Forage Quality Graphs and Stats
Jonathan Spiess
June 30, 2021
Weight Gains
Questions/Hypothesis
Are animals gaining weight each year? -I expect average daily gain to be positive for each livestock class each year
Do weight gains differ between years for any livetock class? -I do not expect years to differ despite variable precipitation
Results
Weights for each year and livestock class are greater than zero.
Weight gains did differ by year for each livestock class, but did not follow the precipitation gradient for the study duration. (2017-drought, 2018-normal, 2019-2x normal, 2020-drought)
For ewes, 2018 had the highest daily gains across all years, and gains in 2020 were lower than the other years with patch-burn grazing.
For cows, 2018 had the highest daily gains across all years, and there was no difference between other years.
For calves, 2018 had the lowest daily gains across all years, and there was no difference between other years.
CI Calc
##
## Simultaneous Tests for General Linear Hypotheses
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecEwe,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Year2016 == 0 0.257569 0.008853 29.09 <2e-16 ***
## Year2017 == 0 0.289330 0.009047 31.98 <2e-16 ***
## Year2018 == 0 0.347224 0.009021 38.49 <2e-16 ***
## Year2019 == 0 0.285604 0.009192 31.07 <2e-16 ***
## Year2020 == 0 0.230511 0.009144 25.21 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Confidence Intervals
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecEwe,
## REML = FALSE)
##
## Quantile = 2.5676
## 95% family-wise confidence level
##
##
## Linear Hypotheses:
## Estimate lwr upr
## Year2016 == 0 0.2576 0.2348 0.2803
## Year2017 == 0 0.2893 0.2661 0.3126
## Year2018 == 0 0.3472 0.3241 0.3704
## Year2019 == 0 0.2856 0.2620 0.3092
## Year2020 == 0 0.2305 0.2070 0.2540##
## Simultaneous Tests for General Linear Hypotheses
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecCow,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Year2016 == 0 0.64684 0.08892 7.275 1.74e-12 ***
## Year2017 == 0 0.45185 0.08955 5.046 2.26e-06 ***
## Year2018 == 0 1.28594 0.08812 14.593 < 2e-16 ***
## Year2019 == 0 0.62724 0.09316 6.733 8.30e-11 ***
## Year2020 == 0 0.61921 0.09363 6.614 1.87e-10 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Confidence Intervals
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecCow,
## REML = FALSE)
##
## Quantile = 2.569
## 95% family-wise confidence level
##
##
## Linear Hypotheses:
## Estimate lwr upr
## Year2016 == 0 0.6468 0.4184 0.8753
## Year2017 == 0 0.4519 0.2218 0.6819
## Year2018 == 0 1.2859 1.0596 1.5123
## Year2019 == 0 0.6272 0.3879 0.8666
## Year2020 == 0 0.6192 0.3787 0.8597##
## Simultaneous Tests for General Linear Hypotheses
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecCalf,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Year2016 == 0 2.98717 0.04546 65.72 <2e-16 ***
## Year2017 == 0 3.09392 0.04520 68.44 <2e-16 ***
## Year2018 == 0 2.64443 0.04505 58.70 <2e-16 ***
## Year2019 == 0 2.96073 0.04691 63.11 <2e-16 ***
## Year2020 == 0 2.97068 0.04751 62.52 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Confidence Intervals
##
## Fit: lmer(formula = ADG ~ Year + 0 + (1 | Pasture1/Year), data = HrecCalf,
## REML = FALSE)
##
## Quantile = 2.569
## 95% family-wise confidence level
##
##
## Linear Hypotheses:
## Estimate lwr upr
## Year2016 == 0 2.9872 2.8704 3.1039
## Year2017 == 0 3.0939 2.9778 3.2100
## Year2018 == 0 2.6444 2.5287 2.7602
## Year2019 == 0 2.9607 2.8402 3.0813
## Year2020 == 0 2.9707 2.8486 3.0927##
## Simultaneous Tests for General Linear Hypotheses
##
## Fit: lmer(formula = ADG ~ Type + 0 + (1 | Pasture1/Year), data = HrecAdgRaw,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## TypeEwe == 0 0.28201 0.01989 14.18 <2e-16 ***
## TypeCalf == 0 2.92754 0.02411 121.42 <2e-16 ***
## TypeCow == 0 0.73571 0.02389 30.80 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Confidence Intervals
##
## Fit: lmer(formula = ADG ~ Type + 0 + (1 | Pasture1/Year), data = HrecAdgRaw,
## REML = FALSE)
##
## Quantile = 2.3642
## 95% family-wise confidence level
##
##
## Linear Hypotheses:
## Estimate lwr upr
## TypeEwe == 0 0.2820 0.2350 0.3290
## TypeCalf == 0 2.9275 2.8705 2.9845
## TypeCow == 0 0.7357 0.6792 0.7922Year Test
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## EweY 8 -3365.31 0.00 1 1 1690.68
## EweNull 4 -3343.68 21.63 0 1 1675.85## Data: HrecEwe
## Models:
## EweNull: ADG ~ 1 + (1 | Pasture1/Year)
## EweY: ADG ~ Year + (1 | Pasture1/Year)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## EweNull 4 -3343.7 -3320.3 1675.8 -3351.7
## EweY 8 -3365.4 -3318.5 1690.7 -3381.4 29.672 4 5.707e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADG ~ Year + (1 | Pasture1/Year), data = HrecEwe,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## 2017 - 2016 == 0 0.031761 0.012082 2.629 0.0651 .
