Search

Landsat time series images are used for the detection of forest disturbance and the classification of causal agents. Various studies have classified disturbance agents with respect to forest disturbance detected using Landsat time series images. However, the accuracy of the finally classified disturbance agents in different approaches is rarely evaluated. In this study, we investigated the effectiveness of using ensemble classification, and multiple spectral and spatio-temporal information for the accuracy of the classification of disturbance agents in two-stage prediction (i.e., disturbance agents are classified with respect to the detected disturbance) and direct prediction (i.e., disturbance agents are directly classified from Landsat temporal information). Predictor variables were derived from the results of the trajectory-based temporal segmentation of five spectral indices using an annual Landsat time series (2000–2018). We compared six approaches of classifying disturbance agents. For two-stage prediction, we investigated four disturbance detection approaches: threshold-based detection with a single spectral index, random forest (RF) model with a single spectral index, RF model with multiple spectral indices, and RF model with spatio-temporal variables. The detected disturbance pixels were aggregated to disturbance patches and classified into disturbance agents. For direct prediction, two RF models one with only temporal variables and the other with spatio-temporal variables were constructed to classify pixel-based disturbance agents. The overall accuracy of the RF model using spatio-temporal variables for direct prediction was 92.4% and significantly higher than that of the RF model for two-stage prediction (90.9%). The use of an RF model based only on a single spectral index in disturbance detection was not effective for improving accuracy compared with threshold-based detection; however, the use of an RF model based on multiple spectral indices in disturbance detection improved the accuracy of the final classification of disturbance agents. Introducing spatial variables in RF models was effective for improving the overall classification accuracy in pixel-based direct prediction. However, it was not necessary in two-stage prediction because of spatial information contained in the patches. Although a spatially discontinuous appearance was observed for the RF model for directly classifying disturbance agents, this could be an alternative approach to two-stage prediction when considering the relative classification performance and simplicity of implementation.

This study quantified the effect of weeding frequency and weeding schedules on weeding operation time in a sugi (Cryptomeria japonica) plantation stand. A weeding operation time estimation model was developed; then the cumulative weeding operation time after six growing seasons was simulated using the developed model. The developed model included weed height, relative height of weeds to sugi, and initial planting density. The simulated cumulative weeding operation time decreased approximately 6% for each one-treatment decrease in weeding frequency. Under a three-treatment weeding frequency scenario, the simulated cumulative operation time when weeding was conducted during non-consecutive years was longer than that when weeding was conducted during three consecutive years. The results suggest that carrying out weeding treatment during consecutive years is the more effective for reduction of weeding costs. We conclude that weeding schedule as well as weeding frequency must be considered for reduction of weeding operation time.

Selective logging is one of the factors contributing to deforestation and forest degradation in tropical forests. A low-cost methodology to monitor selective logging is clearly required. However, this poses a challenge because only a few trees are felled at a given time. Here, we investigate the potential of using repeatedly acquired digital aerial photographs (DAPs) from a lightweight unmanned aerial vehicle (UAV) to detect selective logging in tropical forests in Myanmar. Selective logging was conducted within two 9-ha plots. DAPs were acquired immediately before and after selective logging using a lightweight UAV in this case study. The aboveground biomass (AGB) change related to selective logging was regressed against metrics expressing forest changes calculated at a 0.25-ha resolution from a photogrammetric point cloud created using the DAPs before and after selective logging. The root-mean-square error and coefficient of determination were 0.77 and 9.32 Mg/ha, respectively. This study demonstrates that repeated DAPs taken from a lightweight UAV can be used to estimate changes in the AGB linked to selective logging. This method could be used to quantify the impacts of both legal selective logging and illegal logging in tropical forests.

SummaryIncreasing attention has been given to evaluating the effectiveness of forest conservation projects, but it is not well known whether different methods yield similar results when evaluating changes in forest resources. The present study compares forest cover maps and local people’s perceptions for evaluating the effectiveness of the Chambok community-based ecotourism (CBET) programme in Cambodia. We assessed forest cover changes from 2000 to 2012 using published global maps and used a covariate matching method to compare forest sites in CBET and non-CBET areas. We also analysed local people’s perceptions of forest resource changes by interviewing 174 households. The forest cover maps showed that the Chambok CBET was effective at reducing deforestation, although the outcome was not completely robust to unobserved heterogeneity. Local people’s perceptions concurred with the effectiveness observed in the forest cover maps, in that 64% of the people perceived that forest resources increased and 75% thought that the local community could protect its own forest resources. We conclude that the Chambok CBET performed effectively for forest conservation and suggest that mixed-method approaches are essential for evaluating the effectiveness of conservation programmes.

