Estimating Mediterranean forest parameters using multi seasonal Landsat 8 OLI imagery and an ensemble learning method

Chrysafis, Irene; Mallinis, Giorgos; Gitas, Ioannis Z.; Tsakiri–Strati, Maria

doi:10.1016/j.rse.2017.07.018

Cited by 67 publications

(57 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For example, this research indicated that the SWIR bands (e.g., Landsat TM spectral bands 5 and 7) play a more important role than visible and near-infrared bands. This conclusion is similar to that from previous studies on the moist tropical forests in the Amazon [4,45] and the Mediterranean forests [55]. The more important role of SWIR than near-infrared and visible spectral bands in AGB modeling may be due to the fact that SWIR is more sensitive to moisture and shade components inherent in the forest stand structure and that atmospheric conditions have less impact on spectral signatures than other shorter wavelength (e.g., near-infrared and visible) spectral bands.…”

Section: Selection Of Suitable Variables For Agb Modelingsupporting

confidence: 93%

Comparative Analysis of Modeling Algorithms for Forest Aboveground Biomass Estimation in a Subtropical Region

et al. 2018

View full text Add to dashboard Cite

Remote sensing-based forest aboveground biomass (AGB) estimation has been extensively explored in the past three decades, but how to effectively combine different sensor data and modeling algorithms is still poorly understood. This research conducted a comparative analysis of different datasets (e.g., Landsat Thematic Mapper (TM), ALOS PALSAR L-band data, and their combinations) and modeling algorithms (e.g., artificial neural network (ANN), support vector regression (SVR), Random Forest (RF), k-nearest neighbor (kNN), and linear regression (LR)) for AGB estimation in a subtropical region under non-stratification and stratification of forest types. The results show the following: (1) Landsat TM imagery provides more accurate AGB estimates (root mean squared error (RMSE) values in 27.7-29.3 Mg/ha) than ALOS PALSAR (RMSE values in 30.3-33.7 Mg/ha). The combination of TM and PALSAR data has similar performance for ANN and SVR, worse performance for RF and KNN, and slightly improved performance for LR. (2) Overestimation for small AGB values and underestimation for large AGB values are major problems when using the optical (e.g., Landsat) or radar (e.g., ALOS PALSAR) data. (3) LR is still an important tool for AGB modeling, especially for the AGB range of 40-120 Mg/ha. Machine learning algorithms have limited effects on improving AGB estimation overall, but ANN can improve AGB modeling when AGB values are greater than 120 Mg/ha. (4) Forest type and AGB range are important factors that influence AGB modeling performance. (5) Stratification based on forest types improved AGB estimation, especially when AGB was greater than 160 Mg/ha, using the LR approach. This research provides new insight for remote sensing-based AGB modeling for the subtropical forest ecosystem through a comprehensive analysis of different source data, modeling algorithms, and forest types. It is critical to develop an optimal AGB modeling procedure, including the collection of a sufficient number of sample plots, extraction of suitable variables and modeling algorithms, and evaluation of the AGB estimates.

show abstract

Section: Selection Of Suitable Variables For Agb Modelingsupporting

confidence: 93%

Comparative Analysis of Modeling Algorithms for Forest Aboveground Biomass Estimation in a Subtropical Region

et al. 2018

View full text Add to dashboard Cite

show abstract

“…While the number of trees (ntree) can be as many as possible (default value 100) due to the RF model being fast and not overfitting, in practice the best values for these parameters will depend on the problem and should be treated as tuning parameters. In this study, the number of regression trees grown (ntree) was adjusted based on the out-of-bag (OOB) estimated error, and the optimum number of random variables to be tested in each tree (mtry) was estimated using the tune RF function in the Random Forest package [48]. Initially, we built a RF model based on the complete set of data including all spatial, spectral, and topographic variables with varied parameter settings.…”

Section: Model Specification/parameter Settingmentioning

confidence: 99%

“…Since most GPS-based ground training data collected were biased to locations across from/close to the road network, an additional 600 to 700 (varies on images) stratified random samples were also taken as reference data [28]. As several studies suggest that nonparametric machine learning classifiers such as Random Forest required a large number of reference data to attain the most favorable outcome [48,52], we combined ground training data and randomly chosen samples in order to get enough information on the land cover, while maintaining the spatial distribution of our reference data throughout the study area. Later, large portions of these samples were used as input variables for the Random Forest classifier and some were retained and used for validation for the thematic map of 2015.…”

Section: Collection Of Reference Data and Variable Selectionmentioning

confidence: 99%

“…As a result, we calibrated our model based on a complete dataset, then used the variable importance results to tune parameter settings such as increasing the number of mtry. Using backward elimination processes, we curtailed the least important variables that were ranked far down the list of the variables of importance graph ( Figure 2) [48]. In this way, the final RF models were developed based on a reduced number of datasets for every single period of image classification.…”