## 2018 - 2016 == 0 0.089655 0.012063 7.432 <0.001 ***
## 2019 - 2016 == 0 0.028036 0.012192 2.300 0.1448
## 2020 - 2016 == 0 -0.027057 0.012155 -2.226 0.1700
## 2018 - 2017 == 0 0.057894 0.012205 4.743 <0.001 ***
## 2019 - 2017 == 0 -0.003725 0.012333 -0.302 0.9982
## 2020 - 2017 == 0 -0.058818 0.012296 -4.783 <0.001 ***
## 2019 - 2018 == 0 -0.061619 0.012314 -5.004 <0.001 ***
## 2020 - 2018 == 0 -0.116712 0.012277 -9.506 <0.001 ***
## 2020 - 2019 == 0 -0.055093 0.012404 -4.442 <0.001 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## CalfY 8 144.11 0.00 1 1 -63.84
## CalfNull 4 158.83 14.73 0 1 -75.36## Data: HrecCalf
## Models:
## CalfNull: ADG ~ 1 + (1 | Pasture1/Year)
## CalfY: ADG ~ Year + (1 | Pasture1/Year)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## CalfNull 4 158.72 174.14 -75.359 150.72
## CalfY 8 143.69 174.53 -63.842 127.69 23.033 4 0.0001247 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADG ~ Year + (1 | Pasture1/Year), data = HrecCalf,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## 2017 - 2016 == 0 0.106754 0.064000 1.668 0.454
## 2018 - 2016 == 0 -0.342738 0.063891 -5.364 <1e-04 ***
## 2019 - 2016 == 0 -0.026437 0.065221 -0.405 0.994
## 2020 - 2016 == 0 -0.016490 0.065654 -0.251 0.999
## 2018 - 2017 == 0 -0.449493 0.063711 -7.055 <1e-04 ***
## 2019 - 2017 == 0 -0.133192 0.065044 -2.048 0.243
## 2020 - 2017 == 0 -0.123244 0.065478 -1.882 0.327
## 2019 - 2018 == 0 0.316301 0.064937 4.871 <1e-04 ***
## 2020 - 2018 == 0 0.326249 0.065371 4.991 <1e-04 ***
## 2020 - 2019 == 0 0.009948 0.066671 0.149 1.000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## CowY 8 707.60 0.00 1 1 -345.60
## CowNull 4 721.47 13.87 0 1 -356.68## Data: HrecCow
## Models:
## CowNull: ADG ~ 1 + (1 | Pasture1/Year)
## CowY: ADG ~ Year + (1 | Pasture1/Year)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## CowNull 4 721.36 736.98 -356.68 713.36
## CowY 8 707.19 738.44 -345.60 691.19 22.166 4 0.0001857 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADG ~ Year + (1 | Pasture1/Year), data = HrecCow,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## 2017 - 2016 == 0 -0.194985 0.126192 -1.545 0.533
## 2018 - 2016 == 0 0.639104 0.125185 5.105 <1e-04 ***
## 2019 - 2016 == 0 -0.019595 0.128780 -0.152 1.000
## 2020 - 2016 == 0 -0.027623 0.129119 -0.214 1.000
## 2018 - 2017 == 0 0.834089 0.125633 6.639 <1e-04 ***
## 2019 - 2017 == 0 0.175390 0.129216 1.357 0.655
## 2020 - 2017 == 0 0.167361 0.129553 1.292 0.696
## 2019 - 2018 == 0 -0.658699 0.128233 -5.137 <1e-04 ***
## 2020 - 2018 == 0 -0.666728 0.128573 -5.186 <1e-04 ***
## 2020 - 2019 == 0 -0.008029 0.132076 -0.061 1.000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)
ADG Graph Stack
cowplot::plot_grid( CowPlot,CalfPlot, EwePlot, nrow = 3 )
NIR Ordination
Questions/Hypotheses
How do the measured forage quality parameters relate to each other? - I expected protein to be dissimilar from fibers and lignin
Which treatment and environmental variables influence the ordination? - I expected differences within TSF levels and between months - I expected biomass to be more associated with fibers and lignin - I expected selection index and moisture content to be more associated with protein -
Results Summary
The first axis explains 71% of the variation.
