02 08 2021

Narrative

  • year 1: Plant IWG in fall

  • year 2: IWG establishes, is harveseted once for grain

  • year 3 + 4: IWG is treated as a forage

  • how should we manage the IWG in years 3 and 4?

  • when should we harvest?

  • how many times should we harvest?

Sites

Site Planted Data collected
I2 2011 2017
R70 2015 2017-2018
R100 2016 2018-2019
  • Rosemount MN

Data collection

  • response variables
    • yield
    • quality
      • $
  • treatments
    • timing.1cut
      • boot, anthesis, dough, grain
    • follow.cut
      • none, sept, sept+oct
  • sample times
    • 1.cut, 2.cut, 3.cut,
      • .total

Site conditions 1

Site conditions 2

Timing year

Diagnostics

  • missing data
    • grain cuts
    • follow-up cuts @ R70.2018 & I2.2017
  • problematic data
    • one outlier value at R100.2019 for dough cut
    • I2

Analysis 1

null hypothesis testing

  • repeated measures
    • each plot is measured 1-3 times per year depending on assigned follow.cut treatment
    • each plot is measured over 2 consecutive years
  • random effects
    • environment
    • site:block
  • fixed effects
    • follow.cut
    • timing.1cut

Analysis 2

regression

  • responses vs.
    • GDD accumulation
    • precipitation accumulation?
    • hydrothermal time?

yield of first cut in second year of data collection

car::Anova(lmer(yield.1cut~follow.cut*timing.1cut+
                  (1|site:block)+
                  (1|env),
           data=subset(dat1, field.year=="second")))
## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: yield.1cut
##                          Chisq Df Pr(>Chisq)    
## follow.cut              0.3972  2     0.8199    
## timing.1cut            48.9246  2  2.378e-11 ***
## follow.cut:timing.1cut  1.1413  4     0.8877    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

quality of first cut in second year of data collection

car::Anova(lmer(RFQ.1cut~follow.cut*timing.1cut+
                  (1|site:block)+
                  (1|env),
           data=subset(dat1, field.year=="second")))
## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: RFQ.1cut
##                           Chisq Df Pr(>Chisq)    
## follow.cut               0.5249  2     0.7692    
## timing.1cut            165.1329  2     <2e-16 ***
## follow.cut:timing.1cut   6.2227  4     0.1831    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

quality of second cut in second year

car::Anova(lmer(RFQ.2cut~follow.cut*timing.1cut+
                  (1|block),
           data=subset(dat1, env=="R100.2019" & follow.cut!="none")))
## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: RFQ.2cut
##                         Chisq Df Pr(>Chisq)  
## follow.cut             0.0577  1    0.81022  
## timing.1cut            5.8500  2    0.05367 .
## follow.cut:timing.1cut 0.2704  2    0.87355  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

quality of second cut in both years

car::Anova(lmer(RFQ.2cut~year*follow.cut+
                  (1|block),
           data=subset(dat1, site=="R100" & follow.cut!="none")))
## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: RFQ.2cut
##                  Chisq Df Pr(>Chisq)
## year            1.1689  1     0.2796
## follow.cut      0.3024  1     0.5824
## year:follow.cut 0.1136  1     0.7361
  • compare between years at same site since no obvious carryover effect
  • conclude that no difference in quality in second cut between the first and second year at R100.

the effect of multiple cuttings

Mixed effect model outputs
dataset yield.1cut rfq.1cut .2cut .3cut
R70.2018+R100.2019 ns ns
R100.2019 ns

  • Second year datasets do not find differences among follow.cut treatment levels due to the follow.cut treatments applied in the first year.

  • no carryover effect from years observed

follow.cut contribution to yield

  • NS differences

follow.cut rfq

follow.cut contribution to RFQ

  • NS differences

yield.1cut vs. timing.1cut

yield.2cut vs. timing.1cut

yield.total vs. timing.1cut

rfq.1cut vs timing.1cut

rfq.2cut vs. timing.1cut

  • may want regression here rather than mean comparison

rfq.3cut vs. timing.1cut

car::Anova(glmer(RFQ.3cut~timing.1cut +
                  (1|env) + (1|site:block),
           data=dat3,
           family=Gamma(link="log")))
## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: RFQ.3cut
##              Chisq Df Pr(>Chisq)
## timing.1cut 0.8397  2     0.6571

rfq.total vs. treatment

$ vs treatment

  • NS differences

$ vs treatment

year

## Analysis of Deviance Table (Type II Wald chisquare tests)
## 
## Response: return.total
##                  Chisq Df Pr(>Chisq)    
## treatment       35.862  8  1.861e-05 ***
## year            16.007  2  0.0003343 ***
## treatment:year 107.818 16  1.164e-15 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## year = 2017:
##  treatment emmean   SE   df lower.CL upper.CL .group
##  AO          16.6 79.2 33.2   -144.5      178  1    
##  BO          32.3 79.2 33.2   -128.8      193  1    
##  DO          96.1 79.2 33.2    -65.0      257  12   
##  AS         101.4 79.2 33.2    -59.7      262  12   
##  DS         194.3 79.2 33.2     33.2      355  12   
##  AN         214.8 79.2 33.2     53.7      376  12   
##  DN         304.2 79.2 33.2    143.1      465  12   
##  BS         338.8 79.2 33.2    177.7      500   23  
##  BN         609.5 79.2 33.2    448.4      771    3  
## 
## year = 2018:
##  treatment emmean   SE   df lower.CL upper.CL .group
##  AO         135.4 79.2 33.2    -25.7      297  1    
##  BO         149.2 79.2 33.2    -11.9      310  12   
##  BN         236.3 79.2 33.2     75.2      397  123  
##  BS         272.4 79.2 33.2    111.3      433  123  
##  AS         311.4 79.2 33.2    150.3      473  123  
##  AN         316.3 79.2 33.2    155.2      477  123  
##  DO         321.8 79.2 33.2    160.7      483  123  
##  DS         433.4 79.2 33.2    272.3      594   23  
##  DN         451.8 79.2 33.2    290.7      613    3  
## 
## year = 2019:
##  treatment emmean   SE   df lower.CL upper.CL .group
##  DO         217.5 88.0 43.7     40.0      395  1    
##  BN         263.4 79.2 33.2    102.3      424  1    
##  BO         264.9 79.2 33.2    103.8      426  1    
##  DS         294.4 79.2 33.2    133.3      455  12   
##  DN         323.8 79.2 33.2    162.7      485  12   
##  BS         369.6 79.2 33.2    208.5      531  12   
##  AN         564.8 79.2 33.2    403.7      726   23  
##  AO         580.1 79.2 33.2    419.0      741   23  
##  AS         696.3 79.2 33.2    535.2      857    3  
## 
## Degrees-of-freedom method: kenward-roger 
## Confidence level used: 0.95 
## P value adjustment: tukey method for comparing a family of 9 estimates 
## significance level used: alpha = 0.05

regression

  • We conclude that our maximum yield for the first forage cut is 4,103 kg ha dry forage biomass and that this is achievable when 1,468 GDD have accumulated, which appears to be after anthesis and before dough stage.