Section: Collection Of Reference Data and Variable Selectionmentioning

confidence: 99%

See 1 more Smart Citation

Analyzing Land Cover Change and Urban Growth Trajectories of the Mega-Urban Region of Dhaka Using Remotely Sensed Data and an Ensemble Classifier

Hassan

Southworth

2017

Sustainability

View full text Add to dashboard Cite

Accurate information on, and human interpretation of, urban land cover using satellite-derived sensor imagery is critical given the intricate nature and niches of socioeconomic, demographic, and environmental factors occurring at multiple temporal and spatial scales. Detailed knowledge of urban land and their changing pattern over time periods associated with ecological risk is, however, required for the best use of critical land and its environmental resources. Interest in this topic has increased recently, driven by a surge in the use of open-source computing software, satellite-derived imagery, and improved classification algorithms. Using the machine learning algorithm Random Forest, combined with multi-date Landsat imagery, we classified eight periods of land cover maps with up-to-date spatial and temporal information of urban land between the period of 1972 and 2015 for the mega-urban region of greater Dhaka in Bangladesh. Random Forest-a non-parametric ensemble classifier-has shown a quantum increase in satellite-derived image classification accuracy due to its outperformance over traditional approaches, e.g., Maximum Likelihood. Employing Random Forest as an image classification approach for this study with independent cross-validation techniques, we obtained high classification accuracy, user and producer accuracy. Our overall classification accuracy ranges were between 85% and 97% with kappa values between 0.81 and 0.94. The area statistics derived from the thematic land cover map show that the built-up area in the 43-year study period expanded quickly, from 35 km 2 in 1972 to 378 km 2 in 2015, with a net increase rate of approximately 980% and an average annual growth rate of 6%. This growth rate, however, was higher in peripheral areas, with a 2903% increase and an annual expansion rate of 8%, compared to a 460% increase with an annual growth rate of 4% in the core city area (Dhaka City Corporation). This huge urban expansion took place in the north, northwest, and southwest regions of Dhaka, transforming areas that were previously agricultural land, vegetation cover, wetland, and water bodies. The main factors driving the city towards northern corridors include flood-free higher land, the availability of a transportation network, and the agglomeration of manufacturing-based employment centers. The resulting thematic map and spatial information produced from this study therefore serve to facilitate a detailed understanding of urban growth dynamics and land cover change patterns in the mega-urban region of Dhaka, Bangladesh.

show abstract

“…Remote sensing is an effective tool to estimate the stand density. Results from relevant literature are shown in Table 4, where the remote sensing data used included airborne [40] and terrestrial LiDAR [41], optical imagery [42][43][44], and SAR data [45]. The study reported by Lee and Lucas [40] is most directly comparable to ours as they also implemented airborne LiDAR data (Optech ALTM 1020), while the stand densities were estimated by computing the height-scaled crown openness index for LiDAR data of white cypress pine in the coniferous forest.…”

Section: Discussionmentioning

confidence: 85%

Estimating Stand Density in a Tropical Broadleaf Forest Using Airborne LiDAR Data

Lee

Wang

2018

Forests

View full text Add to dashboard Cite

Forest-related statistics, including forest biomass, carbon sink, and the prevention of forest fires, can be obtained by estimating stand density. In this study, a dataset with the laser pulse density of 225.5 pulses/m2 was obtained using airborne laser scanning in a tropical broadleaf forest. Three digital surface models (DSMs) were generated using first-echo, last-echo, and highest first-echo data. Three canopy height models (CHMs) were obtained by deducting the digital elevation model from the three DSMs. The cell sizes (Csizes) of the CHMs were 1, 0.5, and 0.2 m. In addition, stand density was estimated using CHM data and following the local maximum method. The stand density of 35 sample regions was acquired via in-situ measurement. The results indicated that the root-mean-square error (RMSE) ranged between 1.68 and 2.43; the RMSE difference was only 0.78, indicating that stand density was effectively estimated in both cases. Furthermore, regression models were used to correct the error in stand density estimations; the RMSE after correction was called RMSE′. A comparison of the RMSE and RMSE′ showed that the average value decreased from 12.35 to 2.66, meaning that the regression model could effectively reduce the error. Finally, a comparison of the effects of different laser pulse densities on the RMSE value showed that, in order to obtain the minimum RMSE for stand density, the laser pulse density must be greater than 10, 30, and 125 pulses/m2 at Csizes of 1, 0.5, and 0.2 m, respectively.

show abstract

Estimating Mediterranean forest parameters using multi seasonal Landsat 8 OLI imagery and an ensemble learning method

Cited by 67 publications

References 58 publications

Comparative Analysis of Modeling Algorithms for Forest Aboveground Biomass Estimation in a Subtropical Region

Comparative Analysis of Modeling Algorithms for Forest Aboveground Biomass Estimation in a Subtropical Region

Analyzing Land Cover Change and Urban Growth Trajectories of the Mega-Urban Region of Dhaka Using Remotely Sensed Data and an Ensemble Classifier

Estimating Stand Density in a Tropical Broadleaf Forest Using Airborne LiDAR Data

Contact Info

Product

Resources

About