Some reflections on acquisition, processing and analysis of statistical data in forest soils

Beyond the classical statistical approaches (determination of basic statistics, regression analysis, ANOVA, etc.) a new set of applications of different statistical techniques has increasingly gained relevance in the analysis, processing and interpretation of data concerning the characteristics of forest soils. This is possible to be seen in some of the recent publications in the context of Multivariate Statistics. These new methods require additional care that is not always included or refered in some approaches. In the particular case of geostatistical data applications it is necessary, besides to geo-reference all the data acquisition, to collect the samples in regular grids and in sufficient quantity so that the variograms can reflect the spatial distribution of soil properties in a representative manner. In the case of the great majority of Multivariate Statistics techniques (Principal Component Analysis, Correspondence Analysis, Cluster Analysis, etc.) despite the fact they do not require in most cases the assumption of normal distribution, they however need a proper and rigorous strategy for its utilization. In this work, some reflections about these methodologies and, in particular, about the main constraints that often occur during the information collecting process and about the various linking possibilities of these different techniques will be presented. At the end, illustrations of some particular cases of the applications of these statistical methods will also be presented.

Elaboração e estruturação de um modelo de rating interno

Temos vindo a assistir nos últimos anos a uma evolução no que respeita à avaliação do risco de crédito. As constantes alterações de regulamentação bancária, que resultam dos Acordos de Basileia, têm vindo a impor novas normas que condicionam a quantidade e a qualidade do risco de crédito que as Instituições de Crédito podem assumir nos seus balanços. É de grande importância as Instituições de Crédito avaliarem o risco de crédito, as garantias e o custo de capital, pois têm um impacto direto na sua gestão nomeadamente quanto à afetação de recursos e proteção contra perdas. Desta forma, pretende-se com o presente trabalho elaborar e estruturar um modelo de rating interno através de técnicas estatísticas, assim como identificar as variáveis estatisticamente relevantes no modelo considerado. Foi delineada uma metodologia de investigação mista, considerando na primeira parte do trabalho uma pesquisa qualitativa e na segunda parte uma abordagem quantitativa. Através da análise documental, fez-se uma abordagem dos conceitos teóricos e da regulamentação que serve de base ao presente trabalho. No estudo de caso, o modelo de rating interno foi desenvolvido utilizando a técnica estatística designada de regressão linear múltipla. A amostra considerada foi obtida através da base de dados SABI e é constituída por cem empresas solventes, situadas na zona de Paredes, num horizonte temporal de 2011-2013. A nossa análise baseou-se em três cenários, correspondendo cada cenário aos dados de cada ano (2011, 2012 e 2013). Para validar os pressupostos do modelo foram efetuados testes estatísticos de Durbin Watson e o teste de significância - F (ANOVA). Por fim, para obtermos a classificação de rating de cada variável foi aplicada a técnica dos percentis. Pela análise dos três cenários considerados, verificou-se que o cenário dois foi o que obteve maior coeficiente de determinação. Verificou-se ainda que as variáveis independentes, rácio de liquidez geral, grau de cobertura do ativo total pelo fundo de maneio e rácio de endividamento global são estatisticamente relevantes.

Soil sampling design and sampling techniques for soil properties monitoring – common difficulties related to forest soils

Mathematical models and statistical analysis are key instruments in soil science scientific research as they can describe and/or predict the current state of a soil system. These tools allow us to explore the behavior of soil related processes and properties as well as to generate new hypotheses for future experimentation. A good model and analysis of soil properties variations, that permit us to extract suitable conclusions and estimating spatially correlated variables at unsampled locations, is clearly dependent on the amount and quality of data and of the robustness techniques and estimators. On the other hand, the quality of data is obviously dependent from a competent data collection procedure and from a capable laboratory analytical work. Following the standard soil sampling protocols available, soil samples should be collected according to key points such as a convenient spatial scale, landscape homogeneity (or non-homogeneity), land color, soil texture, land slope, land solar exposition. Obtaining good quality data from forest soils is predictably expensive as it is labor intensive and demands many manpower and equipment both in field work and in laboratory analysis. Also, the sampling collection scheme that should be used on a data collection procedure in forest field is not simple to design as the sampling strategies chosen are strongly dependent on soil taxonomy. In fact, a sampling grid will not be able to be followed if rocks at the predicted collecting depth are found, or no soil at all is found, or large trees bar the soil collection. Considering this, a proficient design of a soil data sampling campaign in forest field is not always a simple process and sometimes represents a truly huge challenge. In this work, we present some difficulties that have occurred during two experiments on forest soil that were conducted in order to study the spatial variation of some soil physical-chemical properties. Two different sampling protocols were considered for monitoring two types of forest soils located in NW Portugal: umbric regosol and lithosol. Two different equipments for sampling collection were also used: a manual auger and a shovel. Both scenarios were analyzed and the results achieved have allowed us to consider that monitoring forest soil in order to do some mathematical and statistical investigations needs a sampling procedure to data collection compatible to established protocols but a pre-defined grid assumption often fail when the variability of the soil property is not uniform in space. In this case, sampling grid should be conveniently adapted from one part of the landscape to another and this fact should be taken into consideration of a mathematical procedure.