Wojciech Krawczyk
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Wojciech Krawczyk has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Piotr Wężyk
Department of Forest Resources Management, Faculty of Forestry, University of Agriculture in Krakow, 31-425 Krakow, Poland
4 shared publications - Erdenetuya Boldbaatar
Department of Forest Resources Management, Faculty of Forestry, University of Agriculture in Krakow, 31-425 Krakow, Poland
2 shared publications Jakub Miszczyszyn
1 shared publication- Jarosław Socha
Department of Forest Resources Management, Faculty of Forestry University of Agriculture in Krakow Kraków Poland
1 shared publication Luiza Tymińska-Czabańska
1 shared publication- Paweł Hawryło
Department of Forest Resources Management, Faculty of Forestry University of Agriculture in Krakow Kraków Poland
1 shared publication
Latest Publications
- Open Access
Fusing Sentinel-2, PlanetScope and Topography with a Machine Learning Approach for Improvement of Tree Species and Forest Disturbance Classification in Northern Mongolia
Tree species classification is challenging in the heterogeneous forest-steppe ecosystem. We evaluated whether bi-sensor data fusion (Sentinel-2, PlanetScope SuperDove) and topographic variables improve tree species classification in Selenge Province, northern Mongolia. Random Forest (RF) classified five common tree species—Siberian larch, Manchurian birch, Scots pine, Siberian spruce, and Siberian pine—as well as burned forests and non-forest areas using spectral bands, vegetation indices, and terrain derivatives. Sentinel-2, PlanetScope, and FUSION models were assessed across eight dates (2021–2025) using overall accuracy (OA), Kappa, and F1, all derived from polygon-based 5-fold cross-validation to avoid spatial autocorrelation bias. The FUSION model achieved the highest accuracy in June 2023 (OA = 0.913, Kappa = 0.895) and September 2022 (OA = 0.909, Kappa = 0.891). The Sentinel-2 model performed stably (OA = 0.850–0.905, mean = 0.883, CV = 2.0%), while the FUSION model was slightly higher on average (mean = 0.889) but more variable (CV = 2.4%). McNemar’s test showed FUSION significantly outperformed both single sensors on all dates (p < 0.001), although the practical gain was small and phenology-dependent. The FUSION model showed the closest aggregate area agreement with the national forest inventory (<1% error). FUSION improved accuracy only under specific phenological conditions, whereas Sentinel-2 alone was temporally more stable and equally accurate under full leaf-on conditions, making it a simpler, more robust baseline.
Open AccessFusing Sentinel-2, PlanetScope and Topography with a Machine Learning Approach for Improvement of Tree Species and Forest Disturbance Classification in Northern Mongolia
Tree species classification is challenging in the heterogeneous forest-steppe ecosystem. We evaluated whether bi-sensor data fusion (Sentinel-2, PlanetScope SuperDove) and topographic variables improve tree species classification in Selenge Province, northern Mongolia. Random Forest (RF) classified five common tree species—Siberian larch, Manchurian birch, Scots pine, Siberian spruce, and Siberian pine—as well as burned forests and non-forest areas using spectral bands, vegetation indices, and terrain derivatives. Sentinel-2, PlanetScope, and FUSION models were assessed across eight dates (2021–2025) using overall accuracy (OA), Kappa, and F1, all derived from polygon-based 5-fold cross-validation to avoid spatial autocorrelation bias. The FUSION model achieved the highest accuracy in June 2023 (OA = 0.913, Kappa = 0.895) and September 2022 (OA = 0.909, Kappa = 0.891). The Sentinel-2 model performed stably (OA = 0.850–0.905, mean = 0.883, CV = 2.0%), while the FUSION model was slightly higher on average (mean = 0.889) but more variable (CV = 2.4%). McNemar’s test showed FUSION significantly outperformed both single sensors on all dates (p < 0.001), although the practical gain was small and phenology-dependent. The FUSION model showed the closest aggregate area agreement with the national forest inventory (<1% error). FUSION improved accuracy only under specific phenological conditions, whereas Sentinel-2 alone was temporally more stable and equally accurate under full leaf-on conditions, making it a simpler, more robust baseline.
