Spring flood reconstruction from tree rings (continuous and discontinuous series)

É"enne  Boucher  Taha  B.M.  J.  Ouarda  

Yves  Bégin    Antoine  Nicault  

Ins<tut  Na<onal  de  la  Recherche  Scien<fique  Centre  Eau-­‐Terre-­‐Environnement  

[email protected]  

Photo  credit:  S.  Popov   World  Dendro  2010,  Rovaniemi,  Finland  

Why  reconstruct  spring  floods?  

Hudson  bay  

LA  GRANDE  (LG)  hydroelectric  complex  

Why  reconstruct  spring  floods?  

Infilling  of  hydroelectric  reservoirs  

Spring  

Why  reconstruct  spring  floods?  

[email protected]  

Pros  

Cons  

A  mul<-­‐proxy  dendrohydrological  approach  to  reconstruct  spring  floods  

Discon<nuous  series  /  uncertain<es  /  non-­‐lineari<es  

Direct  rela<onship  with  streamflow  Con<nuous  series  

Indirect  rela<onships  with  

streamflow  

Discrete  event  series  e.g.  ice  scarred  trees  

Uncertain  discrete  series  

[0.05,  0.5,  0.95]  

Propor<on  of  ice-­‐scarred  trees  around  a  northern  lake  

Prop

or<o

ns  

 N  trees  sampled

[email protected]  

Model  uncertain"es  of  discrete  event  series  Uncertain<es  around  observed  propor<ons  are  binomialy  distributed  1)    

Model  verifica"on  Jackknife  resampling  3)    

Calibra"on  /  Reconstruc"ons    Performed  using  Generalized  Addi<ve  Models  (GAMs)  2)    

€

g(E(Q)) = B0 + f1TR1 + f2TR2 + ...

[email protected]  

Reconstruc<on  of  May  water  supplies  to  the  Caniapiscau  reservoir  (northern  Québec,  Canada)  

[email protected]  

Hudson  bay  

La  Grande  HE  complex  

Dendrochronological  data  

=  Earlywood  density   =  Ice  scar  chronology  at  lake  Montausier  

[email protected]  

Water  supplies  to  the  Caniapiscau  reservoir,    &  hydrological  variables  to  reconstruct  

[email protected]  

Discrete  and  con<nuous  dendroseries  

Propor<on  of  ice  scoured  trees  around  lake  Montausier  

Earlywood  density,  PC1  

Earlywood  density,  PC2  

[email protected]  

The  Montausier  lake  ice  scar  chronology  (an  other  look  at  it…)  

Prop

or<o

ns  

N=300  

[email protected]  

For  each  year,  N=300  possible  propor<ons  were  randomly  generated    from  a  binomial  distribu<on.      

Colors  indicate  «  the  number  of  "mes  a  simulated  propor<on  value  falls  in  each  class  »  

Rela<onships  between  TR  series  &    water  supplies  (QMay)  to  the  Caniapiscau  Reservoir  

Propor<ons  of  ice  scarred  trees,  Montausier  lake  

PC1    Index  

(wet  sites)  

PC2    index  

(mesic  xeric  sites)  

Calibra<on  /  verifica<on  sta<s<cs    &  comparison  with  a  tradi<onal  MLR  approach  

€

g(E(QMay )) = B0 + f1X1 + f2X2 + ...

€

E(QMay ) = B0 + B1X1 + B2X2 + ...

+  

[email protected]  

Historical  May  water  supplies  to  the    Caniapiscau  Reservoir  

[email protected]  

Discussion  and  Conclusion  

-­‐  A  new  approach  to  reconstruct  paleofloods  from  tree  rings  in  northern  environment.  It  allows:    

1)  to  take  advantage  of  both  discrete  and  con"nuous  tree  ring  proxys  

2)  to  model  and  include  the  uncertainty  of  discrete  TR  series  within  the  calibra<on  /  reconstruc<on  process  

3)  to  account  for  non-­‐linear  rela"onships  that  occur  in  natural  environments  

-­‐      Our  approach  can  be  used  in  other  environments  /  situa<ons  where  discrete  and  con<nuous  tree  ring  series  are  available  

BIENVILLE  MONTAUSIER  

CORVETTE  

LEC  

NECOPASTIC  

HERVÉ  

POLARIS  

Many  thanks  to  

+  field  assistants  throughout  the  years  [email protected]