Community-based ecotourism (CBET) has become a popular tool, especially in developing countries, for biodiversity conservation and livelihood improvement, but there is a lack of studies that use quantitative data to evaluate the effectiveness of CBET using a before-after-control-intervention (BACI), BA, or CI design. We conducted a case study of the flagship Chambok CBET program in Cambodia to quantify the contribution of CBET to household income and livelihood changes for CBET and non-CBET members. We conducted an interview survey of 173 households (77 CBET and 96 non-CBET members) that were selected systematically from every two households along roads within the same villages. There was no significant difference between the total income of member and non-member households; the median CBET income was 1.22 USD per month, which is only 1.65% of the total income. The inequality of income from CBET among the CBET members was higher than that from other income sources. The perceptions of livelihood changes differed considerably before and after the establishment of the Chambok CBET program, although this difference was not attributed to CBET but rather to general socioeconomic changes in the country. We conclude that it is challenging to employ CBET to achieve poverty reduction and livelihood improvement.

Firewood and charcoal are the main energy sources in developing countries, but much fewer quantitative data have been collected for charcoal consumption. This study compared firewood and charcoal consumption rates and patterns in a rural area of Yedashe Township, Myanmar. Household interviews were conducted for randomly selected households, resulting in 147 firewood users and 34 charcoal users. Forest inventory data was used to estimate forest area needed to meet woodfuel demand. Average per capita consumption rates were 780 and 280 kg year−1 for firewood and charcoal, respectively. Distinct differences were found in fuel sources and sizes; only 16% of firewood was collected from living trees in natural forests, of which 72% was from trees

This study evaluated the uncertainty of individual tree biomass estimated by allometric models by both including and excluding tree height independently. Using two independent sets of measurements on the same trees, the errors in the measurement of diameter at breast height and tree height were quantified, and the uncertainty of individual tree biomass estimation caused by errors in measurement was calculated. For both allometric models, the uncertainties of the individual tree biomass estimation caused by the use of a specific allometric model were also calculated. Finally, the overall uncertainty of individual tree biomass by combining the two uncertainties was calculated. The allometric model including tree height was 6 % more accurate than that excluding tree height when the uncertainty caused by allometric models became the only consideration. However, in terms of the uncertainty caused by measurement, the allometric model excluding tree height was three times more accurate than allometric model including tree height. As a result, the allometric model excluding tree height was 5 % more accurate than the allometric model including tree height when both causes of uncertainty, the allometric model and measurement errors were considered. In conclusion, errors in tree height measurement have the potential to increase the error of aboveground biomass estimation.

This study investigated the effect of weeding on the height growth of planted sugi (Cryptomeria japonica) saplings in southwestern Japan. First, we developed stand-level sugi and weed annual height growth models that accounted for tree spacing, slope aspect, and mean sugi height at the end of prior growing season and mean weed height just after weeding. Then, we simulated sugi heights after the sixth growing season under 64 different weeding schedules, using the developed model to examine the effects of the weeding frequency and the timing of weeding. The selected model for sugi height growth was expressed in terms of mean sugi height, mean weed height, slope aspect, and the relative height of weeds to sugi. In the case of weeds, the selected model was expressed in terms of mean sugi height, slope aspect, and the relative height of weeds to sugi. The simulation showed that sugi height decreased approximately 20% for every one-time reduction in the number of weeding. The simulation also showed that...

Logging damage to residual trees is an important parameter when considering the sustainability of selective logging in tropical natural forests. Here, we applied a proposed tree-based approach to e...

Detecting forest disturbances is an important task in formulating mitigation strategies for deforestation and forest degradation in the tropics. Our study investigated the use of Landsat time series imagery combined with a trajectory-based analysis for detecting forest disturbances resulting exclusively from selective logging in Myanmar. Selective logging was the only forest disturbance and degradation indicator used in this study as a causative force, and the results showed that the overall accuracy for forest disturbance detection based on selective logging was 83.0% in the study area. The areas affected by selective logging and other factors accounted for 4.7% and 5.4%, respectively, of the study area from 2000 to 2014. The detected disturbance areas were underestimated according to error assessments; however, a significant correlation between areas of disturbance and numbers of harvested trees during the logging year was observed, indicating the utility of a trajectory-based, annual Landsat imagery time series analysis for selective logging detection in the tropics. A major constraint of this study was the lack of available data for disturbances other than selective logging. Further studies should focus on identifying other types of disturbances and their impacts on future forest conditions.

