It is generally assumed that increased map precision (ie map unit homogeneity) and map purity (map unit accuracy) should result from increasing the scale of mapping of the soil resource, since it should enable a more intricate breakdown of the landscape into landform facet based units. This study compared the predictive success of a 1:1 OK scale soil association map with the 1:25K and 1:1OOK scale soil landscape maps within the Birrigai area of the Paddy's river catchment, south west of Canberra, A.C.T. The 1:25K and the 1:100K scale soil landscape maps were also evaluated in a second larger evaluation area in the Paddy's river catchment which allowed more of the larger soil landscape map units to be evaluated. The 1:25K scale soil map was produced by another author for the A.C.T Government, and was surveyed at a substantially lower survey intensity than that for the 1:100K and 1:10K scale soil maps (ie only 0.05 observation sites / cm2 of published map). These maps were evaluated using a set of randomly located independent evaluation sites in each evaluation area, and from these calculating and comparing standard Marsman & de Gruijter(1986) measures of Map Purity. The strength of soil-landscape relationships within this catchment were determined from a Fixed One Way Analysis of Variance, and from more simplistic graphical comparisons of the means and standard deviations of the discrete soil data within these landform based map units. Soil-landscape relationships for the Nominal scale soil data (ie class type data) were evaluated by comparing the Marsman & de Gruijter(1986) Homogeneity index ratings among the soil map units. Intensive survey traverses were also carried out in selected soil landscapes to further evaluate the strength of soil landscapes present. The results revealed obvious improvements in map quality associated with increasing map scale from 1:100,000 to 1:10,000, and these included increases in the predictive success (Map Purity), reductions in the extent of map unit impurities, and planning advantages associated with having individual land facets delineated on the 1:10,000 scale map. The respectable purity ratings achieved by the 1:100,000 scale soil landscape map (ie average purity rating of 63%) was largely attributed to the flexibility of the "soil material" approach to soil landscape mapping. The relatively poor performance of the 1:25K consultancy soil landscape map demonstrated the fact that; any benefit gained from the improved intricacy in the representation of map unit delineation's with increased mapping scale, will be drastically reduced if it is not matched by an associated increase in the intensity of field investigations. Evaluations of the soil-landscape relationships found that the land facets of the Paddy's river catchment generally failed to delineate areas that were both uniform and unique in respect of their soil properties. Soil-landscape relationships were instead found to be quite complex, applying to only certain land facets, and in regards to only certain soil properties. Soil maps with units based on landsurface features were recommended on the basis of the importance of other landscape factors other than soils to land capability ratings, as well as on the useability of such maps. This study recommended the adoption of a " >2 detailed soil profile observations / land facet in each map unit " mapping standard to ensure a reasonable estimate of the variability and modal soil conditions present, as well as a reliable confirmation of the perceived soil-landscape relationships. The error usually associated with small scale mapping was effectively reduced by rapid ground truthing, involving driving along the major roads dissecting the map area and making brief observations of soil exposures on road batters, despite the bias of the road network making such mapping improvements uneven across the map. The major point to come from this study was the re-emphasising of the point that soil spatial variability has to be accepted as a "real landscape attribute" which needs to be accurately described and communicated to land users, and must not be considered as some sort of soil mapping failure. The fact that individual facets of the landscape rarely coincide with unique pockets of uniform and unique soils and soil properties must be considered simply an on the ground reality of nature, and not some mapping failure. It was thought that since other landscape factors (eg hillslope gradient) most often dominate the determination of land use suitability and capability, it is better to effectively describe the range and modal state of the soil conditions within such facets, then to attempt to extrapolate possible soil boundaries using geostatistical techniques which cut across such land facets, and may or may not correlate with real groupings of soil properties, depending on the spatial resolution of the soil variability distribution in the landscape. Even so the results of this investigation do put the validity of the physiographic terrain class mapping model as a predictor of soil traits under question, at least for the more complex landscape settings.