Moisture, Month, TSF, Selection Index and Biomass are significant environmental variables for the forage quality data.
Available biomass vector is more associated with fibers. Selection Index vector is more associated with protein. Moisture vector is opposite of lignin.
TSF: Pairwise factorfit says they are all different
Pretty clear month gradient
Ordination Setup
## importance.MDS1 importance.MDS2 importance.MDS3
## Eigenvalue 1.19 0.47 0.01
## Proportion Explained 0.71 0.28 0.01
## Cumulative Proportion 0.71 0.99 1.00##
## ***VECTORS
##
## MDS1 MDS2 r2 Pr(>r)
## Moisture 0.96763 -0.25239 0.6335 0.002 **
## Powell 0.88703 0.46171 0.1246 0.002 **
## KgHa -0.90506 -0.42529 0.2003 0.002 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Blocks: strata
## Permutation: free
## Number of permutations: 499
##
## ***FACTORS:
##
## Centroids:
## MDS1 MDS2
## TSFRB 0.1109 0.0322
## TSF1to2 -0.0056 -0.0606
## TSF2to3 -0.0730 -0.1032
## TSF3plus -0.1284 -0.0443
## TSFNYB -0.0237 0.0618
## ESDClayey 0.0073 0.0227
## ESDLoamy 0.0148 -0.0202
## ESDSaline Lowland -0.0224 -0.0120
## ESDSandy 0.0096 0.0063
## ESDThin Claypan -0.0182 -0.0827
## MonthMay 0.0769 -0.0019
## MonthJune 0.2541 -0.0566
## MonthJuly 0.0098 -0.0343
## MonthAugust -0.0897 0.0052
## MonthSeptember -0.1193 0.0717
## TreatmentCattle 0.0217 0.0543
## TreatmentSheep -0.0252 -0.0631
##
## Goodness of fit:
## r2 Pr(>r)
## TSF 0.1025 0.002 **
## ESD 0.0063 0.512
## Month 0.2360 0.002 **
## Treatment 0.0451 1.000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Blocks: strata
## Permutation: free
## Number of permutations: 499Ordinations
Figure: NIR Ordi + Vectors
Figure: NIR TSF Ordi
##
## Pairwise comparisons using factor fitting to an ordination
##
## data: NIR.cap3 by NIREnv_2$TSF
## 999 permutations
##
## RB 1to2 2to3 3plus
## 1to2 0.0012 - - -
## 2to3 0.0012 0.0056 - -
## 3plus 0.0012 0.0012 0.0120 -
## NYB 0.0012 0.0012 0.0012 0.0012
##
## P value adjustment method: fdr## Scale for 'colour' is already present. Adding another scale for 'colour',
## which will replace the existing scale.
Figure: NIR Month Ordi
Available Biomass (Kg/Ha)
Hypothesis
Recently burned patches will have lower available biomass than other patches, especially not yet burned and 3 years since fire.
Results
The TSF*Grazer type model was the best overall model, and the TSF single term model was also better than the null.
For cattle pastures, the recently burned patches had lower available biomass than all other patches across years and grazing seasons. There were no other differences between other patches.
For sheep pastures, the recently burned patches had lower available biomass than all other patches across years and grazing seasons. The not yet burned patches had lower available biomass than patches 1-2, 2-3, and 3-4 years since fire. Patches with 2-3 and 3-4 years since fire had higher available biomass than patches with 1-2 years since fire.
Figure: Biomass Overall
Biomass Model Results
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## KgHaInt 12 3597.58 0.00 1 1 -1786.71
## KgHaT 8 3644.98 47.40 0 1 -1814.46
## KgHaTG 9 3646.36 48.78 0 1 -1814.13
## KgHaNull 5 4002.96 405.37 0 1 -1996.46
## KgHaG 6 4004.32 406.74 0 1 -1996.14## Data: HRECNIR
## Models:
## KgHaT: log(KgHa + 1) ~ TSF + (1 | Location/Year/Month)
## KgHaInt: log(KgHa + 1) ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## KgHaT 8 3644.9 3689.7 -1814.5 3628.9
## KgHaInt 12 3597.4 3664.6 -1786.7 3573.4 55.486 4 2.569e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## KgHaNull: log(KgHa + 1) ~ 1 + (1 | Location/Year/Month)
## KgHaT: log(KgHa + 1) ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## KgHaNull 5 4002.9 4030.9 -1996.5 3992.9
## KgHaT 8 3644.9 3689.7 -1814.5 3628.9 364.02 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## KgHaNull: log(KgHa + 1) ~ 1 + (1 | Location/Year/Month)
## KgHaG: log(KgHa + 1) ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## KgHaNull 5 4002.9 4030.9 -1996.5 3992.9
## KgHaG 6 4004.3 4037.9 -1996.1 3992.3 0.6517 1 0.4195##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = log(KgHa + 1) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 0.36618 0.04666 7.848 <0.001
## 3 Years Since Fire - Recently Burned == 0 0.45660 0.08241 5.540 <0.001
## Not Yet Burned - Recently Burned == 0 0.48509 0.04760 10.190 <0.001
## 3 Years Since Fire - Intermediate == 0 0.09042 0.07814 1.157 0.6407
## Not Yet Burned - Intermediate == 0 0.11891 0.05036 2.361 0.0789
## Not Yet Burned - 3 Years Since Fire == 0 0.02849 0.08793 0.324 0.9875
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0 .
