I'm attaching a .csv file with our welfare changes for the FUND regions (% change in the agricultural sector) that we used to parameterize the damage functions. The "Meta-Analysis" is our analysis of the IPCC results (which also has results for the error bars), and AgMIP NoN is our preferred AgMIP ensemble. 

Delavane did both a linear interpolation between these points and fit a quadratic through them to get the damage function. The R code she used to do that based on the attached is copied below.

Let me know if there's anything else you need. Did you end up submitting an SCC session to the World Congress? 

Best,
Fran

# ImpactAgricultureComponent
agrate[t,r,'ref'] = if (t==1) agrbm * (0.005 / 0.04)^agnl * agtime else agrbm * (dtemp / 0.04)^agnl + (1- 1/ agtime) * agrate[t-1,r,'ref']
aglevel[t,r,'ref'] = aglparl * deltat[t] + aglparq * (deltat[t]^2)
agco2[t,r,'ref'] = if (t==1) 0 else agcbm /log(2) * log(conc[t - 1] / concpre) 

# Ag damages are fractional loss of Regional Ag Product - damages are negative
AgLossFUND[t,r,'ref']=max(-1,agrate[t,r,'ref'] + aglevel[t,r,'ref'] + agco2[t,r,'ref'])

damages[t,r,'Agriculture',i] = -1*income[t, r]*agrish0*( ypc_F[t,r] / ypc90[r])^-agel*AgLossFUND[t,r,'ref']



# Read data for GTAP DFs ---------------------------------------------------------------
load("welfarewithFUNDregions.Rdat")
# Formulate alternative, linear damage functions of temperature change using point estimates 
# midDF takes temperature as x and returns welfare % change (neg is loss)
midDF=acast(regiondata[which(regiondata$Model=='Meta-Analysis'),c('Temp','FUND','mid')], Temp ~ FUND)
lowDF=acast(regiondata[which(regiondata$Model=='Meta-Analysis'),c('Temp','FUND','low')], Temp ~ FUND)
highDF=acast(regiondata[which(regiondata$Model=='Meta-Analysis'),c('Temp','FUND','high')], Temp ~ FUND)
AgMIP_NoNDF=acast(regiondata[which(regiondata$Model=='AgMIP_NoN'),c('Temp','FUND','mid')], Temp ~ FUND)
AgMIP_AllDF=acast(regiondata[which(regiondata$Model=='AgMIP_All'),c('Temp','FUND','mid')], Temp ~ FUND)


# GTAP damages are not time dependent, so compute outside of the time loop
# selectedAgDF is one of the DFs constructed above
# deltat and Pulsedeltat are the timeseries of temperature change
for (r in FUNDregions) {
  AgLossGTAP[,r,'ref']=pmax(-100,approxExtrap(0:3,c(0,selectedAgDF[,r]),xout=deltat)$y)/100
  AgLossGTAP[,r,'pulse']=pmax(-100,approxExtrap(0:3,c(0,selectedAgDF[,r]),xout=Pulsedeltat)$y)/100
}

damages[t,r,'Agriculture',i] = -1*income[t, r]*agrish0*( ypc_F[t,r] / ypc90[r])^-agel*AgLossGTAP[t,r,'ref']

