Hochschule für nachhaltige Entwicklung Eberswalde
•••
Menü

ABSTRACTS

Abstracts

Use of GIS for statistics monitoring, planning, and development
Albulena Hashani

Kosovo’s Agency of Statistics, respectfully Office of Cartography is currently working on procedures to create a Statistical Web Map where all statistical data will be available to the public, meanwhile the comparison research was done to use the pros and cons in Kosovo’s WebMap. Swiss, Slovenian, and Albanian National Statistical Institute Websites  - Study of Comparison. These three countries have a similar area, and population but with differences in economy, demography, and organization of their WebMaps as well. This study’s results will be used for creating Kosovo’s Web map, a project which is still in progress due to organizational and bureaucratic procedures. Kosovo is a newly developing country with a young population and new constructions can be easily seen in every area, the geodatabase needs to be updated to keep in time with changes in the field. Detection of changes in new orthophotos will keep the database updated and will be used on the next population census that will be held in the next year, and other statistical processes. Since January, another pilot census was being held in the field; the data collection was done by using Survey Solution Application. This process has been ongoing for 3months; this process includes data collection control on application and in the field as well. Maps are being designed for specific municipalities with updated data and these maps will be used also on the next census and other field surveys.

Presentation on Friday 13 of May of 2022 at 15:30




Monitoring of deforestation and forest degradation using remote sensing and GIS: A case study of Edo-Ekiti, Nigeria.

Ogunsina, Ayodeji

The forest is an essential component of our environment. The major function of forests is to absorb carbon dioxide and create oxygen in order to maintain a balanced and healthy environment. Climate control and carbon accumulation are also essential functions of forests. One-third of the world's forest has been gone, an area twice the size of the United States.
Therefore, forest change must be monitored so that deforestation and development may be planned while maintaining the ecosystem's balance. The present study focuses on the use of Landsat imagery to track deforestation and forest degradation in Ado-Ekiti, Nigeria. Using the Landsat 5 TM image from 1998 and the Landsat 8 OLI image from 2020, we look at the variance in forest cover dynamics from 1998 to 2020. The LULCC detection results show a significant rise in agricultural land from 16% in 1998 to 34% in 2020, while forest cover decreased by 81% in 1998 and 63%t in 2020. According to the study, agricultural expansion is the primary cause of deforestation. Other factors contributing to the loss of forest cover include population increase, need for wood for construction and fuel, illiteracy, poverty, land scarcity, and a lack of environmental education. Overall, our findings point to the necessity for participatory forest management and public education in order to preserve the remaining forest.

Presentation on Friday 13 of May of 2022 at 17:00



Remote Sensing: Land cover mapping of agricultural areas of Multan and Bahawalpur.
Iqbal, Hafiz Abdul Rehman

Land Use Land Cover change has several impacts on human beings, the environment, and the economy of a country. In this study, we derived a crop type ground truth dataset for land cover classification in Punjab Province of Pakistan, especially Multan and Bahawalpur. The agricultural land cover mapping was focused on cotton, as this cash crop plays an important role in the economy of Pakistan. Preliminary results indicate that Multan is an area in which cotton production already shifted to a more diversified agricultural land use. For Bahawalpur, the results were less clear. The causes, effects, and other information about the agrarian transformation will be evaluated with the support of local farmers of each area in the future.

Keywords: Agrarian transformation, ground-truthing

Presentation on Friday 13 of May of 2022 at 16:00.




Investigating a Compatible Modeling System for Pinus sylvestris
Seymour, Jackson

Through the pairing of a variable-form segmented taper model with total volume equations, simultaneous estimation of carbon concentration, biomass, and green-weight can be achieved at the single-tree level. This accomplished via the inclusion of density and weight constants. This concept is known as a compatible modelling system. The focus of this research investigation is to assess the viability of applying compatible biometric models for Pinus Sylvestris based on the work of Quiñonez-Barraza et al. (2018) and others. No such investigation has yet been attempted for Pinus Sylvestris.

The database of biometric measurements used in this study originated from the Gubin forest district of the Regional Directorate of State Forests in Zielona GĂłra, in the west of Poland. Contained in the database is over 1400 measurements taken from 90 trees over 18 plots. The compatible system was comprised of 13 equations which were fitted via generalized nonlinear least squares. Two model validation procedures were conducted to assess the results.