## Not Yet Burned - 3 Years Since Fire == 0
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = log(KgHa + 1) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 0.82962 0.05170 16.047 <0.001
## 3 Years Since Fire - Recently Burned == 0 1.07080 0.08593 12.462 <0.001
## Not Yet Burned - Recently Burned == 0 0.56987 0.05445 10.466 <0.001
## 3 Years Since Fire - Intermediate == 0 0.24118 0.08150 2.959 0.0153
## Not Yet Burned - Intermediate == 0 -0.25976 0.05711 -4.548 <0.001
## Not Yet Burned - 3 Years Since Fire == 0 -0.50094 0.09349 -5.358 <0.001
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0 *
## Not Yet Burned - Intermediate == 0 ***
## Not Yet Burned - 3 Years Since Fire == 0 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)ADF
Hypothesis
Recently burned patches will have lower ADF than other patches, especially not yet burned and 3 years since fire.
Results
The TSF only, TSF + Grazer Type, and TSF*Grazer Type models were better than the null model. There was no difference between the top three models.
For cattle pastures, the recently burned patches had lower ADF than all other patches across years and grazing seasons. Patches with 3-4 years since fire also had higher ADF than not yet burned and patches with 1-2 years since fire.
For Sheep pastures, the recently burned patches had lower ADF than all other patches across years and grazing seasons. Patches with 3-4 years since fire also had higher ADF than not yet burned and patches with 1-2 years since fire.
Figure: ADF Overall
ADF Overall Models
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## ADFTG 9 9425.69 0.00 0.49 0.49 -4703.80
## ADFT 8 9426.55 0.86 0.32 0.80 -4705.24
## ADFInt 12 9427.47 1.78 0.20 1.00 -4701.66
## ADFG 6 9769.77 344.08 0.00 1.00 -4878.86
## ADFNull 5 9770.57 344.88 0.00 1.00 -4880.27## Data: HRECNIR
## Models:
## ADFT: ADF ~ TSF + (1 | Location/Year/Month)
## ADFTG: ADF ~ TSF + Treatment + (1 | Location/Year/Month)
## ADFInt: ADF ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADFT 8 9426.5 9471.3 -4705.2 9410.5
## ADFTG 9 9425.6 9476.0 -4703.8 9407.6 2.8825 1 0.08955 .
## ADFInt 12 9427.3 9494.5 -4701.7 9403.3 4.2829 3 0.23249
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## ADFNull: ADF ~ 1 + (1 | Location/Year/Month)
## ADFT: ADF ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADFNull 5 9770.5 9798.5 -4880.3 9760.5
## ADFT 8 9426.5 9471.3 -4705.2 9410.5 350.06 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## ADFNull: ADF ~ 1 + (1 | Location/Year/Month)
## ADFG: ADF ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADFNull 5 9770.5 9798.5 -4880.3 9760.5
## ADFG 6 9769.7 9803.3 -4878.9 9757.7 2.8116 1 0.09359 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADF ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 2.4739 0.2199 11.248 <0.001
## 3 Years Since Fire - Recently Burned == 0 3.2084 0.3878 8.273 <0.001
## Not Yet Burned - Recently Burned == 0 2.1870 0.2242 9.755 <0.001
## 3 Years Since Fire - Intermediate == 0 0.7345 0.3681 1.996 0.1793
## Not Yet Burned - Intermediate == 0 -0.2869 0.2365 -1.213 0.6047
## Not Yet Burned - 3 Years Since Fire == 0 -1.0214 0.4128 -2.474 0.0597
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADF ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 2.7442 0.1994 13.759 < 0.001
## 3 Years Since Fire - Recently Burned == 0 3.1044 0.3306 9.391 < 0.001
## Not Yet Burned - Recently Burned == 0 1.8191 0.2095 8.685 < 0.001
## 3 Years Since Fire - Intermediate == 0 0.3602 0.3140 1.147 0.64887
## Not Yet Burned - Intermediate == 0 -0.9251 0.2187 -4.231 < 0.001
## Not Yet Burned - 3 Years Since Fire == 0 -1.2853 0.3579 -3.591 0.00153
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0 ***
## Not Yet Burned - 3 Years Since Fire == 0 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)NDF
Hypothesis
Recently burned patches will have lower NDF than other patches, especially not yet burned and 3 years since fire.
