[163] | 1 | ;+ |
---|
[93] | 2 | ; |
---|
| 3 | ; compute point-wise slope of linear regression of two 3D (x,y,t) fields |
---|
[163] | 4 | ; |
---|
| 5 | ; @version |
---|
| 6 | ; $Id$ |
---|
| 7 | ; |
---|
| 8 | ;- |
---|
[93] | 9 | FUNCTION make_linfit, file_name, ncdf_db, BOXZOOM = boxzoom, TIME_1 = time_1, TIME_2 = time_2, ALL_DATA = all_data |
---|
| 10 | |
---|
| 11 | @common |
---|
| 12 | @com_eg |
---|
| 13 | |
---|
| 14 | IF debug_w THEN print, ' ENTER make_linfit... ' |
---|
| 15 | |
---|
| 16 | var_name1 = (strsplit(macro_base_fld, ',', ESCAPE = ' ', /EXTRACT))[0] |
---|
| 17 | var_name2 = (strsplit(macro_base_fld, ',', ESCAPE = ' ', /EXTRACT))[1] |
---|
| 18 | |
---|
| 19 | ; Build file_name2 if different |
---|
| 20 | |
---|
| 21 | IF strpos (cmd1_back.grid, '#') NE -1 THEN BEGIN |
---|
| 22 | varpos = strpos(file_name, var_name1) |
---|
| 23 | file_name2= strmid(file_name, 0, varpos)+var_name2+'.nc' |
---|
| 24 | IF debug_w THEN print, ' file_name2 = ', file_name2 |
---|
| 25 | ENDIF ELSE file_name2 = file_name |
---|
| 26 | |
---|
| 27 | ; Read the variables in the correspondant netcdf file |
---|
| 28 | var1 = nc_read(file_name, var_name1, ncdf_db, BOXZOOM = boxzoom, TIME_1 = time_1, TIME_2 = time_2) |
---|
| 29 | var2 = nc_read(file_name2, var_name2, ncdf_db, BOXZOOM = boxzoom, TIME_1 = time_1, TIME_2 = time_2) |
---|
| 30 | |
---|
| 31 | varname = var1.name+'/'+var2.name+' corr.' |
---|
| 32 | varunit = '[-1/1]' |
---|
| 33 | |
---|
| 34 | nxa = (size(var1.data))[1] |
---|
| 35 | nya = (size(var1.data))[2] |
---|
| 36 | |
---|
| 37 | IF debug_w THEN print, ' nxa, nya', nxa, nya |
---|
| 38 | |
---|
| 39 | pt_linfit = fltarr(nxa, nya) |
---|
| 40 | pt_err = fltarr(nxa, nya) |
---|
| 41 | pt_corr = fltarr(nxa, nya) |
---|
| 42 | pt_linfit[*, *] = 0 |
---|
| 43 | pt_err[*, *] = 1.1 |
---|
| 44 | pt_corr[*, *] = 0. |
---|
| 45 | |
---|
| 46 | ; Sampling of data and computation of new numbers of values |
---|
| 47 | @mth_decode |
---|
| 48 | |
---|
| 49 | FOR imth = 0, nmth-1 DO BEGIN |
---|
| 50 | |
---|
| 51 | IF debug_w THEN print, ' month idx/value: ', imth, strd(imth) |
---|
| 52 | |
---|
| 53 | data1 = (var1.data)[*, *, reform(idxm(imth,*), njpt)] |
---|
| 54 | data2 = (var2.data)[*, *, reform(idxm(imth,*), njpt)] |
---|
| 55 | nval = njpt |
---|
| 56 | |
---|
| 57 | ; Compute linear regression |
---|
| 58 | |
---|
| 59 | |
---|
| 60 | FOR idx = 0, nxa-1 DO FOR idy = 0, nya-1 DO BEGIN |
---|
| 61 | |
---|
[106] | 62 | IF data1(idx, idy, 0) LT valmask/10. THEN BEGIN |
---|
[93] | 63 | |
---|
| 64 | x1i = reform(data1(idx, idy, *), nval) |
---|
