Lauri Korhonen, Daniela Ali-Sisto, Timo Tokola. (2015). Tropical forest canopy cover estimation using satellite imagery and airborne lidar reference data. Silva Fennica vol. 49 no. 5 article id 1405. https://doi.org/10.14214/sf.1405

Keywords: logistic regression; beta regression; forest area; international forest definition; ALOS AVNIR-2; vegetation index

Highlights: The fusion of airborne lidar data and satellite images enables accurate canopy cover mapping; The zero-and-one inflated beta regression is demonstrated in large area estimation; Forest/non-forest classification should be done directly, for example by using logistic regression.

Abstract | Full text in HTML | Full text in PDF | Author Info

The fusion of optical satellite imagery, strips of lidar data and field plots is a promising approach for the inventory of tropical forests. Airborne lidars also enable an accurate direct estimation of the forest canopy cover (CC), and thus a sample of lidar strips can be used as reference data for creating CC maps which are based on satellite images. In this study, our objective was to validate CC maps obtained from an ALOS AVNIR-2 satellite image wall-to-wall, against a lidar-based CC map of a tropical forest area located in Laos. The reference CC values which were needed for model training were obtained from a sample of four lidar strips. Zero-and-one inflated beta regression (ZOINBR) models were applied to link the spectral vegetation indices derived from the ALOS image with the lidar-based CC estimates. In addition, we compared ZOINBR and logistic regression models in the forest area estimation by using >20% CC as a forest definition. Using a total of 409 217 30 × 30 m population units as validation, our model showed a strong correlation between lidar-based CC and spectral satellite features (root mean square error = 12.8%, R² = 0.82). In the forest area estimation, a direct classification using logistic regression provided better accuracy than the estimation of CC values as an intermediate step (kappa = 0.61 vs. 0.53). It is important to obtain sufficient training data from both ends of the CC range. The forest area estimation should be done before the CC estimation, rather than vice versa.

Korhonen, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland; (current) University of Helsinki, Department of Forest Sciences, P.O. Box 27, FI-00014 University of Helsinki, Finland http://orcid.org/0000-0002-9352-0114 E-mail: lauri.z.korhonen@helsinki.fi
Ali-Sisto, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: dheikkil@student.uef.fi
Tokola, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland. E-mail: timo.tokola@uef.fi

article id 275, category Research article

Lauri Korhonen, Kari T. Korhonen, Pauline Stenberg, Matti Maltamo, Miina Rautiainen. (2007). Local models for forest canopy cover with beta regression. Silva Fennica vol. 41 no. 4 article id 275. https://doi.org/10.14214/sf.275

Keywords: beta regression; canopy cover; forest canopy

Abstract | View details | Full text in PDF | Author Info

Accurate field measurement of the forest canopy cover is too laborious to be used in extensive forest inventories. A possible alternative to the separate canopy cover measurements is to utilize the correlations between the percent canopy cover and easier-to-measure forest variables, especially the basal area. A fairly new analysis technique, the beta regression, is specially designed for modelling percentages. As an extension to the generalized linear models, the beta regression takes into account the distribution of the model residuals, and uses a logistic link function to ensure logical predictions. In this study, the beta regression method was found to perform well in conifer dominated study area located in central Finland. The same model shape, with basal area, tree height and an additional predictor (Scots pine: site fertility, Norway spruce: percentage of hardwoods) as independent variables, produced good results for both pine and spruce dominated sites. The models had reasonably high pseudo R-squared values (pine: 0.91, spruce: 0.87) and low standard errors (pine: 6.3%, spruce: 5.9%) for the fitting data, and also performed well in a cross validation test. The models were also tested on separate test plots located in a different geographical area, where the prediction errors were slightly larger (pine: 8.8%, spruce: 7.4%). In pine plots, the model fit was further improved by introducing additional predictors such as stand age and density. This improved also the performance of the models in the cross validation test, but weakened the results for the external data set. Our results indicated that the beta regression method offers a noteworthy alternative to separate canopy cover measurements, especially if time is limited and the models can be applied in the same region where the modelling data were collected.

Korhonen, University of Joensuu, Faculty of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: lauri.korhonen@joensuu.fi
Korhonen, Finnish Forest Research Institute, Joensuu Research Unit, P.O. Box 68, FI-80101 Joensuu, Finland E-mail: ktk@nn.fi
Stenberg, Univ. of Helsinki, Dept of Forest Resource Management, P.O. BOX 27, FI-00014 University of Helsinki, Finland E-mail: ps@nn.fi
Maltamo, University of Joensuu, Faculty of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: mm@nn.fi
Rautiainen, Univ. of Helsinki, Dept of Forest Resource Management, P.O. BOX 27, FI-00014 University of Helsinki, Finland E-mail: mr@nn.fi

Click this link to register to Silva Fennica.

If you are a registered user, log in to save your selected articles for later access.

Contents alert

Your selected articles

Follow @SilvaFennica