Results
The interaction term and the additive models were the top two models for NDF. The TSF only model was also better than the null model and Grazer Type only model, but was slighlty worse than the top two models.
For cattle pastures, the recently burned patches had lower NDF than all other patches across years and grazing seasons. There were no other significant differences between patches when looking across years and grazing seasons.
For Sheep pastures, the recently burned patches had lower NDF than all other patches across years and grazing seasons. Not yet burned patches also had higher NDF than patches with 1-2 and 2-3 years since fire.
Figure: NDF Overall
NDF Overall Models
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## NDFInt 12 11047.41 0.00 0.62 0.62 -5511.63
## NDFTG 9 11048.46 1.05 0.36 0.98 -5515.18
## NDFT 8 11054.19 6.78 0.02 1.00 -5519.06
## NDFG 6 11270.35 222.94 0.00 1.00 -5629.15
## NDFNull 5 11276.16 228.76 0.00 1.00 -5633.07## Data: HRECNIR
## Models:
## NDFT: NDF ~ TSF + (1 | Location/Year/Month)
## NDFTG: NDF ~ TSF + Treatment + (1 | Location/Year/Month)
## NDFInt: NDF ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## NDFT 8 11054 11099 -5519.1 11038
## NDFTG 9 11048 11099 -5515.2 11030 7.7477 1 0.005378 **
## NDFInt 12 11047 11114 -5511.6 11023 7.1163 3 0.068283 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## NDFNull: NDF ~ 1 + (1 | Location/Year/Month)
## NDFT: NDF ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## NDFNull 5 11276 11304 -5633.1 11266
## NDFT 8 11054 11099 -5519.1 11038 228.02 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## NDFNull: NDF ~ 1 + (1 | Location/Year/Month)
## NDFG: NDF ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## NDFNull 5 11276 11304 -5633.1 11266
## NDFG 6 11270 11304 -5629.2 11258 7.827 1 0.005147 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = NDF ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 3.2200 0.3647 8.830 <0.001
## 3 Years Since Fire - Recently Burned == 0 4.3084 0.6426 6.705 <0.001
## Not Yet Burned - Recently Burned == 0 2.6888 0.3715 7.237 <0.001
## 3 Years Since Fire - Intermediate == 0 1.0884 0.6101 1.784 0.2681
## Not Yet Burned - Intermediate == 0 -0.5312 0.3913 -1.358 0.5115
## Not Yet Burned - 3 Years Since Fire == 0 -1.6196 0.6832 -2.371 0.0771
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = NDF ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 3.2210 0.2592 12.425 <0.001
## 3 Years Since Fire - Recently Burned == 0 3.0297 0.4303 7.040 <0.001
## Not Yet Burned - Recently Burned == 0 1.8467 0.2727 6.771 <0.001
## 3 Years Since Fire - Intermediate == 0 -0.1913 0.4084 -0.468 0.9642
## Not Yet Burned - Intermediate == 0 -1.3743 0.2857 -4.811 <0.001
## Not Yet Burned - 3 Years Since Fire == 0 -1.1830 0.4674 -2.531 0.0516
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0 ***
## Not Yet Burned - 3 Years Since Fire == 0 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = NDF ~ Treatment + (1 | Location/Year/Month), data = HRECNIR,
## REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Sheep - Cattle == 0 2.1985 0.6242 3.522 0.000428 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)ADL
Hypothesis
Recently burned patches will have lower ADL than other patches, especially not yet burned and 3 years since fire.
Results
The TSF only, additive, and the interaction term models were better than the null model. The TSF only model was better than the interaction model, but not different than the additive model.
For cattle and sheep pastures, the recently burned patches had lower ADL than all other patches across years and grazing seasons. There were no other significant differences between patches when looking across years and grazing seasons.
