Supplementary Material Title: Seasonality of hydraulic-redistribution by trees to grasses and changes in water-source use that change tree-grass interactions Running Head: Hydraulic-redistribution by trees to grasses and changes in their water-source use AUTHORS Priyadarshini K V R1ab, Herbert. H. T Prins1c, Steven de Bie1d, Ignas. M. A. Heitkönig1e, Stephan Woodborne2 , Gerrit Gort 3, Kevin Kirkman4, Fulco Ludwig5, Todd E. Dawson6, Hans de Kroon7 AUTHOR AFFILIATIONS 1 Resource Ecology Group, Wageningen University, Wageningen, The Netherlands, priya.kvr@gmail.com, c herbert.prins@wur.nl, dsteven.debie@gemeynt.nl , e ignas.heitkonig@wur.nl a 2 iThemba Laboratories, Private Bag 11, WITS, 2050, South Africa, Swoodborne@tlabs.ac.za 3 Biometris, Wageningen University, The Netherlands, gerrit.gort@wur.nl 4 University of KwaZulu-Natal, Scottsville, 3209, Pietermaritzburg, South Africa, Kirkmank@ukzn.ac.za 5 Earth system science and Climate change group, Wageningen University, Wageningen, The Netherlands, Fulco.ludwig@wur.nl 6 Departments of Integrative Biology and Environmental Science, Policy & Management, University of California at Berkeley, Berkeley, CA, 94720, USA, tdawson@berkeley.edu 7 Institute for Water and Wetland Research, Department of Experimental Plant Ecology, Radboud University, Nijmegen, The Netherlands, h.dekroon@science.ru.nl b Corresponding author 1 Appendix 1 Data transformations and the linear mixed model p We used Box-Cox transformation functions f(OR) = (OR –1)/p for hydrogen and for oxygen, with OR the RSAMPLE/RSTANDARD ratios, where the power ‘p’ was optimized using PROC TRANSREG of SAS by fitting fixed effects models to the transformed odds ratio. The power p = –10 was chosen for hydrogen, and p = –75 for oxygen. The transformation functions for oxygen and hydrogen were small for the majority of the observations (429 out 2 of 440 δ H values were in the range –75 to 75, corresponding to the Odd Ratio (OR) values 18 0.925 to 1.075, yielding transformed values ranging from –0.12 to 0.051; the range for δ O was –13 to 25, corresponding to OR values 0.987 to 1.040, yielding transformed values within the range –0.019 to 0.015). In the mixed linear model, random effects were allowed for tree pairs, trees within pairs, and tree/grass observational units, on which repeated measurements were made. Of these, only random effects of tree pairs were found. The residual variances were different in some subgroups even after data transformation. Variances for tree and grass were found to be different for the dry-wet transition, wet-dry transition and dry seasons. Measurements after spiking were more variable than pre-spike measurements or measurements on control trees. Decisions about this heterogeneity were made by comparisons of models using Akaike’s Information Criterion. No repeated measurements correlation structure (like auto-regression) was found. The test statistics for the different parts of the analysis were obtained using CONTRAST statements within the larger overall mixed model analysis. For oxygen, a simpler model than that used for hydrogen could be used since an effect of group (pre-spike and control only) was neither expected, nor found. The fixed part of the model contained main effects and interactions of species, plant functional type, and season. In the random part of the model, random effects of tree pairs, trees within pairs, and tree/grass 2 observational units were allowed to take care of the experimental design, but none of these appeared to be important. Residual variances were different for different grass/tree and season combinations. The largest variance was found in the dry-wet transition (Nov 2010), and smallest variance in the wet 2011 (Jan 2011) period. Trees had slightly larger variance than grass. 3 Appendix 2 14 Vachellia nilotica Topsoil Subsoil soil moisture content (%) 12 10 8 6 4 2 0 Philenoptera violacea Topsoil Subsoil soil moisture content (%) 14 12 10 8 6 4 2 0 14 Terminalia sericea Topsoil soil moisture content (%) 12 Subsoil 10 8 6 4 2 0 Monthly variation in average top and sub-soil volumetric moisture content (%) under three savanna tree species. The dotted arrow indicates start of dry top-soil and the long-dashed arrow depicts the start of dry sub-soil (2.5m) during the study period (Jan 2010 to Jan 2011). The black solid arrow represents sub-soil water recharge (Jan 2010 data is not shown since the loggers were not set up during that time and Oct data is not available due to the loss of loggers in September 2010) 4 Appendix 3 2 Table 1: Summary of ANOVA for the differences in δ H between pre-spiked and control groups. The F-tests for the differences were extracted from a larger mixed linear model. The analysed responses are Box-Cox transformed Odds ratio’s. Effects Pre-spike Vs. Control F P DF Tree species 0.64 0.530 2, 286 Plant functional type 1.96 0.162 1, 286 Season 6.17 0.0002* 3, 286 Species x Plant functional type 0.01 0.990 2, 286 Species x Season 1.83 0.071 8,286 Plant functional type x Season 3.08 0.016* 4, 286 Species x Plant functional type x Season 1.06 0.392 8, 286 *P<0.05 2 Table 2: Summary of comparisons of differences in least square means of δ H as obtained from the repeated measures mixed linear model. The term “ftype” indicates the plant functional type. Group comparisons of pre-spike Vs. control P T Wet 10 (season) Differences between LS Means (± 1SEM) 0.002 (± 0.003) 0.60 0.59 Wet-dry transition (season) -0.003(±0.004) 0.45 -0.75 Dry (season) 0.020 (±0.004) <0.0001* 4.84 Dry-wet transition (season) 0.006 (±0.003) 0.043* 2.03 Wet 11 (season) 0.006 (±0.002) 0.025* 2.26 Wet 10 (ftype x season) Grass -0.002 (±0.003) 0.555 -0.59 Wet-dry transition (ftype x season) Grass -0.014 (±0.001) <0.0001* -9.91 Dry (ftype x season) Grass 0.030 (±0.008) <0.0001* 4.00 Dry-wet transition (ftype x season) Grass 0.004 (±0.004) 0.268 1.11 Wet 11 (ftype x season) Grass 0.002 (±0.003) 0.410 0.83 Wet 10 (ftype x season) Tree 0.005(±0.005) 0.305 1.03 Wet-dry transition (ftype x season) Tree 0.007 (±0.008) 0.360 0.92 Dry (ftype x season) Tree 0.010 (±0.004) 0.004* 2.87 Dry-wet transition (ftype x season) Tree 0.009 (±0.005) 0.087 1.71 Wet 11 (ftype x season) Tree 0.010 (±0.005) 0.034* 2.13 *P<0.05 5