Submitted by Jorge Silva (jorgelmsilva@ufrrj.br) on 2018-08-29T20:27:02Z No. of bitstreams: 1 2015 - Claudia Cseko Nolasco de Carvalho.pdf: 7996736 bytes, checksum: 4342b7f850b677568441806a801d9047 (MD5) / Made available in DSpace on 2018-08-29T20:27:02Z (GMT). No. of bitstreams: 1 2015 - Claudia Cseko Nolasco de Carvalho.pdf: 7996736 bytes, checksum: 4342b7f850b677568441806a801d9047 (MD5) Previous issue date: 2015-05-27 / Sustainable development requires more detailed soil maps with representative information of local reality, helping to conciliate increasing agriculture production and environmental impacts reduction. It is a hard task to achieve considering the Brazil territorial extension and the absence of governmental programs and resources. The Digital Soil Mapping (DSM), through methodologies implementation, has been consolidated as support technique for conventional mapping. Using DSM techniques, we aimed to evaluate a methodological procedure to generate soil maps with refined scale by fuzzy inference, and to map simpler units, which means more adequate areas for the rural settlement development plans. We carried out the study on an area at the microregion of Mata of Alagoas State and the following attributes were integrated by pedologist knowledge: elevation, slope, curvature and profile, topographic wetness index ? derived from a Digital Elevation Model (built by aerophotogrammetric restitution at scale 1:8 000), and geological data at scale 1:50 000. In a first moment, we integrate the thematic layers of 5m resolution pixel related to the attributes. Possibility maps of predefined classes were generate based on rules established by pedological expertise using fuzzy inference. Then, we integrate the evidence maps of classes by weighted mean inference to produce the soils digital map. The predicted digital soil map was opposed to a map elaborated traditionally, and validated by field observation of transects and pre-existent profiles. The comparison between the digital soil map and the conventional indicated a accuracy of 81.2% between the predicted classes of the digital map and the ones on the conventional map units (described on the legend, however not spatialized). Confronting the predicted classes of the digital soil map and the ones identified on field, the observed accuracy was of 74.65%. The Kappa index found was very good, with 0.696 values for the comparison between the digital map and the conventional map and 0.705 between the digital map and field observations. The results demonstrate that legacy maps data may, through the pedologists knowledge and more detailed attribute data, aggregate information and produce more detailed soil maps. Considering the concepts of minimum mappable area, the definitions of pixel sizes suggested for the digital cartographic scale, and the rules for soil survey, the applied methodology allowed, on the used cartographic base, the amplification of the mapping scale from 1:25 000 to 1:15 000. / O desenvolvimento sustent?vel requer mapas de solo mais detalhados com informa??es representativas da realidade local que permitem conciliar aumento da produ??o agropecu?ria com minimiza??o de impactos ambientais. Essa ? uma tarefa dif?cil considerando a extens?o territorial do Brasil e a falta de programas e recursos governamentais. O Mapeamento Digital de Solos-MDS, atrav?s da implementa??o de metodologias, vem se consolidando como t?cnica de apoio ao mapeamento convencional. Usando t?cnicas de MDS, esse trabalho objetivou avaliar um procedimento metodol?gico para gerar mapa de solos com refinamento de escala usando infer?ncia fuzzy para mapear unidades mais simples, portanto mais adequadas aos planos de desenvolvimento dos assentamentos rurais. A ?rea escolhida para teste est? localizada na microrregi?o da Mata do Estado de Alagoas. Para execu??o do trabalho os atributos do terreno eleva??o, declividade, curvatura em perfil, plano de curvatura e ?ndice de umidade topogr?fica derivados de Modelo Digital de Eleva??o-MDE (gerado por restitui??o aerofotogram?trica na escala 1:8.000) e dados geol?gicos em escala 1:50:000, foram integrados atrav?s do conhecimento de ped?logos. Os planos de informa??es, com pixel de 5 m, a partir de regras estabelecidas pela expertise de ped?logos foram fuzzificados e integrados para gerar os mapas de possibilidades de ocorr?ncia para as classes de solo predefinidas. O mapa digital de solos foi produzido pela integra??o dos mapas de possibilidades das classes por infer?ncia m?dia ponderada. O mapa digital de solos predito foi confrontado com mapa elaborado por m?todo tradicional e validado por observa??es em campo ao longo de transectos e por perfis preexistentes. A an?lise comparativa entre o mapa digital de solos e o convencional obteve acur?cia de 81,2 % entre as classes preditas no mapa digital e as contidas nas unidades do mapa convencional (descritas na legenda, por?m n?o espacializadas). Confrontando as classes preditas no mapa digital e as identificadas no campo a acur?cia encontrada foi de 74,65%. A aferi??o das classes pelo ?ndice de Kappa foi considerada muito boa com valores de 0,696 para a compara??o entre o mapa digital versos mapa convencional e 0,705 entre o mapa digital e as observa??es de campo. A pesquisa mostrou que dados contidos em mapas legados de solo podem atrav?s do conhecimento de ped?logos e de dados detalhados de atributos do terreno agregar informa??es e produzir mapa digital de solo com amplia??o da escala e que neste trabalho ? compat?vel com a de levantamentos detalhados (1:15.000).