Browsing by Subject "Model development"

Now showing 1 - 2 of 2

Entwicklung eines computergestützten Assessment-Tools zur Erfassung des Ernährungszustandes von Senioren
(2011) Ott-Renzer, Cornelia; Biesalski, Hans-Konrad
Introduction: Worldwide a demographic change of population is to be observed, thus also in Germany. Special expertise in geriatric diagnostics and therapy will gain in importance. Thereby, a special relevance comes up to determination of nutritional status by corresponding screening and assessment. Conventional assessment mostly equates to questionnaire methods. Purpose/Question: Focus was on developing a software (geroMAT-Malnutrition Assessment and Therapy for gerontologic patients) for identification Senior´s nutritional state, a kind of "management tool" for diagnostics and intervention. Methods: Investigation was open, cooperative and multi-centric, as well as clinical-experimentally invested. In three partial studies (I: Suitability of the MNA® as a reference method; II: Anthropo-metry, biochemistry, body composition; III: Food patterns and intake, not-nutritive factors) indicators of malnutrition were initially selected. In a final analysis Model I (prognosis of malnutritional risk) and accordingly Model II (prognosis of the MNA®) have been developed for their use in geroMAT. Results: Prevalence of malnutrition was (in this random sample) 5%, 44% were at risk and 51% were well nourished. Due to inhomogeneity in class range by assessment with the MNA®, modelling of a dichotomic risk variable ("RiskMal", homogeneous) occurred. All up 25 features and 12 (optional) additional items from the partial studies I-III could be generated and attached to further analysis. Model I prognosticated "RiskMal" reliably (auROC=.739). Although firstly Model II predicted MNA® well (r=0,5167), model quality could be improved even further by the well-chosen parametres of a feature subset selection for Models I and II (Í: r=0,822;II: r=0,6634). Discussion: The Models I/II reached the requirements made on developing geroMAT. According to the features, geroMAT would be multi-centric usable, simple to learn and operable, documentable and reproduceable, interprofessionally and without high expense, as well as modern. Advantages of geroMAT, compared with the MNA® lay in its detailedness and its choice of further options, its capture of documented information off the normal anamnesis process and the initiation or monitoring of individual interventions. Conclusions: Mean aim of the study, the identification of indicators and model development, was reached. Other model validation studies should follow before the final clinical practice of geroMAT.
Land use change and its impact on soil properties using remote sensing, farmer decision rules and modelling in rural regions of Northern Vietnam
(2017) Nguyen, Thanh Thi; Cadisch, Georg
After the Indo China war in 1954, a dramatic rise in population in Northwest Vietnam led to an increased demand of agricultural land for food security requirements. Slash and burn systems which existed for many hundreds of years were replaced by intense cash crop systems, particularly maize production. Maize cropping was further expanded to steeper sloping areas, resulting in a risk of soil degradation. Therefore, investigating Land Use Change (LUC) and its impact on soil properties were considered in this study. The study aimed to identify LUC in 1954, 1973, the 1990s and 2007 in Chieng Khoi commune, Yen Chau district, Son La province, Vietnam using available remote sensing data. Furthermore, a detailed land use map classification method was developed using farmers’ decision rules. Based on farmers’ crop decision rules and, food requirement and population information, a simple LUC model was developed to simulate LUC annually from 1954 to 2007. Moreover, total soil nitrogen and carbon were determined under a chronosequence of intense cultivation. Thus, developing a modelling tool had the aim to assess the impacts of LUC on soil fertility at watershed level. The first case study (Chapter 3) presented the LUC assessment, using available remote sensing data combined with farmer information. Forest areas decreased from 1954 to 2007, except in the 1990s because of policies that aimed to encourage and support afforestation programmes to increase forest land. However, planted forest has since decreased again since 1999 whereas agricultural land has increased dramatically. Agricultural land expanded to both natural forest and planted forest areas until 2007 legally (with encouragement of agroforestry) and illegally thereafter (at the border between cultivated land and forest). The establishment of an artificial lake in Chieng Khoi commune opened the accessibility to forest land surrounding the lake, with a forest area of 929 ha remaining in 2007 compare to more than 2,500 ha in 1954. Paddy rice areas did not change because of their specific location (lower and flat lands), but production increased and was intensified by two cropping seasons per year due to irrigation improvements and a continuous water supply from the artificial lake. The second case study (Chapter 4) presented the development of a LUC model, using the outputs from the first case study comprising farmers’ decision rules and food requirements for an increased population. For later periods, the influence of market orientation factor was considered. The model successfully simulated the expansion of cultivation areas and replacement of forest land by agricultural land. Simulations were at accepted level of accuracy comparing actual and simulated LUC (Goodness-of-fit – GOF values greater than 0.7 and Figure of merit - FOM values greater than 50%). The third case study (Chapter 5) demonstrated an investigation of the soil fertility dynamic under intense cultivation and the development of a simple dynamic and spatially-explicit modelling tool to assess the changes in soil fertility. The Dynamic of total Carbon and Nitrogen distribution (DyCNDis) model was constructed using field data combined with literature information. The field data showed that, under a decade of maize mono cultivation in slope areas, both nitrogen and carbon were largely depleted. Furthermore, the DyCNDis model showed an acceptable level of validation (modelling efficiency – EF of 0.71 and root mean square error - RMSE of 0.42) to simulate nitrogen and carbon under intense maize cultivation at watershed level. Additionally, the model identified hotspot areas of 134 ha (18.9% of total upland cultivation areas) that are threatened by soil degradation through intense cultivation over a long-term period. In conclusion, the combination of qualitative and quantitative approaches allowed assessing impacts of LUC on environmental services such as soil fertility through the developed DyCNDis modeling tool. The combination of improved LUC analysis with a simple spatial dynamic soil fertility modeling tool may assist policy makers in developing alternative implementation strategies for local stakeholders in regions which face data limitations. The modelling tools developed in this study were able to successfully simulate LUC and to identify locations where soil conservation methods at watershed level need most urgently to be applied to avoid soil degradation. The model tools were able to simulate the trends rather than values of agricultural area expansion and reduction of soil nitrogen and carbon. The developed approaches could be linked and coupled to other modelling tools to economically consider benefits or ecological concerns toward sustainable crop production in remote and rural regions.