 All equations and parameters in the investigation were fitted and achieved statistical significance. A high degree of success was observed in predicting observed data. Despite the taper equation being the basis for all proceeding biometric models, simultaneous fitment was not achieved.

A final section will be reserved at the end of the discussion of results to detail experiences from the internship itself. This discussion will include learning processes, difficult moments, and errors that were encountered along the timeline of this fascinating experience.

Presentation on Friday 13 of May of 2022 at 11:00.




GNSS measurements quality in forested area
Grabek, Karol

The GNSS (Global Navigation Satellite System) system is designed to accurately determine the position of a point or other object. It is also widely used in the forest environment in many situations. Unfortunately, this technology is associated with many problems, such as reducing the penetration of satellite signals into the forest, which results in less accurate results. Therefore, in this study the accuracy of the GNSS receiver measurements in the forest environment was examined.

Two types of data were collected in the Herby Forest District. The point data, which constituted the centers of the sample plots measured with the GNSS receiver together with the trees, were manually calibrated to the CHM (Canopy Height Model) obtained from the ALS (Airborne Laser Scanning).

Each of the 906 sample plots were visually inspected and necessary actions such as no action, move, rotate, or both were performed. The offset vector for each point was then calculated. After completion of the study, a statistical analysis of the obtained results was performed according to the following criteria: all plots, species, height and age classes. The study showed that 14.46% of the surfaces did not require any action. It was difficult to select among the species the best result due to the large influence of the abundance of individual species. It was also found that the height class of 20m - 40m turned out to be the most accurate. The GNSS and individual category results mean that the measured location is not as accurate and the sample plot determination technique with a GNSS receiver alone may not be sufficient.

Keywords: GNSS accuracy, GNSS in forest, sample plot position, CHM, offset vector

Presentation on Friday 13 of May of 2022 at 11:30




Mapping crown condition survey’s trees with individual tree delineation based on LiDAR data.
Fares, Lotfi

Forests play a large role in our lives, monitoring, and managing the forest has become reality in the focus of remote sensing research. Individual tree delineation has shown promising results in providing precise tree metrics quickly and efficiently. This has put the lidar (Light Detection and Ranging) at the top of the list of remote sensing techniques for plot and stand-level forest inventory. From a wide array of algorithms to segment individual trees at the point cloud level, AMS3D (Adaptive Mean Shift algorithm) has presented promising results in this regard. This report tested the response of AMS3D on temperate mixed forest in the Traunstein area in southern Germany according to different parameters combinations. The segmentation results were used to predict DBH (Diameter at Breast Height) and then compare it at a single tree level with an inventory data set by calculating statistical metrics. The highest RÂČ value is 0.511 and the lowest RMSE (Root Mean Square Error) is 0.134 m. The results showed that AMS3D performance is influenced by the value of the parameters, site structure, and tree heights distribution. Results of previous analyses were then used to segment 26 plots for the crown condition survey and the results of the segmentation were then used to build multiple linear regression models using 5 predictors to estimate stand age. Lastly, backward stepwise selection was applied to optimize the final model, resulting in an RÂČ for the final model of 0.6258. The average age of the stand was compared against the predicted age through statistical metrics, RÂČ = 0.6241, RMSE = 24.297 were found to be acceptable considering the small size of the data. This analysis can be a good opportunity to carried out with more precise and bigger data and maybe apply ML ( machine learning) to produce a better sophisticated model.


Presentation on Friday 13 of May of 2022 at 13:00



Data Management for Machine learning: Data collection and organizing of Prötzel forest area near Berlin for machine learning
Anwar, Muddassir

Image processing become more essential components of data quality control, with the advancements of Artificial intelligence and machine learning. As traditional methods of image processing by individuals were very time-consuming and has less accuracy. This project is based on managing and creating big data for image processing machine learning. As in this modern world, data is the future. Organizations that havemore data will have more opportunities to towards progress and development. In this project, additional data is created from limited available data for machine learning. Big data creates more accurate machine learning outputs of the given input data. The project is consisting of correcting the GPS points of already available GPS points of the forest area. The area of interests is then cropped from the mosaic through polygons. To create more data sets from the existing available limited data, the single individual tree image is crop into more images. And in the last part, the cropped images are augmented in different ways to have more data for machine learning models that will use the data in the future. In this case study, we have selected a forest area in Brandenburg called Prötzel. For this project the tools used are QGIS, R studio for R language, Visual Studio Code for Python.

