Interconnection of a Forest Growth Model and a Structural Model for Young Poplar Trees (Populus spp.)
by Christoph Stiehm
Date of Examination:2019-09-27
Date of issue:2020-03-20
Advisor:Prof. Dr. Winfried Kurth
Referee:Prof. Dr. Jürgen Nagel
Referee:Prof. Dr. Christian Ammer
Files in this item
Name:Diss.pdf
Size:21.3Mb
Format:PDF
Abstract
English
When planting fast-growing tree species such as poplars and willows on agricultural land in short rotation coppice plantations, site selection and the associated yield potential pose a central decision for the practitioner. In connection with the cultivar aspect there has been a need for research on the interaction between site and genotype in terms of growth performance. The aim of this work is to examine these questions on several levels. For this purpose, a multi-scale approach was chosen in the framework of which two model complexes are developed which are then connected by an interface. The first model complex incorporates the implementation of a yield simulator which depicts single tree based growth and mortality as a function of competition and site conditions. The data basis for this is growth data from the joint research project ProLoc funded by the BMEL. For this purpose, 18 trial sites are chosen which were initiated on a broad amplitude of environmental conditions. Following a uniform experimental design, monoclonal trial plots with three poplar and two willow clones (interspecific crossed hybrids) were supervised in two tri-annual rotations and cut back after the third year. Based on the model of the forest growth simulator BWINPro and the associated TreeGross program library, several models are parameterized which, in addition to the survival rates after planting and harvest, estimate the height increment in the first and second rotation. With the distance-independent competition index ``basal area of larger trees'' the development within the stands can be predicted. Regarding the growth performance on the site level, the parameters of planting date, available water capacity, German agricultural soil quality rating, sum of precipitation in May and June and mean temperature in June and July are identified as influential by variable selection. To estimate the height increment and survival after pruning, tree height before harvest is regarded as an independent variable. The factor clone indicates differences in the growth processes within the models but interactions with site variables can not be determined as significant. Missing variables such as the mean annual increment in dry matter yield in oven-dry tons ha^-1 a^-1 are estimated by additional functions parameterized with the dataset. The individual models are connected to a simulation procedure and the overall predictive power is assessed. Good results can be achieved for the first rotation with squared correlations of the observed and estimated mean stand height of 0.79. However, in the second rotation the estimation quality decreases to 0.53. There are single sites with considerable deviations. The depth of the soil sampling and missing extended information on the water supply are suspected as problematic here. The second model complex includes a structural model focused on the poplar genotypes and the second rotation. First, several measuring methods were identified which are deemed suitable for determining the tree architecture in terms of geometry and topology of the above-ground woody biomass, as well as the morphology of foliage in terms of leaf architecture and leaf shape. For the branch architecture, a manual method and a semi-automatic method with an electromagnetic digitizer for determining branch curvature have been selected and employed. The leaf architecture was measured by a manual method. The leaf shape could be determined by digitizing collected leaves. After analyzing the obtained data, several models are parameterized. As a result, the probability of bud growth and the dimensions and orientation in space of developing shoots can be estimated for apical and lateral buds. The models differentiate between main and minor stems, prolongation and lateral shoots, long and short shoots and, within the lateral shoots, sylleptic and regular shoots. The starting point here is the estimation of the number of internodes per shoot which in turn influences other parameters such as the branch angle and the curvature through the shoot length. Other factors underlying several models are the age, branch order and the genotypic influence. Parameters such as foliage and leaf size can mainly be estimated by the relative height with regard to the absolute tree height. The leaf shape in turn is determined by contour points whose coordinates are calculated as a function of the leaf blade length. As part of the analysis of these models, only slight differences in the structure between the clones are found. Exceptions are the curvature and branching angles of the lateral shoots for one of the clones, for which the models reproduce the observable slender habitus. Significant differences also occur in the leaf shape which reflect the leaf shapes of the underlying parent species of the hybrids. The individual model functions are then implemented into a structural model in the model platform GroIMP. The resulting model can simulate the development of the tree structure for each of the three clones in annual steps. Arbitrarily large stands can be simulated that have realistically varying tree sizes through stochastic components in the model. The interconnection of the two model complexes is realized through the import of single tree data from the yield model into the structural model. Two further models are parameterized to determine the number of internodes from the shoot length as annual height increment of the yield model for the structural model and to modify the growth of the minor stems in dependence of the main stem growth. Additionally, the single tree mortality generated by the yield simulator is incorporated into the structural model. Further research will show whether it is possible to improve the yield model by validation with data from other experiments to include deeper soil layers here. The structural model could be extended to a complete functional structural plant model by incorporating a physiology module. By extending the interconnection to return data from the structural model to the yield model, the predictive power could be improved, for example by means of extended possibilities for modeling the within-stand competition dynamics.
Keywords: Populus; Poplar; Short Rotation Coppice; Forest Growth Model; Yield Model; Structural Plant Model; Tree Architecture; Model Interconnection; Branch Architecture; Leaf Morphology; GroIMP; BWinPro; Multi-Scale Approach