Figure: ADL Overall
ADL Overall Models
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## ADLT 8 4098.97 0.00 0.59 0.59 -2041.45
## ADLTG 9 4100.52 1.55 0.27 0.87 -2041.21
## ADLInt 12 4101.96 2.99 0.13 1.00 -2038.90
## ADLNull 5 4361.23 262.26 0.00 1.00 -2175.60
## ADLG 6 4362.77 263.81 0.00 1.00 -2175.37## Data: HRECNIR
## Models:
## ADLT: ADL ~ TSF + (1 | Location/Year/Month)
## ADLTG: ADL ~ TSF + Treatment + (1 | Location/Year/Month)
## ADLInt: ADL ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADLT 8 4098.9 4143.7 -2041.5 4082.9
## ADLTG 9 4100.4 4150.8 -2041.2 4082.4 0.4677 1 0.4940
## ADLInt 12 4101.8 4169.0 -2038.9 4077.8 4.6252 3 0.2014## Data: HRECNIR
## Models:
## ADLNull: ADL ~ 1 + (1 | Location/Year/Month)
## ADLT: ADL ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADLNull 5 4361.2 4389.2 -2175.6 4351.2
## ADLT 8 4098.9 4143.7 -2041.5 4082.9 268.31 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## ADLNull: ADL ~ 1 + (1 | Location/Year/Month)
## ADLG: ADL ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## ADLNull 5 4361.2 4389.2 -2175.6 4351.2
## ADLG 6 4362.7 4396.3 -2175.4 4350.7 0.4677 1 0.494##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADL ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 0.47025 0.05749 8.179 <0.001
## 3 Years Since Fire - Recently Burned == 0 0.44913 0.10181 4.411 <0.001
## Not Yet Burned - Recently Burned == 0 0.59167 0.05876 10.069 <0.001
## 3 Years Since Fire - Intermediate == 0 -0.02111 0.09637 -0.219 0.996
## Not Yet Burned - Intermediate == 0 0.12142 0.06255 1.941 0.200
## Not Yet Burned - 3 Years Since Fire == 0 0.14254 0.10910 1.306 0.544
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = ADL ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 0.58539 0.04915 11.909 <1e-04
## 3 Years Since Fire - Recently Burned == 0 0.52043 0.08190 6.354 <1e-04
## Not Yet Burned - Recently Burned == 0 0.52925 0.05195 10.187 <1e-04
## 3 Years Since Fire - Intermediate == 0 -0.06497 0.07754 -0.838 0.829
## Not Yet Burned - Intermediate == 0 -0.05615 0.05486 -1.023 0.725
## Not Yet Burned - 3 Years Since Fire == 0 0.00882 0.08966 0.098 1.000
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)CP
Hypothesis
Recently burned patches will have higher CP than other patches, especially not yet burned and 3 years since fire.
Results
The TSF*Grazer Type model was the best for CP, and the TSF only and additive models were also better than the null.
For cattle pastures, CP was higher in recently burned and not yet burned patches than all other patches across years and grazing seasons. There were no other
For sheep pastures, recently burned patches had higher CP than all other patches across years and grazing seasons. Not yet burned patches had higher CP than patches with 1-2 and 3-4 years since fire, and patches with 1-2 years since fire had higher CP than those with 3-4 years since fire.
Figure: CP Overall
CP Overall Models
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## CPInt 12 -593.23 0.00 1 1 308.70
## CPTG 9 -540.77 52.46 0 1 279.43
## CPT 8 -536.85 56.38 0 1 276.46
## CPG 6 -170.42 422.82 0 1 91.23
## CPNull 5 -166.36 426.87 0 1 88.20## Data: HRECNIR
## Models:
## CPTG: log(CP + 1) ~ TSF + Treatment + (1 | Location/Year/Month)
## CPInt: log(CP + 1) ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## CPTG 9 -540.86 -490.49 279.43 -558.86
## CPInt 12 -593.39 -526.23 308.70 -617.39 58.53 3 1.211e-12 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## CPNull: log(CP + 1) ~ 1 + (1 | Location/Year/Month)
## CPT: log(CP + 1) ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## CPNull 5 -166.39 -138.41 88.196 -176.39
## CPT 8 -536.93 -492.16 276.464 -552.93 376.54 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## CPNull: log(CP + 1) ~ 1 + (1 | Location/Year/Month)
## CPG: log(CP + 1) ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## CPNull 5 -166.39 -138.41 88.196 -176.39
## CPG 6 -170.46 -136.88 91.229 -182.46 6.0667 1 0.01378 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = log(CP + 1) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 -0.224577 0.018439 -12.180 <1e-06
## 3 Years Since Fire - Recently Burned == 0 -0.399031 0.032592 -12.243 <1e-06
## Not Yet Burned - Recently Burned == 0 -0.002973 0.018823 -0.158 0.999
## 3 Years Since Fire - Intermediate == 0 -0.174454 0.030888 -5.648 <1e-06
## Not Yet Burned - Intermediate == 0 0.221604 0.019953 11.106 <1e-06
## Not Yet Burned - 3 Years Since Fire == 0 0.396058 0.034820 11.374 <1e-06
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0
## 3 Years Since Fire - Intermediate == 0 ***
## Not Yet Burned - Intermediate == 0 ***
## Not Yet Burned - 3 Years Since Fire == 0 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = log(CP + 1) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 -0.20893 0.01505 -13.879 <0.001
## 3 Years Since Fire - Recently Burned == 0 -0.26172 0.02503 -10.455 <0.001
## Not Yet Burned - Recently Burned == 0 -0.14974 0.01587 -9.435 <0.001
## 3 Years Since Fire - Intermediate == 0 -0.05279 0.02373 -2.225 0.1099
## Not Yet Burned - Intermediate == 0 0.05919 0.01669 3.547 0.0019
## Not Yet Burned - 3 Years Since Fire == 0 0.11198 0.02729 4.104 <0.001
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0
## Not Yet Burned - Intermediate == 0 **
## Not Yet Burned - 3 Years Since Fire == 0 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = log(CP + 1) ~ Treatment + (1 | Location/Year/Month),
## data = HRECNIR, REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Sheep - Cattle == 0 -0.12164 0.04409 -2.759 0.0058 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)Moisture Content
Hypothesis
Recently burned patches will have higher moisture content than other patches, especially not yet burned and 3 years since fire.
