Mapping soil texture classes using field textuing, particle size distribution and local knowledge by both conventional and geostatisical methods

Abstract

Summary We investigated the utility of three interpolation techniques that ignored descriptive `soft' information and one that used it for mapping topsoil texture classes: re‐coding of soil map units within Geographical Information Systems (GIS), Thiessen polygons, and classifcation of probability vectors estimated by ordinary indicator kriging and simple indicator kriging with local prior means. The results were compared with texture maps based on a classifcation of kriged maps of particle size distribution. The methods were applied to two distinct regions, which represent large areas in rain‐fed rice ecosystems and irrigated rice ecosystems. The `hard' databases for both areas contained soil information needed for mapping at regional scales (1:100 000‐1:150 000). These data were complemented with 'soft' information derived from farmers and soil experts (Northeast Thailand) and soil maps (Nueva Ecija, Philippines). Interpolated maps agreed with the texture map based on interpolation of particle size distribution, and ®eld estimates of soil texture proved to be valuable surrogates for laboratory measurements of soil texture classes. The interpolation of categorical data such as soil texture classes allows for upgrading and increasing the resolution of maps in sparsely sampled regions by using simple field measurements. Validation using independent test sets showed that indicator kriging with local prior means performed best in the rain‐fed lands, whereas soft information did not improve the predictions in Nueva Ecija. Local knowledge in a formalized form was valuable in Northeast Thailand and the interpolated soil texture map for this area had an accuracy and resolution to support agronomic decisions at the village scale. The poor quality of the soil map and the fact that the gradually changing variability in young alluvial soils can be modelled with fewer data explained the lower accuracy of simple indicator kriging with local prior means in Nueva Ecija. Thiessen polygons performed well in the undulating rain‐fed lands but were not as reliable as indicator kriging in the gradually changing irrigated lands.

Uncertainties in estimating regional methane emissions from rice paddies due to data scarcity in the modeling approach

“…Figure 2 illustrates the data abundance of the five model variables. The data for soil sand content is a 10 km by 10 km raster data set constructed from soil profiles via spatial interpolation (Oberthür et al, 1999;Shi et al, 2004Shi et al, , 2006. Although a certain proportion of the immense spatial variation in soil properties may be lost after spatial interpolation (Goovaerts, 2001;van Bodegom et al, 2002), the gridded soil data are still the most detailed of the five model inputs.…”

Mapping clay content variation using electromagnetic induction techniques

“…Various geostatistical approaches have been employed to estimate spatial variation in topsoil clay content. These include, ordinary- (Voltz and Webster, 1990;Kalivas and Kollias, 1999), block- (Mapa and Kumaragamage, 1996), intrinsic random function of order k- (McBratney et al, 1991), indicator- (Oberthur et al, 1999), co- (Vauclin et al, 1983;Zhang et al, 1992), universal- (Odeh et al, 1995), regression- (Odeh and McBratney, 2000), and compositional-kriging (Odeh et al, 2003). Several studies have compared methods (Gallichand and Marcotte, 1993;Odeh et al, 1995) to map subsurface clay.…”

Upscaling Regional Emissions of Greenhouse Gases from Rice Cultivation: Methods and Sources of Uncertainty

“…For the detailed analysis in the Mun˜oz area we used an extensive data set of soil properties, crop management and yields compiled by Oberthu¨r et al (1996Oberthu¨r et al ( , 1999. Soil data in this area were based on a detailed, quasi-systematic soil survey, made in 1993.…”