- Open Access
- 18 March 2026
- Published by Adam Mickiewicz University Poznan in Quaestiones Geographicae
- Vol. 45 (1) , 139-154
Preliminary forest tree species classification in northern provinces of Mongolia using Sentinel-2 and machine learning approach
This research addresses the need for precise, wide-scale monitoring of Mongolia’s boreal forests, as a part of a critical ecosystem that stores nearly 30% of global terrestrial carbon. Covering approximately 7.37 Mha in Mongolia, these forests form the southern fringe of the Siberian taiga and are increasingly affected by wildfires (95.9% of total forest losses) and logging (2.5% of losses) since 2000. The study focuses on Selenge, Darkhan-Uul and Tuv provinces located in northern Mongolia, where traditional forest inventory methods do not work well due to the extensive and difficult or inaccessible terrain. To overcome these challenges and needs of precise forest monitoring, we developed a classification framework using Sentinel-2 (European Space Agency [ESA]) multi-temporal satellite imageries (period 2020–2024), acquiring key phenological stages. We applied a random forest (RF) algorithm to classify five dominant tree species, that is, Siberian pine (SBP) (Pinus sibirica Ledeb.), Scotch pine (SP) (Pinus sylvestris L.), Siberian Larch (SBL) (Larix sibirica Ledeb.), Siberian spruce (SBS) (Picea obovata Ledeb.) and Manchurian birch (MB) (Betula platyphylla Sukaczev), forming the forest stand cover. The results of Sentinel-2 imageries processing demonstrate very high classiication overall accuracy (OA = 96.19%, κ = 0.949). Compared with existing forest management maps (based on in situ surveys), SBS (P. obovata Ledeb.) area share was underestimated, whereas MB (B. platyphylla) area share was overestimated, indicating observable differences in traditional forest inventories. This Sentinel-2 classification approach offers timely, cost-effective and accurate data tailored to Mongolian conditions, supporting sustainable forest management, conservation, reforestation, afforestation and REDD + programmes (Reducing Emissions from Deforestation and Forest Degradation).
- 18 March 2026
- Published by Adam Mickiewicz University Poznan in Quaestiones Geographicae
- Vol. 45 (1) , 139-154
Open AccessPreliminary forest tree species classification in northern provinces of Mongolia using Sentinel-2 and machine learning approach
This research addresses the need for precise, wide-scale monitoring of Mongolia’s boreal forests, as a part of a critical ecosystem that stores nearly 30% of global terrestrial carbon. Covering approximately 7.37 Mha in Mongolia, these forests form the southern fringe of the Siberian taiga and are increasingly affected by wildfires (95.9% of total forest losses) and logging (2.5% of losses) since 2000. The study focuses on Selenge, Darkhan-Uul and Tuv provinces located in northern Mongolia, where traditional forest inventory methods do not work well due to the extensive and difficult or inaccessible terrain. To overcome these challenges and needs of precise forest monitoring, we developed a classification framework using Sentinel-2 (European Space Agency [ESA]) multi-temporal satellite imageries (period 2020–2024), acquiring key phenological stages. We applied a random forest (RF) algorithm to classify five dominant tree species, that is, Siberian pine (SBP) (Pinus sibirica Ledeb.), Scotch pine (SP) (Pinus sylvestris L.), Siberian Larch (SBL) (Larix sibirica Ledeb.), Siberian spruce (SBS) (Picea obovata Ledeb.) and Manchurian birch (MB) (Betula platyphylla Sukaczev), forming the forest stand cover. The results of Sentinel-2 imageries processing demonstrate very high classiication overall accuracy (OA = 96.19%, κ = 0.949). Compared with existing forest management maps (based on in situ surveys), SBS (P. obovata Ledeb.) area share was underestimated, whereas MB (B. platyphylla) area share was overestimated, indicating observable differences in traditional forest inventories. This Sentinel-2 classification approach offers timely, cost-effective and accurate data tailored to Mongolian conditions, supporting sustainable forest management, conservation, reforestation, afforestation and REDD + programmes (Reducing Emissions from Deforestation and Forest Degradation).
