HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text Free Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (7) Citing Articles via Google Scholar Google Scholar Articles by Pringle, M. J. Articles by Lark, R. M. Search for Related Content PubMed Articles by Pringle, M. J. Articles by Lark, R. M. GeoRef GeoRef Citation Agricola Articles by Pringle, M. J. Articles by Lark, R. M. Related Collections Carbon Sequestration Other Models Spatial Distribution

Spatial Analysis of Model Error, Illustrated by Soil Carbon Dioxide Emissions

Rothamsted Research, Harpenden, Hertfordshire, AL5 2JQ, UK

View larger version (13K): [in a new window]   Fig. 1. The inter-block correlation between two hypothetical variables. In the first simulation, the structural correlation between the two variables increases with spatial scale, and the best pixel size for aggregation is approximately 80 units. In the second simulation, the structural correlation between the two variables is strongest at an intermediate spatial scale, and the best pixel size for aggregation is as small as possible.

View larger version (20K): [in a new window]   Fig. 2. Map of Great Britain, indicating the location of Silsoe, the site of the case study. Map provided by Ordnance Survey (www.ordnancesurvey.co.uk/).

View larger version (26K): [in a new window]   Fig. 3. Histograms and summary statistics of: (a) observed soil CO2 flux, Fco2o; (b) log. of observed soil CO2 flux, ln Fco2o; (c) predicted soil CO2 flux, Fco2p; (d) log. of predicted soil CO2 flux, ln Fco2p; (e) model error, ; and (f) error of the log. of the fluxes, ln.

View larger version (12K): [in a new window]   Fig. 4. (a) Log. of observed soil CO2 flux (ln Fco2o) vs. log. of predicted soil CO2 flux (ln Fco2p), and summary statistics; and (b) ln Fco2o and ln Fco2p by position on the transect. ME = mean error (observed – predicted), RMSE = root-mean-square-error, rPM = product-moment correlation, and rMVE = correlation based on a minimum-volume ellipsoid estimate of the covariance matrix (Rousseeuw, 1985).

View larger version (20K): [in a new window]   Fig. 5. (a) Auto-variogram of log. of observed soil CO2 flux (ln Fco2o), (b) auto-variogram of log. of predicted soil CO2 flux (ln Fco2p), and (c) cross-variogram of ln Fco2o and ln Fco2p. Solid lines are the fits from the linear model of coregionalization. Part (d) shows the codispersion function and structural correlations associated with (a) to (c).

View larger version (10K): [in a new window]   Fig. 6. (a) Log. of observed soil CO2 flux (ln Fco2o) vs. log. of predicted soil CO2 flux (ln Fco2p), and summary statistics, after removal of nugget variation; and (b) ln Fco2o and ln Fco2p by position on the transect, after removal of nugget variation. ME = mean error (observed – predicted), RMSE = root-mean-square-error.

View larger version (14K): [in a new window]   Fig. 7. (a) Variogram of model error (ln) with model estimated directly, and model implied from the linear model of coregionalization; and (b) model error cokriged over blocks the extent of each soil type.

View larger version (19K): [in a new window]   Fig. 8. Scatterplots of model error (ln) (ordinate) and (a) soil organic carbon (C) content, (b) soil clay content, and (c) soil volumetric water content (v).

View larger version (8K): [in a new window]   Fig. 9. The inter-block correlation of log. of observed soil CO2 flux (ln Fco2o) and log. of predicted soil CO2 flux (ln Fco2p). The best pixel size for aggregation is 10 to 20 m.