Modeling net primary productivity and leaf area index with remote sensing techniques in pure crimean pine stands in Ankara Regional Directorate of Forestry

In this study, the possibilities of modeling the net primary production (NBU) calculated using the Carnegie-Ames-Stanford Approach (CASA) model in pure Crimean pine stands spread in Ankara Regional Directorate of Forestry with local and remote sensing data were investigated. A total of 180 temporary samples were taken in the study. Classical inventory measurements were made in each sample area and some stand parameters (stand basal area, stand volume, stand mean diameter, stand number of trees, stand top height and leaf area index) were calculated for each sample area. In addition, Landsat 8 OLI and Sentinel-2 satellite images were used as remote sensing data in the study. The reflectance, vegetation index and texture values were obtained from satellite images. Relationships between NBU values obtained for each sample area using the CASA model, stand parameters and data obtained from satellite images were modeled using multiple regression analysis, support vector machines and deep learning techniques. With the applied modeling techniques, 90° for stand basal area (Model R2=0,91, Test R2=0,90), 90° for stand volume (Model R2=0,85, Test R2=0,85), stand mean diameter 9x9 filter (Model R2=0,88, Test R2=0,88), 45° for stand number of trees (Model R2 = 0,98, Test R2 = 0,96), 135° for stand top height (Model R2=0,90, Test R2=0,89), 0° for leaf area index (Model R2=0,96, Test R2=0,96) and 45° for NBU (Model R2=0,96, Test R2=0,95) using the texture values, the highest level of model success was obtained. In order to check the CASA model validity, 30 sample areas were established and litterfall amounts were obtained. In the validity check of the CASA model, the highest correlation level was obtained with the male flower fall (r=-0,75). In addition, a negative relationship was found with the elevation and stand age of the NBU, and a positive relationship with the temperature. It has been determined that NDVI and solar radiation are effective in monitoring the temporal distribution of NBU, and stand features when evaluated spatially. It has been observed that the combination of filter and degree texture data improves the modeling success.