- Open Access
- 2 November 2023
- Published by Elsevier in Forest Ecology and Management
- Vol. 551
How to adequately determine the top height of forest stands based on airborne laser scanning point clouds?
- 2 November 2023
- Published by Elsevier in Forest Ecology and Management
- Vol. 551
Open AccessHow to adequately determine the top height of forest stands based on airborne laser scanning point clouds?
- Open Access
- 9 February 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (4) , 951
Using Satellite Imagery and Aerial Orthophotos for the Multi-Decade Monitoring of Subalpine Norway Spruce Stands Changes in Gorce National Park, Poland
In the last 50 years, forest disturbances, caused mainly by insect outbreaks and windstorms, had a significant impact on the subalpine Norway spruce (Picea abies (L.) H. Karst.) stands across Europe. The high intensity of these factors often led to complete dieback of existing forest stands, as in Gorce National Park (Southern Poland). The aim of this study was to monitor land cover changes in subalpine Norway spruce stands and their dynamics in Gorce NP in the years 1977–2020 (43 years), with the use of archival remote sensing data. The study area was divided into two subareas: A—the Kudłoń and B—the Jaworzyna range. Changes were tracked in six defined land cover classes, based on available aerial orthophotos and Landsat (NASA) imagery, with the help of the authors’ photointerpretation key. The results showed that almost 50% of old-growth Norway spruce stands died in the analyzed time period (50.9% in subarea A; 48.8% in subarea B). However, young forests appeared in almost 17% of the study area (20.7% and 14.2% in subarea A and B, respectively). The dynamics of land cover changes were different for the analyzed subareas; in subarea A Norway spruce dieback processes weakened at the end of the analyzed time period, whereas in subarea B they maintained high intensity. The process of old-growth Norway spruce stands dieback is still occurring in Gorce NP, but it does not result in the disappearance of the whole subalpine spruce forest ecosystem but is rather a generational change, due to emerging young forests.
- 9 February 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (4) , 951
Open AccessUsing Satellite Imagery and Aerial Orthophotos for the Multi-Decade Monitoring of Subalpine Norway Spruce Stands Changes in Gorce National Park, Poland
In the last 50 years, forest disturbances, caused mainly by insect outbreaks and windstorms, had a significant impact on the subalpine Norway spruce (Picea abies (L.) H. Karst.) stands across Europe. The high intensity of these factors often led to complete dieback of existing forest stands, as in Gorce National Park (Southern Poland). The aim of this study was to monitor land cover changes in subalpine Norway spruce stands and their dynamics in Gorce NP in the years 1977–2020 (43 years), with the use of archival remote sensing data. The study area was divided into two subareas: A—the Kudłoń and B—the Jaworzyna range. Changes were tracked in six defined land cover classes, based on available aerial orthophotos and Landsat (NASA) imagery, with the help of the authors’ photointerpretation key. The results showed that almost 50% of old-growth Norway spruce stands died in the analyzed time period (50.9% in subarea A; 48.8% in subarea B). However, young forests appeared in almost 17% of the study area (20.7% and 14.2% in subarea A and B, respectively). The dynamics of land cover changes were different for the analyzed subareas; in subarea A Norway spruce dieback processes weakened at the end of the analyzed time period, whereas in subarea B they maintained high intensity. The process of old-growth Norway spruce stands dieback is still occurring in Gorce NP, but it does not result in the disappearance of the whole subalpine spruce forest ecosystem but is rather a generational change, due to emerging young forests.