Results
The TSF only, additive, and interaction models were better than the null model.
For cattle pastures, the available forage in recently burned patches had higher moisutre content than all other patches. Moisture content in the not yet burned and 1-2 years since fire patches was higher than patches with 2-3 and 3-4 years since fire.
For sheep pastures, the available forage in recently burned patches had higher moisutre content than all other patches. Moisture content in the not yet burned and 1-2 years since fire patches was higher than patches with 3-4 years since fire.
Figure: Moisture Content
Moisture Overall Models
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## MoistureT 8 13883.84 0.00 0.57 0.57 -6933.88
## MoistureTG 9 13884.52 0.68 0.41 0.98 -6933.21
## MoistureInt 12 13890.38 6.54 0.02 1.00 -6933.11
## MoistureNull 5 14249.23 365.39 0.00 1.00 -7119.60
## MoistureG 6 14249.87 366.03 0.00 1.00 -7118.92## Data: HRECNIR
## Models:
## MoistureT: Moisture ~ TSF + (1 | Location/Year/Month)
## MoistureTG: Moisture ~ TSF + Treatment + (1 | Location/Year/Month)
## MoistureInt: Moisture ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## MoistureT 8 13884 13928 -6933.9 13868
## MoistureTG 9 13884 13935 -6933.2 13866 1.3397 1 0.2471
## MoistureInt 12 13890 13957 -6933.1 13866 0.2058 3 0.9766## Data: HRECNIR
## Models:
## MoistureNull: Moisture ~ 1 + (1 | Location/Year/Month)
## MoistureT: Moisture ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## MoistureNull 5 14249 14277 -7119.6 14239
## MoistureT 8 13884 13928 -6933.9 13868 371.43 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## MoistureNull: Moisture ~ 1 + (1 | Location/Year/Month)
## MoistureG: Moisture ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## MoistureNull 5 14249 14277 -7119.6 14239
## MoistureG 6 14250 14283 -7118.9 14238 1.3669 1 0.2423##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = Moisture ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Cattle"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 -7.2543 0.5759 -12.598 <0.001
## 3 Years Since Fire - Recently Burned == 0 -9.7309 1.0199 -9.541 <0.001
## Not Yet Burned - Recently Burned == 0 -6.8054 0.5886 -11.561 <0.001
## 3 Years Since Fire - Intermediate == 0 -2.4766 0.9653 -2.566 0.0470
## Not Yet Burned - Intermediate == 0 0.4490 0.6268 0.716 0.8848
## Not Yet Burned - 3 Years Since Fire == 0 2.9255 1.0932 2.676 0.0348
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0 *
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: lmer(formula = Moisture ~ TSF + (1 | Location/Year/Month), data = subset(HRECNIR,
## Treatment == "Sheep"), REML = FALSE)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 -7.4643 0.6640 -11.241 <0.001
## 3 Years Since Fire - Recently Burned == 0 -10.3286 1.1061 -9.337 <0.001
## Not Yet Burned - Recently Burned == 0 -6.7505 0.7016 -9.622 <0.001
## 3 Years Since Fire - Intermediate == 0 -2.8643 1.0474 -2.735 0.0295
## Not Yet Burned - Intermediate == 0 0.7138 0.7404 0.964 0.7603
## Not Yet Burned - 3 Years Since Fire == 0 3.5781 1.2101 2.957 0.0152
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0 *
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)Selection Index
Hypothesis
Recently burned patches will have a higher selection index than other patches, especially not yet burned and 3 years since fire.
