『Abstract
　We investigated soil carbon (C) and nitrogen (N) distribution and developed a model, using readily available geospatial data, to predict that distribution across a mountainous, semi-arid, watershed in southwestern Idaho (USA). Soil core samples were collected and analyzed from 133 locations at 6 depths (n = 798), revealing that aspect dramatically influences the distribution of C and N, with north-facing slopes exhibiting up to 5 times more C and N than adjacent south-facing aspects. These differences are superimposed upon an elevation precipitation) gradient, with soil C and N contents increasing by nearly a factor of 10 from the bottom (1100 m elevation) to the top (1900 m elevation) of the watershed. Among the variables evaluated, vegetation cover, as represented by a Normalized Difference Vegetation Index(NDVI), is the strongest, positively correlated, predictor of C; potential insolation (incoming solar radiation) is a strong, negatively correlated, secondary predictor. Approximately 62％ (as R²) of the variance in the C data is explained using NDVI and potential insolation, compared with an R² of 0.54 for a model using NDVI alone. Soil N is similarly correlated to NDVI and insolation. We hypothesized that the correlation between soil C and N and slope, aspect and elevation reflect, in part, the inhibiting influence of insolation on semi-arid ecosystem productivity via water limitation. Based on these identified relationships, two modeling techniques (multiple linear regression and cokriging) were applied to predict the spatial distribution of soil V and N across the watershed. Both methods produce similar distributions, successfully capturing observed trends with aspect and elevation. This easily applied approach may be applicable to other semi-arid systems at larger scales.

Keywords: Soil carbon; Insolation; NDVI; Statistical model; Semi-arid』

1. Introduction
2. Material and methods
　2.1. Study sites and land use history
　2.2. Sample collection and laboratory analysis methods
　2.3. Predictor variables
　　2.3.1. Elevation and precipitation
　　2.3.2. Potential insolation
　　2.3.3. NDVI/vegetation
　2.4. Statistical analysis and modeling
3. Results
　3.1. General trends in spatial distribution of C and N
　3.2. Predictors of soil C and N
　3.3. Modeled spatial distribution of C and N
4. Discussion
　4.1. Underlying controls on soil carbon distribution
　4.2. Importance of disturbance
　4.3. Comparison and appropriateness of modeling approaches
　4.4. Future modeling efforts
　4.5. Implications for soil carbon management and climate change
5. Conclusions
Acknowledgements
References