, the issue doesn’t lie in the fact that the CRU researchers used interpolation. The issue is the weight of the valadj array with respect to the raw data. valadj simply introduces too large of an influence to the original data to do anything productive with it.

Here is the graph I plotted of the valadj array. When we’re talking about trying to interpret temperature data that grows on the scale of one-tenths of a degree over a period of time, “fudging” a value by 2.5 is going to have a significant impact on the data set.

No proof exists that shows this code was used in publishing results.

Correct! That’s why I am (and always have) taken the following stand: Enough proof exists that the CRU had both the means and intent to intentionally falsify data. This means that all of their research results cannot be trusted until they are verified. Period.

The fact that the “fudge-factor” source code exists in the first place is reason enough for alarm. Hopefully, they didn’t use fudged results in the CRU research results, but the truth is, we just don’t know.

You need the raw climate data to prove that foul play occurred.

This is assuming the raw data are valid, which I maintain that it probably is. Several people question the validity of the climate data gathering methods used by the different climate research institutions, but I am not enough of a climate expert to have an opinion one way or the other. Furthermore, It simply doesn’t matter if the raw climate data are correct or not to demonstrate the extreme bias the valadj array forces on the raw data.

So, the raw data could actually be temperature data or corporate sales figures, the result is the same; a severe manipulation of data.

Full Source Listing

As promised, here is the entire source listing for: harris-tree/briffa_sep98_e.pro

[sourcecode language=”text”]

1. ;

2. ; PLOTS ‘ALL’ REGION MXD timeseries from age banded and from hugershoff

3. ; standardised datasets.

4. ; Reads Harry’s regional timeseries and outputs the 1600-1992 portion

5. ; with missing values set appropriately. Uses mxd, and just the

6. ; "all band" timeseries

7. ;****** APPLIES A VERY ARTIFICIAL CORRECTION FOR DECLINE*********

8. ;

9. yrloc=[1400,findgen(19)*5.+1904]

10. valadj=[0.,0.,0.,0.,0.,-0.1,-0.25,-0.3,0.,-0.1,0.3,0.8,1.2,1.7,2.5,2.6,2.6,$

11. 2.6,2.6,2.6]*0.75 ; fudge factor

12. if n_elements(yrloc) ne n_elements(valadj) then message,’Oooops!’

13. ;

14. loadct,39

15. def_1color,20,color=’red’

16. plot,[0,1]

17. multi_plot,nrow=4,layout=’large’

18. if !d.name eq ‘X’ then begin

19. window, ysize=800

20. !p.font=-1

21. endif else begin

22. !p.font=0

23. device,/helvetica,/bold,font_size=18

24. endelse

25. ;

26. ; Get regional tree lists and rbar

27. ;

28. restore,filename=’reglists.idlsave’

29. harryfn=[‘nwcan’,’wnam’,’cecan’,’nweur’,’sweur’,’nsib’,’csib’,’tib’,$

30. ‘esib’,’allsites’]

31. ;

32. rawdat=fltarr(4,2000)

33. for i = nreg-1 , nreg-1 do begin

34. fn=’mxd.’+harryfn(i)+’.pa.mean.dat’

35. print,fn

36. openr,1,fn

37. readf,1,rawdat

38. close,1

39. ;

40. densadj=reform(rawdat(2:3,*))

41. ml=where(densadj eq -99.999,nmiss)

42. densadj(ml)=!values.f_nan

43. ;

44. x=reform(rawdat(0,*))

45. kl=where((x ge 1400) and (x le 1992))

46. x=x(kl)

47. densall=densadj(1,kl) ; all bands

48. densadj=densadj(0,kl) ; 2-6 bands

49. ;

50. ; Now normalise w.r.t. 1881-1960

51. ;

52. mknormal,densadj,x,refperiod=[1881,1960],refmean=refmean,refsd=refsd

53. mknormal,densall,x,refperiod=[1881,1960],refmean=refmean,refsd=refsd

54. ;

55. ; APPLY ARTIFICIAL CORRECTION

56. ;

57. yearlyadj=interpol(valadj,yrloc,x)

58. densall=densall+yearlyadj

59. ;