We quantified the differences in growth responses between Cryptomeria japonica and Chamaecyparis obtusa trees in approximately 0.1-ha group selection openings in southern Japan. We measured the position, height (H), and diameter at ground level (D) of 6- and 13-year-old C. japonica and C. obtusa trees that were planted in two respective openings. The annual H and D growth of C. japonica increased until trees were 10 and 8 m, respectively, distant from the nearest edge of the opening. For C. obtusa, the annual H and D growth increased slightly until the trees were 6 m distant from the nearest edge, but they did not significantly differ between 6 and 18 m distant from the nearest edge. While the shortest distance from the edge explained 63% and 50% of the variationsal H and D growth, respectively, of C. obtusa. Our result demonstrates that C. obtusa is less sensitive to edge effects than C. japonica. We conclude that C. obtusa is likely to be more suitable than C. japonica for planting within 0.1-ha...

Estimation of the current successional status of long-lived coniferous forests is difficult especially after several hundred years of human disturbance. However, in Cryptomeria japonica forest on Yakushima Island, the stumps generated by logging activities from the seventeenth to nineteenth centuries remain almost intact without intense rotting. From an inventory of these stumps and long-term monitoring of current stands for more than 25 years, we estimated the forest condition before logging and evaluated the current status of secondary succession. The size structure of Cr. japonica stumps suggested the stumps were composed of two different generations: the stumps of trees present before initiation of logging; and those of trees that regenerated after initiation of logging. In the current forest, the stem number of canopy species is decreasing and that of understory broad-leaved species is increasing. Regenerating Cr. japonica individuals are rare. Before the logging, the forest contained a large...

SUMMARY Reduced-impact logging (RIL) is widely expected to maximize conservation values of selectively logged tropical forests; however, there remains a lack of supporting data to confirm the effectiveness of individual RIL practices. This study evaluates the extent of damage to residual stands and soil caused by directional felling, elephant skidding, and road construction in a tropical mixed deciduous forest under the Myanmar Selection System (MSS). The felling damage number was consistently larger for bamboo clumps than for trees over the range of felled tree size and felling intensity. Soil disturbed by road construction made up 4.6% of the 9-ha study area, but no visible damage to residual trees and soil from elephant skidding was found three months after the operation. Directional felling toward bamboos and elephant skidding of MSS are effective as RIL practices, producing the lowest level of damage to residual trees and soil as compared with other RIL studies.

In this study, we evaluated the effects of topographic correction and gap filling of Landsat Enhanced Thematic Mapper Plus ETM+ images on the accuracy of forest change detection through a trajectory-based approach. Four types of Landsat time series stacks LTSS were generated. These stacks resulted from combinations of topographically corrected and uncorrected imagery combined with gap-filled and unfilled stacks. These combinations of stacks were then used as input into a trajectory-based change detection. The results of change detection from trajectory-based analysis using these LTSS were compared in order to assess the effects of both topographic correction and gap-filling procedures on the ability to detect forest disturbances. The results showed that overall accuracies of change detection were improved after gap filling 10.5% and 7.5%, but were only slightly improved after topographic correction 3.6% and 0.6%. Although the gap-filling process introduced some uncertainty that might have caused false change detection, the number of pixels whose detection of disturbance was enhanced after gap filling exceeded those detecting false change. The results also showed that the topographic correction did not contribute much to improve the change detection in this study area. However, topographic correction has a potential to increase the accuracy of change detection in areas of more rugged terrain and steep slopes. This is because a direct relationship between the slope of the topography with topographic correction and an enhanced detection of disturbance in pixels from year to year was observed in this study. For robust change detection, we recommend that a gap-filling process should be included in the trajectory-based analysis procedures such as the one used in this study where a single image per year is used to characterize change. We also recommend that in areas of rugged terrain, a topographic correction in the image pre-processing should be implemented.