- 31 December 2020
- Published by Association of Polish Surveyors in Archives of Photogrammetry, Cartography and Remote Sensing
- p. 13
MAPPING OF WINDSTORM DAMAGE OCCURRING IN THE FOREST STANDS OF CZERNIEJEWO FOREST DISTRICT (POLISH STATE FORESTS NATIONAL HOLDING) USING AERIAL PHOTOGRAPHS AND SENTINEL-2 SATELLITE IMAGERY
On the night of 11 and 12.08.2017 a severe hurricane passed over Poland, in a belt of almost 300 km, causing damage in forest stands of area exceeding 100 000 ha. The study aimed to demonstrate the implementation of remote sensing technologies in the process of determining the extent of forest stand damages in the Czerniejewo Forest District (RDLP Poznań) caused by wind and monitoring the progress of clean-up work. In this study were used digital aerial orthophotos commissioned by the State Forests National Forest Holding (PGL LP) and Sentinel-2 (ESA) satellite images as well. The area of damaged stands was determined with the use of two approaches, i.e.: supervised classification (approach A) and thresholding of values of Normalised Difference Vegetation Index – NDVI (approach B). The obtained results were compared to reference data obtained by visual interpretation of high resolution RGB aerial orthophotos by RDLP Poznań experts. Monitoring of the progress of the clean-up works in damaged stands was carried out in 9-time intervals. The conducted image classification and spatial GIS analyses showed that the area of stands damaged by the wind was for methods A and B: 579.16 ha and 516.01 ha, respectively, with 631.00 ha as the reference. The results obtained in the study indicate errors in underestimating the area of forest stand damage based on Sentinel-2, i.e.: 51.84 ha (8.2%) in the case of method A and by 114.99 ha (18.2%) for method B. In the whole analysed time, clean-up operations were carried out on the total area of 762.33 ha of damaged forest stands, and their highest intensity was observed in the first 4 months after the storm. The work showed the applicability of free of charge Sentinel-2 (ESA) satellite imagery in the process of determining the extent of forest stand damages, pointing to the supervised classification method (Maximum Likelihood algorithm; ML) as more accurate than using the threshold of NDVI.
- 31 December 2020
- Published by Association of Polish Surveyors in Archives of Photogrammetry, Cartography and Remote Sensing
- p. 13
MAPPING OF WINDSTORM DAMAGE OCCURRING IN THE FOREST STANDS OF CZERNIEJEWO FOREST DISTRICT (POLISH STATE FORESTS NATIONAL HOLDING) USING AERIAL PHOTOGRAPHS AND SENTINEL-2 SATELLITE IMAGERY
On the night of 11 and 12.08.2017 a severe hurricane passed over Poland, in a belt of almost 300 km, causing damage in forest stands of area exceeding 100 000 ha. The study aimed to demonstrate the implementation of remote sensing technologies in the process of determining the extent of forest stand damages in the Czerniejewo Forest District (RDLP Poznań) caused by wind and monitoring the progress of clean-up work. In this study were used digital aerial orthophotos commissioned by the State Forests National Forest Holding (PGL LP) and Sentinel-2 (ESA) satellite images as well. The area of damaged stands was determined with the use of two approaches, i.e.: supervised classification (approach A) and thresholding of values of Normalised Difference Vegetation Index – NDVI (approach B). The obtained results were compared to reference data obtained by visual interpretation of high resolution RGB aerial orthophotos by RDLP Poznań experts. Monitoring of the progress of the clean-up works in damaged stands was carried out in 9-time intervals. The conducted image classification and spatial GIS analyses showed that the area of stands damaged by the wind was for methods A and B: 579.16 ha and 516.01 ha, respectively, with 631.00 ha as the reference. The results obtained in the study indicate errors in underestimating the area of forest stand damage based on Sentinel-2, i.e.: 51.84 ha (8.2%) in the case of method A and by 114.99 ha (18.2%) for method B. In the whole analysed time, clean-up operations were carried out on the total area of 762.33 ha of damaged forest stands, and their highest intensity was observed in the first 4 months after the storm. The work showed the applicability of free of charge Sentinel-2 (ESA) satellite imagery in the process of determining the extent of forest stand damages, pointing to the supervised classification method (Maximum Likelihood algorithm; ML) as more accurate than using the threshold of NDVI.