KEYWORDS : QGIS, R, Studio, Visual Studio Code, Python, Rlanguage

Presentation on Friday 24 of September of 2021 at 14:00



Application of deep learning approach for crown detection in inventory of forests in Poland
Raczka, Rafal

Deep learning (DL) is recently very fast developing field of machine learning. It showed breaking records for pattern recognition in various fields. Many researches proven broad application capabilities of deep learning in ecology. However, there are not many application of this technology in forestry and even less applied in actual forest management tasks. The forest inventory in Poland is still based on manual field work and did not change much for decades. Despite many remote sensing methods already developed, there are still many obstacles for their implementation. DL can help to pass some of these obstacles highly lowering operational cost and offering high flexible models which can be easily retrained for different site types. In this study, deep learning was used for tree crown detection from RGB imagery acquired from aerial vehicle. Different approaches for it was tested. These approaches include imagery of different parameters and different models. The proposed methods were evaluated on different forest site types. The implementation of the results was proposed and initially tested for estimating forest parameters according to requirements of Polish forest inventory and modelling of forest ecosystems


Presentation on Friday 13 of May of 2022 at 14:30.




Fixed effects taper modelling and volume estimation of Scots pine (Pinus sylvestris) with TLS-derived data in Poland.
Magnuson, Robert

It is common knowledge that having an accurate understanding of an object’s shape can lend itself to an accurate estimation of volume. In forest metrics, volume is a parameter that is paramount to forest economics and management. Precise wood
volume estimation lies within a mathematical understanding of the relationships between the height and diameter of the tree stem, and adjacent to, tree form. A traditional method of estimating wood volume requires physical measurements of trees: diameter at breastheight (DBH) and total height measurements, at a minimum. Historically, these two measurement parameters have led the baseline to time-efficient “boots on the ground” estimation. With advances in remote sensing technology, we are increasing our ability to measure forests more efficiently, less invasively, with acceptable precision. This study analyzes terrestrial laser scanned (TLS) data of 2500 Scots pine (Pinus sylvestris) in six forest districts in Poland through the REMBIOFOR project, with the objective to compare the efficacy of volume estimation of remote sensed data to traditional volume estimation methods. Seven fixed-effects models were trained and validated with an independent dataset of traditionally felled and measured Scots pine trees to assess the accuracy of TLS data with regard to volume estimation. The results of wood volume from fixedeffects models are promising, when compared to traditional methods. With progression into mixed-effects models and exploration of newer equations, these results can aid in the search and selection of less invasive survey techniques of forests in Poland.

Presentation on Friday 13 of May of 2022 at 10:30.



Documentation of avian fauna in Chatri Lake, Amravati (Ms, India): Data analysis a case study.
Avinash Wagh, Shubham

India is very rich in avian diversity. Chatri lake which is near to the Amravati city facing major habitat loss due to increase in human activities, anthropogenic pressure, biotic and abiotic factors which contributes in loss of habitat and therefore, many species ground fostering and roosting in this region also get affected due to this. The current assessment was accomplished for the documentation of diversity, species richness, abundance, and evenness of birds for the year 2017, 2018 and 2019 and to know the current status of avian fauna in and around the Chatri lake of Amravati explicitly utilizing eBird information. For this study continuously 3 years from 2017 to 2019 of data has been obtained through eBird website to analyse data. Total 179 species have been recorded in these three years out of which 74 species observed in 2017, 154 observed in 2018 and 53 observed in 2019. In year 2017 a total of 223 individuals from all the species were reported, for 2018, 1077 individuals were reported and for 2019, 206 individuals were reported. The diversity index, species evenness, species abundance for 2017 were 3.95, 0.22, and 225.26 whereas for 2018, 4.44, 0.19, 1060.66 and for 2019, 3.6, 0.24, 204 respectively. All these parameters show equitable presence of faunal diversity at Chatri Lake area Amravati, but at the same time after comparing the data between 2018 and 2019 it shows an Alarming-call for future conservation measures. In this startle circumstance this study showed a preliminary documentation of avian fauna in Chatri lake and this perception would be of incredible use henceforth for avian studies.


Presentation on Friday 13 of May of 2022 at 16:30