Transferability of ALS-based forest attribute models when predicting with drone-based image point cloud data

Toivonen, Janne; Korhonen, Lauri; Kukkonen, Mikko; Kotivuori, Eetu; Maltamo, Matti; Packalén, Petteri

doi:10.1016/j.jag.2021.102484

Cited by 3 publications

(3 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One reason for the relatively small rate of improvement for Ho in particular, is that models for Ho most frequently include a large height percentile as explanatory variable, which is less affected by sensor effect and flying altitude compared to smaller height percentiles and many other metrics representing density and height variation (Naesset 2009). Height was also the forest attribute most successfully predicted using an external model by van Ewijk et al (2020) and Toivonen et al (2021).…”

Section: Discussionmentioning

confidence: 99%

Estimating forest attributes in airborne laser scanning based inventory using calibrated predictions from external models

Garrido¹,

Gobakken²,

Ørka³

et al. 2022

Silva Fenn.

View full text Add to dashboard Cite

Forest management inventories assisted by airborne laser scanner data rely on predictive models traditionally constructed and applied based on data from the same area of interest. However, forest attributes can also be predicted using models constructed with data external to where the model is applied, both temporal and geographically. When external models are used, many factors influence the predictionsâ accuracy and may cause systematic errors. In this study, volume, stem number, and dominant height were estimated using external model predictions calibrated using a reduced number of up-to-date local field plots or using predictions from reparametrized models. We assessed and compared the performance of three different calibration approaches for both temporally and spatially external models. Each of the three approaches was applied with different numbers of calibration plots in a simulation, and the accuracy was assessed using independent validation data. The primary findings were that local calibration reduced the relative mean difference in 89% of the cases, and the relative root mean squared error in 56% of the cases. Differences between application of temporally or spatially external models were minor, and when the number of local plots was small, calibration approaches based on the observed prediction errors on the up-to-date local field plots were better than using the reparametrized models. The results showed that the estimates resulting from calibrating external models with 20 plots were at the same level of accuracy as those resulting from a new inventory.

show abstract

Section: Discussionmentioning

confidence: 99%

Estimating forest attributes in airborne laser scanning based inventory using calibrated predictions from external models

Garrido¹,

Gobakken²,

Ørka³

et al. 2022

Silva Fenn.

View full text Add to dashboard Cite

show abstract

“…It would be an interesting option to interpolate general diameter-height relationship rasters for the whole of Finland and to test the information as additional variables in nationwide models. Most recently, Toivonen et al (2021) predicted Hdom and V for the Liperi test plots by examining multiple ALS-based models from different regions in DIPC-based predictions. They used the same data and estimation methods presented in study III to fit fixed ALSbased Hdom and V models for all 22 regions.…”

Section: Further Studiesmentioning

confidence: 99%

Prediction of forest attributes using airborne laser scanning-based models without new in-situ field measurements

Kotivuori¹

2022

Diss. For.

Self Cite

View full text Add to dashboard Cite

The era of airborne laser scanning (ALS) and the development of new forest inventory methods has reduced the need for field visits and overall inventory costs over the last two decades. Although the development of inventory methods has been considerable, some systematic field visits are usually always required. For example, the most common ALS inventory method, the area-based approach (ABA), leans on field sample plot measurements. Likewise in the ALS inventory, the ABA method can also be used in drone-based inventories with image point cloud (IPC) data. Due to the small areal coverage of the drones, local sample plot measurements in drone image point cloud (DIPC) inventories are not usually profitable. The objective of this thesis was to examine the performance of ALS-based forest attribute models in ALS- and DIPC-based ABA inventories without new in-situ field measurements. In this study, nationwide ALS models for three forest attributes (stem volume, above ground biomass and dominant height) were fitted for the whole of Finland, and regional-level error rates of the nationwide model predictions were assessed. As the nationwide models tended to exhibit systematic region-wise under- and over-predictions, different calibration methods were examined. First, calibration of nationwide models with a small number of new field measurements from the target area was simulated. Second, the nationwide stem volume model or its regional predictions was calibrated without new in-situ field measurements by three test scenarios: a) using additional calibration variables in the models to account for geographical and environmental conditions throughout the country, b) refitting of the models by using existing sample plots from nearby regions, and c) matching the regional-level predictions with national forest inventory data. The DICP-based forest inventory without new in-situ field measurements was evaluated by replacing the ALS metrics from the ALS-based models with DIPC metrics when the models were applied. In the DIPC inventory, the metrics used in the ALS models were selected carefully so that they would be similar to the corresponding DIPC metrics. The results showed that forest attributes can be predicted without new in-situ field measurements using nationwide ALS-based models with moderate error rates. The systematic errors associated with the nationwide models decreased when the models were fitted with additional calibration variables, such as degree days, precipitation, and tree species proportions. However, the measurement of a carefully selected set of sample plots (e.g., 20 plots) from the target area for the calibration of the nationwide model is recommended, in instances where it is economically feasible. Prediction of forest attributes using ALS-based models with DIPC metrics is possible provided the predictor variables describe the upper canopy layer. The lowest error rates in DIPC-based inventories were obtained when the ALS-based model was fitted in a nearby region and the inventory units were disaggregated to coniferous and deciduous dominated areas before the prediction.

show abstract

“…For instance, [23] achieved similar prediction accuracy for shrubs in sagebrush steppe across different elevations but found inconsistencies in the prediction of grasses and bare ground. [24] showed how spatial heterogeneity due to diverse forest structures results in substantial differences between drone data and field measurements (more than 50%) when models were transferred to test sites for predicting forest attributes like stem volume. However, in agricultural systems, the transferability of drone-based models appears to be less affected, possibly due to their typically homogeneous setup [25].…”

Section: Introductionmentioning

confidence: 99%

Spatio-Temporal Transferability of Drone-Based Models to Predict Forage Supply in Drier Rangelands

Amputu,

Männer,

Tielbörger

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

Unmanned aerial systems offer a cost-effective and reproducible method for monitoring natural resources in expansive areas. But the transferability of developed models, which are often based on single snapshots, is rarely tested. This is particularly relevant in rangelands where forage resources are inherently patchy in space and time, which may limit model transfer. Here, we investigated the accuracy of drone-based models in estimating key proxies of forage provision across two land tenure systems and between two periods of the growing season in semi-arid rangelands. We tested case-specific models and a landscape model, with the expectation that the landscape model performs better than the case-specific models as it captures the highest variability expected in the rangeland system. The landscape model did achieve the lowest error when predicting herbaceous biomass and predicted land cover with better or similar accuracy to the case-specific models. This reinforces the importance of incorporating the widest variation of conditions in predictive models. This study contributes to understanding model transferability in drier rangeland systems characterized by spatial and temporal heterogeneity. By advancing the integration of drone technology for accurate monitoring of such dynamic ecosystems, this research contributes to sustainable rangeland management practices.

show abstract

Transferability of ALS-based forest attribute models when predicting with drone-based image point cloud data

Cited by 3 publications

References 23 publications

Estimating forest attributes in airborne laser scanning based inventory using calibrated predictions from external models

Estimating forest attributes in airborne laser scanning based inventory using calibrated predictions from external models

Prediction of forest attributes using airborne laser scanning-based models without new in-situ field measurements

Spatio-Temporal Transferability of Drone-Based Models to Predict Forage Supply in Drier Rangelands

Contact Info

Product

Resources

About