| 65 | x2i = reform(data2(idx, idy, *), nval) |
---|
| 66 | x1 = x1i-mean(x1i) |
---|
| 67 | x2 = x2i-mean(x2i) |
---|
| 68 | |
---|
[106] | 69 | no_compute = 0 |
---|
| 70 | |
---|
[93] | 71 | IF linfit_map NE '' THEN BEGIN |
---|
| 72 | ; compute regression for value above (p - default) or below (m) linfit_sep |
---|
| 73 | CASE linfit_map OF |
---|
| 74 | 'm': idt = where (x2 LE linfit_sep) |
---|
| 75 | ELSE: idt = where (x2 GE linfit_sep) |
---|
| 76 | ENDCASE |
---|
[109] | 77 | IF (size(idt))[1] GE 3 THEN BEGIN |
---|
| 78 | IF idt[0] NE -1 THEN BEGIN |
---|
| 79 | x1 = x1(idt) |
---|
| 80 | x2 = x2(idt) |
---|
| 81 | ENDIF |
---|
[106] | 82 | ENDIF ELSE BEGIN |
---|
| 83 | no_compute = 1 |
---|
| 84 | ENDELSE |
---|
[93] | 85 | ENDIF |
---|
| 86 | |
---|
[106] | 87 | ; BEGIN |
---|
[93] | 88 | ; print, ' idx, idy, linfit_map ', idx, idy, linfit_map |
---|
| 89 | ; print, ' size(idt)',size(idt) |
---|
| 90 | ; stop |
---|
| 91 | ; ENDIF |
---|
| 92 | |
---|
[107] | 93 | IF no_compute EQ 0 THEN BEGIN |
---|
[93] | 94 | |
---|
[106] | 95 | coeffl = linfit(x2, x1, CHISQ = linerrl, PROB = proberrl, SIGMA = sigmaerrl) |
---|
| 96 | correl = c_timecorrelate(x2,x1) |
---|
| 97 | pt_linfit(idx, idy) = pt_linfit(idx, idy)+coeffl(1) |
---|
| 98 | pt_err(idx, idy) = min(pt_err(idx, idy), proberrl) |
---|
| 99 | pt_corr(idx, idy) = pt_corr(idx, idy) + correl |
---|
| 100 | |
---|
| 101 | IF debug_w THEN BEGIN |
---|
| 102 | IF idx EQ 30 AND idy EQ 15 THEN BEGIN |
---|
| 103 | print, ' idx, idy, linfit_map ', idx, idy, linfit_map |
---|
| 104 | print, ' size(idt)',size(idt) |
---|
| 105 | print, ' pt_linfit, correl = ',coeffl(1), correl |
---|
| 106 | ENDIF |
---|
[93] | 107 | ENDIF |
---|
[106] | 108 | ENDIF ELSE BEGIN |
---|
| 109 | pt_linfit(idx, idy) = valmask |
---|
| 110 | pt_err(idx, idy) = 0. |
---|
| 111 | pt_corr(idx, idy) = 0. |
---|
| 112 | ENDELSE |
---|
[93] | 113 | |
---|
| 114 | ENDIF ELSE BEGIN |
---|
| 115 | pt_linfit(idx, idy) = valmask |
---|
| 116 | pt_err(idx, idy) = 0. |
---|
| 117 | pt_corr(idx, idy) = 0. |
---|
| 118 | ENDELSE |
---|
| 119 | |
---|
| 120 | ENDFOR |
---|
| 121 | ENDFOR |
---|
| 122 | |
---|
| 123 | ; make mean of period |
---|
| 124 | |
---|
| 125 | idm = where(pt_linfit GE valmask/10.) |
---|
| 126 | pt_linfit = pt_linfit/float(nmth) |
---|
| 127 | pt_corr = pt_corr/float(nmth) |
---|
| 128 | IF idm[0] NE -1 THEN pt_linfit(idm) = valmask |
---|
| 129 | |
---|
[106] | 130 | ; only take points where correlation larger than 0.2 |
---|
[93] | 131 | |
---|
[112] | 132 | idc = where(abs(pt_corr) LT corr_thres) |
---|