60. ; Now plot them

61. ;

62. filter_cru,20,tsin=densall,tslow=tslow,/nan

63. cpl_barts,x,densall,title=’Age-banded MXD from all sites’,$

64. xrange=[1399.5,1994.5],xtitle=’Year’,/xstyle,$

65. zeroline=tslow,yrange=[-7,3]

66. oplot,x,tslow,thick=3

67. oplot,!x.crange,[0.,0.],linestyle=1

68. ;

69. endfor

70. ;

71. ; Restore the Hugershoff NHD1 (see Nature paper 2)

72. ;

73. xband=x

74. restore,filename=’../tree5/densadj_MEAN.idlsave’

75. ; gets: x,densadj,n,neff

76. ;

77. ; Extract the post 1600 part

78. ;

79. kl=where(x ge 1400)

80. x=x(kl)

81. densadj=densadj(kl)

82. ;

83. ; APPLY ARTIFICIAL CORRECTION

84. ;

85. yearlyadj=interpol(valadj,yrloc,x)

86. densadj=densadj+yearlyadj

87. ;

88. ; Now plot it too

89. ;

90. filter_cru,20,tsin=densadj,tslow=tshug,/nan

91. cpl_barts,x,densadj,title=’Hugershoff-standardised MXD from all sites’,$

92. xrange=[1399.5,1994.5],xtitle=’Year’,/xstyle,$

93. zeroline=tshug,yrange=[-7,3],bar_color=20

94. oplot,x,tshug,thick=3,color=20

95. oplot,!x.crange,[0.,0.],linestyle=1

96. ;

97. ; Now overplot their bidecadal components

98. ;

99. plot,xband,tslow,$

100. xrange=[1399.5,1994.5],xtitle=’Year’,/xstyle,$

101. yrange=[-6,2],thick=3,title=’Low-pass (20-yr) filtered comparison’

102. oplot,x,tshug,thick=3,color=20

103. oplot,!x.crange,[0.,0.],linestyle=1

104. ;

105. ; Now overplot their 50-yr components

106. ;

107. filter_cru,50,tsin=densadj,tslow=tshug,/nan

108. filter_cru,50,tsin=densall,tslow=tslow,/nan

109. plot,xband,tslow,$

110. xrange=[1399.5,1994.5],xtitle=’Year’,/xstyle,$

111. yrange=[-6,2],thick=3,title=’Low-pass (50-yr) filtered comparison’

112. oplot,x,tshug,thick=3,color=20

113. oplot,!x.crange,[0.,0.],linestyle=1

114. ;

115. ; Now compute the full, high and low pass correlations between the two

116. ; series

117. ;

118. perst=1400.

119. peren=1992.

120. ;

121. openw,1,’corr_age2hug.out’

122. thalf=[10.,30.,50.,100.]

123. ntry=n_elements(thalf)

124. printf,1,’Correlations between timeseries’

125. printf,1,’Age-banded vs. Hugershoff-standardised’

126. printf,1,’ Region Full <10 >10 >30 >50 >100′

127. ;

128. kla=where((xband ge perst) and (xband le peren))

129. klh=where((x ge perst) and (x le peren))

130. ts1=densadj(klh)

131. ts2=densall(kla)

132. ;

133. r1=correlate(ts1,ts2)

134. rall=fltarr(ntry)

135. for i = 0 , ntry-1 do begin

136. filter_cru,thalf(i),tsin=ts1,tslow=tslow1,tshigh=tshi1,/nan

137. filter_cru,thalf(i),tsin=ts2,tslow=tslow2,tshigh=tshi2,/nan

138. if i eq 0 then r2=correlate(tshi1,tshi2)

139. rall(i)=correlate(tslow1,tslow2)

140. endfor

141. ;

142. printf,1,’ALL SITES’,r1,r2,rall,$

143. format='(A11,2X,6F6.2)’

144. ;

145. printf,1,’ ‘

146. printf,1,’Correlations carried out over the period ‘,perst,peren

147. ;

148. close,1

149. ;

150. end

[/sourcecode]

0 0 votes
Article Rating
209 Comments
Oldest
Newest Most Voted