Watt,M.S. and Palmer,D.J.(2012): Use of regression kriging to develop a Carbon:Nitrogen ratio surface for New Zealand. Geoderma, 183-184, 49-57.

『ニュージーランドについての炭素:窒素比面を作るための回帰空間予測法の使用』


Abstract
 Although the ratio of Carbon:Nitrogen (C:N ratio) is a very important determinant of site fertility little research has characterised spatial variation in this variable at broad scales. Using an extensive national dataset (n=1573) the objectives if this research were to (i) develop a model of soil C:N ratio using regression kriging, (ii) characterise the functional form of relationships between C:N ratio and ancillary variables included in the model and (iii) use the final model to predict C:N ratio spatially across New Zealand. The final model accounted for 65% of the variance in the validation dataset (n=315) using five significant (P<0.01) driving variables and an isotrophic exponential model to account for the significant (P<0.001) spatial covariance in the data.
 C:N ratio was most sensitive to New Zealand soil order, followed by land cover, rainfall, air temperature and then soil depth. Total annual rainfall and mean annual air temperature were included in the model as continuous variables with positive and negative slopes, respectively. All other variables were included as categorical variables. Soil orders with the lowest C:N ratios were Semi-arid, Melanic and Recent Soils while Organic, Podzol and Raw Soils had the highest C:N ratio. C:N ratio ranged widely across land covers in the following order from lowest to highest C:N ratio (with least square means in brackets): high producing pasture (12.8)<horticulture/cropping (12.9)<low producing pasture (14.0)<exotic forest (16.3)<shrubland (19.0)<native forest (19.6). C:M ratio showed a significant but small increase with increasing soil depth from 0-5 to 5-10 cm.
 C:N ratio predicted from the model varied widely throughout New Zealand. Values of C:N ratio were highest in moderately to extremely wet environments in high altitude regions of the North Island, and the west coast of the South Island where land cover has a high proportion of exotic and indigenous forest. In contrast, relatively low values of C:N ratio were predicted in drier eastern regions of New Zealand that are dominated by pasture.

Keywords: C:N ratio; Regression kriging; Spatial modelling』

1. Introduction
2. Methods
 2.1. C:N ratio data
 2.2. Predictive variables included in the model
 2.3. Regression kriging model
 2.4. Data analysis
 2.5. Partial response functions and LSmeans for ancillary variables
 2.6. Spatial predictions of C:N ratio
3. Results
 3.1. Final model
 3.2. Model validation
 3.3. Regression model partial responses and least square means
 3.4. Projections
4. Discussion
Acknowledgements
Appendix 1. Comparison between soil orders of New Zealand Soil Classification (NZSC) and the nearest equivalent soil orders of Soil Taxonomy. General diagnostic features are also given for NZSC soil order
Appendix 2. Regression model coefficients (Eq. (4)) used to describe the natural logarithm of C:N ratio
References


戻る