[122] | 133 | IF idc[0] NE -1 THEN pt_linfit(idc) = !values.f_nan |
---|
[93] | 134 | |
---|
[112] | 135 | corr_txt = '[r>'+strtrim(string(corr_thres, format = '(f5.2)'), 2)+']' |
---|
| 136 | |
---|
[93] | 137 | varname = varname+' '+ntxt |
---|
| 138 | |
---|
| 139 | print, ' Linfit min/max pt_corr ', min(pt_corr), max(pt_corr) |
---|
| 140 | |
---|
| 141 | ; Define the outputs of the function |
---|
| 142 | field = {name: varname, data: pt_linfit, legend: '', units: '', origin: '', dim: 0, direc:'xy'} |
---|
| 143 | field.origin = var1.origin |
---|
| 144 | field.dim = var1.dim |
---|
| 145 | |
---|
| 146 | CASE linfit_map OF |
---|
| 147 | 'p': BEGIN |
---|
[101] | 148 | print, ' Linear fit computed for anomalies ABOVE ',linfit_sep |
---|
[93] | 149 | lintxt = ' > '+ strtrim(string(linfit_sep), 2) |
---|
| 150 | END |
---|
| 151 | 'm': BEGIN |
---|
[101] | 152 | print, ' Linear fit computed for anomalies BELOW ',linfit_sep |
---|
[93] | 153 | lintxt = ' < '+ strtrim(string(linfit_sep), 2) |
---|
| 154 | END |
---|
| 155 | ELSE: lintxt = '' |
---|
| 156 | ENDCASE |
---|
| 157 | |
---|
[112] | 158 | field.legend = lintxt+' for '+ntxt+' in ['+cmdm.date1+'-'+cmdm.spec+'] '+corr_txt+' -' |
---|
[93] | 159 | |
---|
| 160 | ; additional computations (pac_5 nino_3 nino_4 averages) *** requires whole domain **** |
---|
| 161 | |
---|
| 162 | IF nxt EQ jpi AND nyt EQ jpj THEN BEGIN |
---|
| 163 | |
---|
| 164 | n3_linfit = moyenne(pt_linfit,'xy', boite = [210, 270, -5, 5], NaN = valmask) |
---|
| 165 | n4_linfit = moyenne(pt_linfit,'xy', boite = [160, 210, -5, 5], NaN = valmask) |
---|
| 166 | zl_linfit = moyenne(pt_linfit,'xy', boite = [130, 280, -5, 5], NaN = valmask) |
---|
| 167 | |
---|
[101] | 168 | print, ' Nino 3 average of slope of linear fit = ', n3_linfit |
---|
| 169 | print, ' Nino 4 average of slope of linear fit = ', n4_linfit |
---|
| 170 | print, ' Zonal 5N/5S average of slope of linear fit = ', zl_linfit |
---|
[93] | 171 | |
---|
| 172 | ENDIF |
---|
| 173 | |
---|
| 174 | ; jpt=nval |
---|
| 175 | ; meants=grossemoyenne(data2, 't') |
---|
| 176 | ; meantsp=reform(meants,nxt*nyt) |
---|
| 177 | ; meanlf=reform(pt_linfit,nxt*nyt) |
---|
| 178 | ; IF min (meantsp) GE 100 THEN meantsp=meantsp-273.15 |
---|
| 179 | ; pltsc, meanlf,meantsp,-70,30,20,34,"SST "+mth[strd-1], window=2 |
---|
| 180 | ; pltsc, meanlf,meantsp,-70,30,20,34,"SST "+mth[strd-1], window=2, /noerase, /ov1d, col1d=2 |
---|
| 181 | |
---|
| 182 | ; stop |
---|
| 183 | |
---|
| 184 | IF debug_w THEN print, ' ... EXIT make_linfit' |
---|
| 185 | |
---|
| 186 | return, field |
---|
| 187 | |
---|
| 188 | END |
---|