Examining soil parent material influence over Douglas-fir stem growth response to fertilization: Taking advantage of information from spatiotemporally distributed experiments

Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco) in the Inland Northwest region of the USA are nitrogen (N) deficient; however stem growth responses to N fertilizers are unpredictable, which may be due to poor accounting of other limiting nutrients. Screening trial experiments, including potassium (K), sulfur (S), and boron (B) multiple nutrient treatments, have been conducted to learn about Douglas-fir nutritional status and fertilizer growth response. The data from the screening trial experiments were compiled to test whether the soil parent materials of the region could be used to predict nutritional status. Estimating effects of fertilizers and soil parent materials on Douglas-fir growth from compilations of such experiments, however, poses challenges and opportunity; experiments clustered in time and space introduce latent variables that drive between-site variation. We used a two-stage modeling approach to efficiently take advantage of the information in these data. First, we employed a mixed model approach to test the primary hypothesis of soil parent material influence upon stem growth response to fertilizer. As the second-stage to the analysis, the predicted random effects estimated from the mixed model were used as a response variable to test how strongly precipitation drives between-site variation. As expected, including the random site effect significantly improved the model fit of the growth model (Λ = 436.5, P < 0.0001). The full mixed model accounted for 85% of the variation in the growth data (R2 = 0.85) and revealed an interaction between fertilizer treatment and soil parent material class (P = 0.0179). Post hoc analysis suggested that Douglas-fir growing on loessal soils are not constrained by K, S, or B, but no general consistency was apparent with tephra or underlying geology. The second stage modeling suggested that winter precipitation explains variation in predicted random site effects (r2 = 0.23), and hence the growth difference, better than total precipitation. Also, the annual lag precipitation explains variation in predicted random effects comparably well (r2 = 0.22).

► Douglas-fir are nutrient deficient in the Inland Northwest United States.
► Growth data were compiled from multi-nutrient fertilizer screening experiments.
► We tested soil parent material influence over fertilizer growth response.
► Douglas-fir were not limited by potassium, sulfur, or boron on loessal soils.
► Winter and annual lag precipitation explain between-experiment growth differences.