| """Rebuild core quantitative panels at 183 mm with vector exports and source tables.""" |
| from pathlib import Path |
| import json, hashlib |
| import numpy as np |
| import pandas as pd |
| import matplotlib |
| matplotlib.use('Agg') |
| import matplotlib.pyplot as plt |
| ROOT=Path(__file__).resolve().parents[1] |
| OUT=ROOT/'figures/publication_20261007' |
| plt.rcParams.update({'font.family':'DejaVu Sans','font.size':7,'axes.labelsize':7,'axes.titlesize':7, |
| 'xtick.labelsize':7,'ytick.labelsize':7,'legend.fontsize':7,'axes.spines.top':False,'axes.spines.right':False, |
| 'axes.linewidth':.7,'lines.linewidth':1.0,'pdf.fonttype':42,'ps.fonttype':42,'svg.fonttype':'none','savefig.dpi':300}) |
| BLUE='#0072B2';ORANGE='#D55E00';GREEN='#009E73';GRAY='#777777' |
| manifest=[] |
| def source(rel,name): |
| p=ROOT/rel;d=pd.read_csv(p);d.to_csv(OUT/(name+'.csv'),index=False) |
| manifest.append({'path':rel,'sha256':hashlib.sha256(p.read_bytes()).hexdigest()});return d |
|
|
| def save(fig,name): |
| for ax in fig.axes: |
| title=ax.get_title(loc='left') |
| if title and title[0] in 'abcd': |
| ax.set_title(title[1:].strip(),loc='left',fontsize=7) |
| ax.text(-.10,1.05,title[0],transform=ax.transAxes,fontsize=8,fontweight='bold') |
| for ext in ['pdf','svg','png']:fig.savefig(OUT/f'{name}.{ext}') |
| plt.close(fig) |
|
|
| def main(): |
| OUT.mkdir(exist_ok=True,parents=True) |
| d=source('data/derived/trajectory_audit_20261007/endpoint_year_summary.csv','figure2_source') |
| fig,axs=plt.subplots(1,2,figsize=(183/25.4,82/25.4),layout='constrained',gridspec_kw={'width_ratios':[1.2,1]}) |
| ax=axs[0] |
| for col,label,color,style in [('stored_mean','Stored: retained transitions',ORANGE,'-'),('recomputed_model_mean','Area-derived: same transitions',BLUE,'-'),('all_pair_mean','Area-derived: all transitions',GRAY,'--')]: |
| ax.plot(d.year,d[col],marker='o',ms=3,color=color,ls=style,label=label) |
| ax.set(xlabel='Starting survey year',ylabel='Mean endpoint (dimensionless)',xticks=[2014,2016,2018,2020]);ax.legend(frameon=False) |
| ax.set_title('a Response and sampling sensitivity',loc='left',fontweight='bold') |
| ax=axs[1];ax.bar(d.year,d.all_pair_n,color='#DDE2E5',label='All archived transitions (957)');ax.bar(d.year,d.model_n,color=BLUE,label='Retained transitions (854)') |
| for _,r in d.iterrows():ax.text(r.year,r.all_pair_n+7,str(int(r.model_n))+'/'+str(int(r.all_pair_n)),ha='center',fontsize=6.5) |
| ax.set(xlabel='Starting survey year',ylabel='Transition records',ylim=(0,440),xticks=[2014,2016,2018,2020]);ax.legend(frameon=False,loc='upper right');ax.set_title('b Positive-growth selection',loc='left',fontweight='bold') |
| save(fig,'figure2_endpoint') |
| d=source('data/derived/strict_benchmark_20261007/pooled_metrics.csv','figure3_source') |
| d=d[d.target=='log_area_ratio'];fig,axs=plt.subplots(1,2,figsize=(183/25.4,78/25.4),layout='constrained') |
| splits=list(d.split.unique());labels={'buffered_quadrant':'Spatial holdout\n954 predictions','forward_year':'Forward years\n455 predictions'} |
| print('Benchmark splits:',splits) |
| for ax,metric,title in zip(axs,['r2','mae'],['a Squared-error performance','b Absolute-error performance']): |
| for j,(model,color,label) in enumerate([('training_mean',GRAY,'Training mean'),('ridge',BLUE,'Ridge'),('extra_trees',ORANGE,'Extremely randomized trees')]): |
| g=d[d.model==model].set_index('split').loc[splits];x=np.arange(len(splits))+(j-1)*.2;ax.scatter(x,g[metric],s=24,c=color,label=label) |
| ax.set(xticks=np.arange(len(splits)),xticklabels=[labels.get(s,s) for s in splits],xlim=(-.5,len(splits)-.5),ylabel='Pooled R²' if metric=='r2' else 'MAE of log area ratio') |
| ax.set_title(title,loc='left',fontweight='bold') |
| if metric=='r2':ax.axhline(0,color='#AAAAAA',lw=.7,ls='--');ax.legend(frameon=False,fontsize=6.5,loc='lower left') |
| save(fig,'figure3_prediction') |
| d=source('data/derived/ccav_20261007/annual_class_areas_grid_sensitivity.csv','figure4_source') |
| fig,axs=plt.subplots(1,2,figsize=(183/25.4,80/25.4),layout='constrained',sharey=True) |
| for ax,fp,letter,title in zip(axs,['full_image','historical_roi'],'ab',['Full aerial-image footprint','Historical patch ROI']): |
| for code,color,label in [(1,ORANGE,'S. alterniflora'),(3,GREEN,'Mangrove'),(255,GRAY,'Unmapped')]: |
| |
| actual=d.loc[d.label.str.contains('mangrove',case=False),'code'].unique() |
| if code==3:assert len(actual)==1;code=int(actual[0]) |
| sub=d[(d.footprint==fp)&(d.code==code)];bounds=sub.groupby('year').area_ha.agg(['min','max']);base=sub[(sub.shift_x_m==0)&(sub.shift_y_m==0)] |
| ax.fill_between(bounds.index,bounds['min'],bounds['max'],color=color,alpha=.18) |
| ax.plot(base.year,base.area_ha,'o-',color=color,ms=3,label=label) |
| ax.set(xlabel='Satellite map year',xticks=[2016,2018,2020,2022]);ax.set_title(f'{letter} {title}',loc='left',fontweight='bold') |
| axs[0].set_ylabel('Classified area (ha)');axs[1].legend(frameon=False) |
| save(fig,'figure4_landscape') |
| (OUT/'figure_manifest.json').write_text(json.dumps({'width_mm':183,'font':'DejaVu Sans','font_pt':7,'sources':manifest,'uncertainty':'Figure 4 shading is four grid phases, not classification uncertainty. Figure 2 annual means are descriptive. Figure 3 pooled holdout metrics are point estimates.'},indent=2)) |
| if __name__=='__main__':main() |
|
|