Results
The TSF only, additive, and interaction models were better than the null model. There was no difference between the top three models.
For cattle pastures, recently burned patches had higher selection index values than all other patches across years and grazing seasons.
For sheep pastures, recently burned patches had higher selection index values than all other patches across years and grazing seasons. Not yet burned patches had lower selection index values than patches with 2-3 and 3-4 years since fire, and patches with 3-4 years since fire had higher selection index values than patches with 1-2 years since fire.
Figure: Selection Index Overall
Selection Index Overall Model
##
## Model selection based on AICc:
##
## K AICc Delta_AICc AICcWt Cum.Wt LL
## SIT 8 -1237.51 0.00 0.65 0.65 626.79
## SITG 9 -1235.50 2.02 0.24 0.89 626.79
## SIInt 12 -1234.03 3.48 0.11 1.00 629.09
## SINull 5 -1078.43 159.09 0.00 1.00 544.23
## SIG 6 -1076.41 161.10 0.00 1.00 544.23## Data: HRECNIR
## Models:
## SIT: (Powell + 0.001) ~ TSF + (1 | Location/Year/Month)
## SITG: (Powell + 0.001) ~ TSF + Treatment + (1 | Location/Year/Month)
## SIInt: (Powell + 0.001) ~ TSF * Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## SIT 8 -1237.6 -1192.8 626.79 -1253.6
## SITG 9 -1235.6 -1185.2 626.79 -1253.6 0.0027 1 0.9587
## SIInt 12 -1234.2 -1167.0 629.09 -1258.2 4.5999 3 0.2035## Data: HRECNIR
## Models:
## SINull: (Powell + 0.001) ~ 1 + (1 | Location/Year/Month)
## SIT: (Powell + 0.001) ~ TSF + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## SINull 5 -1078.5 -1050.5 544.23 -1088.5
## SIT 8 -1237.6 -1192.8 626.79 -1253.6 165.13 3 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Data: HRECNIR
## Models:
## SINull: (Powell + 0.001) ~ 1 + (1 | Location/Year/Month)
## SIG: (Powell + 0.001) ~ Treatment + (1 | Location/Year/Month)
## npar AIC BIC logLik deviance Chisq Df Pr(>Chisq)
## SINull 5 -1078.5 -1050.5 544.23 -1088.5
## SIG 6 -1076.5 -1042.9 544.23 -1088.5 1e-04 1 0.9916##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: glmer(formula = (Powell + 0.001) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Cattle"), family = Gamma)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 0.9576 0.1742 5.498 < 0.001
## 3 Years Since Fire - Recently Burned == 0 2.8323 0.8317 3.405 0.00258
## Not Yet Burned - Recently Burned == 0 0.7609 0.1506 5.052 < 0.001
## 3 Years Since Fire - Intermediate == 0 1.8747 0.8447 2.219 0.09775
## Not Yet Burned - Intermediate == 0 -0.1967 0.2109 -0.933 0.76005
## Not Yet Burned - 3 Years Since Fire == 0 -2.0714 0.8401 -2.466 0.05282
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 **
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0 .
## Not Yet Burned - Intermediate == 0
## Not Yet Burned - 3 Years Since Fire == 0 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)##
## Simultaneous Tests for General Linear Hypotheses
##
## Multiple Comparisons of Means: Tukey Contrasts
##
##
## Fit: glmer(formula = (Powell + 0.001) ~ TSF + (1 | Location/Year/Month),
## data = subset(HRECNIR, Treatment == "Sheep"), family = Gamma)
##
## Linear Hypotheses:
## Estimate Std. Error z value Pr(>|z|)
## Intermediate - Recently Burned == 0 1.4261 0.2059 6.927 < 0.001
## 3 Years Since Fire - Recently Burned == 0 3.8482 0.9435 4.079 < 0.001
## Not Yet Burned - Recently Burned == 0 0.6242 0.1282 4.870 < 0.001
## 3 Years Since Fire - Intermediate == 0 2.4221 0.9621 2.517 0.04490
## Not Yet Burned - Intermediate == 0 -0.8019 0.2280 -3.518 0.00184
## Not Yet Burned - 3 Years Since Fire == 0 -3.2239 0.9485 -3.399 0.00246
##
## Intermediate - Recently Burned == 0 ***
## 3 Years Since Fire - Recently Burned == 0 ***
## Not Yet Burned - Recently Burned == 0 ***
## 3 Years Since Fire - Intermediate == 0 *
## Not Yet Burned - Intermediate == 0 **
## Not Yet Burned - 3 Years Since Fire == 0 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## (Adjusted p values reported -- single-step method)