Mapping soil carbon, particle-size fractions, and water retention in tropical dry forest in Brazil

Abstract

-The objective of this work was to compare ordinary kriging with regression kriging to map soil properties at different depths in a tropical dry forest area in Brazil. The 11 soil properties evaluated were: organic carbon content and stock; bulk density; clay, sand, and silt contents; cation exchange capacity; pH; water retention at field capacity and at permanent wilting point; and available water. Samples were taken from 327 sites at 0.0-0.10, 0.10-0.20, and 0.20-0.40-m depths, in a tropical dry forest area of 102 km 2 .Stepwise linear regression models for particle-size fractions and water retention properties had the best fit. Relief and parent material covariates were selected in 31 of the 33 models (11 properties at three depths) and vegetation covariates in 29 models. Based on external validation, ordinary kriging obtained higher accuracy for 21 out of 33 property x depth combinations, indicating that the inclusion of a linear trend model before kriging does not necessarily improve predictions. Therefore, for similar studies, the geostatistical methods employed should be compared on a case-by-case basis.Index terms: caatinga, digital soil mapping, gamma radiometric survey, geostatistics, pedometrics. Mapeamento de carbono, frações granulométricas e água do solo em Floresta Tropical Seca no BrasilResumo -O objetivo deste trabalho foi comparar krigagem ordinária com regressão-krigagem para mapear atributos do solo, em diferentes profundidades, em área de Floresta Tropical Seca no Brasil. Os 11 atributos do solo avaliados foram: conteúdo e estoque de carbono orgânico; densidade do solo; conteúdos de argila, areia e silte; capacidade de troca catiônica; pH; retenção de água na capacidade de campo e no ponto de murcha permanente; e água disponível. As amostras foram retiradas de 327 locais a 0,0-0,10, 0,10-0,20 e 0,20-0,40 m de profundidade, em área de Floresta Tropical Seca de 102 km 2 . Os modelos de regressão linear "stepwise" tiveram o melhor ajuste para as frações granulométricas e as propriedades de retenção de água. Foram selecionadas covariáveis de relevo e material de origem em 31 dos 33 modelos (11 atributos em três profundidades) e de vegetação em 29 modelos. Com base na validação externa, a krigagem ordinária obteve maior acurácia para 21 das 33 combinações atributo vs. profundidade, o que é indicativo de que a inclusão de um modelo linear de tendência antes da krigagem não necessariamente melhora as predições. Portanto, para estudos semelhantes, os métodos geoestatísticos empregados devem ser comparados caso a caso.Termos para indexação: caatinga, mapeamento digital de solos, levantamento gamarradiométrico, geoestatística, pedometria.

No silver bullet for digital soil mapping: country-specific soil organic carbon estimates across Latin America

“…Despite the aforementioned limitations, across Latin America, there is an increasing availability of relevant SOC information across site-and country-specific regions (ReyesRojas et al, 2018;Vasques et al, 2016;Angelini et al, 2017;Samuel-Rosa et al, 2015;Angelini et al, 2016;Padarian et al, 2017), which could serve for validating and calibrating global SOC estimates. Thus, regional approaches considering multiple Latin American countries and SOC models could be a valuable resource to explain discrepancies between site-or country-specific and global SOC models.…”

Soil Carbon Stock and Particle Size Fractions in the Central Amazon Predicted from Remotely Sensed Relief, Multispectral and Radar Data

“…The clay content at the subsurface was better predicted by RK than OK, with the regression model selecting relief, multispectral, and radar variables. Although this was the only soil attribute better predicted by RK, the preference of OK over RK is not unusual, and has been reported elsewhere [37][38][39][40]. Thus, there is potential to improve soil attribute predictions by adding remote sensing covariates.…”

Sample planning for quantifying and mapping magnetic susceptibility, clay content, and base saturation using auxiliary information

“…In these protocols, previously acquired information such as geological, geomorphological, or pedological maps, or even property maps, on less detailed scales, can be used along with quantitative analyses (e.g. geostatistical analysis) to refine the mapping units and increase the understanding and reliability of the spatial patterns (Castrignanò et al, 2009;Cambule et al, 2013;Hengl et al, 2014Hengl et al, , 2017Vasques et al, 2016).…”

“…Measures such as the Shannon diversity index (Minasny et al, 2010) can be used as the first indication of soil pedodiversity intensity (variability) at large scales. For more detailed scales (regional or local), the study of incorporation of secondary information into geostatistical models (Castrignanò et al, 2009;Cambule et al, 2013;Vasques et al, 2016) and use of properties with potential for identifying the variation of soil formation processes (magnetic susceptibility-MS, electrical conductivity, and diffuse reflectance spectroscopy) (Bilgili et al, 2011;Siqueira et al, 2014;Mirzaeitalarposhti et al, 2017) represent an increasing research activity. However, the secondary information often used have quantitative (satellite information, electrical conductivity, and MS) (Benedetto et al, 2012) and non-qualitative or categorical nature (Castrignanò et al, 2009).…”

“…Hengl et al (2014) state that information on climate, lithology, biomass indexes and taxonomic units are the main covariates for modeling soil properties on a global scale. For regional and local scales, information on geomorphology, lithology, and pedology present a great potential (Anderson et al, 2003;Vasques et al, 2016). The hypothesis of this research is that the knowledge on soil formation factors (geology and landscape shape), often previously mapped and available at no charge, should be considered at the time of mathematical modeling.…”