Pregled bibliografske jedinice broj: 142922
Mapping soil properties from an existing national soil data set using freely available ancillary data
Mapping soil properties from an existing national soil data set using freely available ancillary data // Proceedings of the 17th World Congress of Soil Science / Van Meirvenne, Marc (ur.).
Bangkok: International Union of Soil Sciences, 2002. (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 142922 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Mapping soil properties from an existing national soil data set using freely available ancillary data
Autori
Hengl, Tomislav ; Rossiter, David G. ; Husnjak, Stjepan
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
Proceedings of the 17th World Congress of Soil Science
/ Van Meirvenne, Marc - Bangkok : International Union of Soil Sciences, 2002
Skup
17th World Congress of Soil Science
Mjesto i datum
Bangkok, Tajland, 14.08.2002. - 21.08.2002
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
soil survey; environmental regression; CLORPT; NOAA's AVHRR; Croatia
Sažetak
The paper demonstrates how NOAA's 1x1 km NDVI images, downloaded from the web, together with free coarse resolution elevation and climatic data, can be used to improve spatial detail of the Croatian national soil data-set consisting of 2198 profile observations. Two regression models were developed: to map pH (measured in H2O) and organic matter (%) in topsoil. Environmental predictors used were standard landform parameters (elevation, slope, curvature, aspect, wetness index), climatic data (rainfall, temperature) and vegetation components derived from the annual NDVI time series. Results show that these two soil properties can be mapped using the CLORPT approach with equal or better precision than with using the existing 1:50´000 soil class map and averaging the properties per soil mapping unit. While the precision of prediction for pH did not improve significantly (residual standard error: 0.60 versus 0.61), the precision for OM was considerably better when compared with the soil map (residual standard error: 2.81 versus 3.85). The models accounted for 54% (pH) and 66% (organic matter) of the total variation. The models coincided with Jenny's empirical models of soil variation. Principal components of the NDVI time series proved to be most significant predictors of the soil properties, representing general vegetation types and their dynamics. The prediction of pH was more difficult than the prediction of OM. The achieved coefficient of variation for pH was 16.8%, while the model for OM it was 10.8%.
Izvorni jezik
Engleski