Appropriate cutting cycles and annual allowable cuts are crucial to ensure sustainability of tropical selective logging, but there have been limited field data to verify long-term effects of different cutting cycles. This study reveals some evidence of forest degradation in selectively logged production forests of Myanmar, which are subject to inappropriate cutting frequency. We compared stand structure, commercial species composition, and incidence of illegal logging between two compartments with low (LCF; 1 time) and high (HCF; 5 times) cutting frequency over a recent 18 years. Prior to the latest cutting, LCF had 176 trees ha − 1 with an inverted-J shape distribution of diameter at breast height (DBH), including a substantial amount of teak ( Tectona grandis ) and other commercially important species in each DBH class. HCF prior to the latest cut had only 41 trees ha − 1 without many commercially important species. At HCF, nearly half the standing trees of various species and size were illegally cut following legal operations; this was for charcoal making in nearby kilns. At LCF, two species, teak and Xylia xylocarpa , were cut illegally and sawn for timber on the spot. More extensive and systematic surveys are needed to generalize the findings of forest degradation and illegal logging. However, our study calls for urgent reconsideration of logging practices with high cutting frequency, which can greatly degrade forests with accompanying illegal logging, and for rehabilitating strongly degraded, bamboo-dominated forests. To reduce illegal logging, it would be important to pay more attention on a MSS regulation stating that logging roads should be destroyed after logging operations.

We developed an empirical model to estimate aboveground carbon density with variables derived from airborne Light Detection and Ranging (LiDAR) in tropical seasonal forests in Cambodia, and assessed the effects of LiDAR pulse density on the accurate estimation of aboveground carbon density. First, we tested the applicability of variables used for estimating aboveground carbon density with the original LiDAR pulse density data (26 pulse m−2). Aboveground carbon density was regressed against variables derived from airborne LiDAR. Three individual height variable models were developed along with a canopy density model, and three other models combined canopy height and canopy density variables. The influence of forest type on model accuracy was also assessed. Next, the relationship between pulse density and estimation accuracy was investigated using the best regression model. The accuracy of the models were compared based on seven LiDAR point densities consisting of 0.25, 1, 2, 3, 4, 5 and 10 pulse m−2. The best model was obtained using the single mean canopy height (MCH) model (R 2 = 0.92) with the original pulse density data. The relationship between MCH and aboveground carbon density was found to be consistent under different forest types. The differences between predicted and measured residual mean of squares of deviations were less than 1.5 Mg C ha−1 between each pulse density. We concluded that aboveground carbon density can be estimated using MCH derived from airborne LiDAR in tropical seasonal forests in Cambodia even with a low pulse density of 0.25 pulse m−2 without stratifying the study area based on forest type.

Community forestry (CF) is increasingly used in developing countries to achieve both the socioeconomic outcome of poverty reduction and an ecological outcome. There have been many single case studies in a specific region to identify the factors affecting the success or failure of CF. Other studies have used large-N data collected from multiple countries. However, there is a dearth of large-N studies within a single country. In this study, we used a country scale dataset of 197 CF projects, established between 1994 and 2005 across Cambodia, to identify the biophysical factors that affected forest cover changes from 2005 to 2016. A mixed-effects logistic regression model was used for a total of 71,252 randomly sampled data pixels nested in the 197 CF. Results showed that deforestation in CF was likely to increase with increasing size of CF area at lower elevations and on gentler slopes. Deforestation also increased if CF was located close to villages, markets and CF boundaries, but further away from main roads. These findings on biophysical factors can help the government to decide on priority locations for further conservation interventions or for the establishment of new CF projects.

Monitoring forest changes based on numerous satellite images has been recently conducted in the tropics. Preparation of a time series of satellite images, sometimes referred to as preprocessing, is essential for conducting robust detection of forest change. To create consistent and stable conditions in satellite images, the best methods have to be used in each step to correct various sources of noise. This study assessed three atmospheric correction methods, six topographic correction methods, and eight gap-filling methods to produce the best possible time series of Landsat images of tropical seasonal forests. The results showed that the best methods for atmospheric and topographic correction were relative corrections using a Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS), which is based on a 6S radiative transfer model, and the C-correction, respectively. Weighted linear regression and multiple linear regression models were selected as the best models for the filling of data gaps associated with Scan Line Corrector-off and clouds, respectively. This study provided the best possible image preprocessing for trajectory-based change detection using annual Landsat images. Although the best possible preprocessing methods might vary depending on the change detection methods used in different study areas, the results highlight the preferable preprocessing methods, even for different types of time series analysis.

We investigated the capabilities of a canopy height model (CHM) derived from aerial photographs using the Structure from Motion (SfM) approach to estimate aboveground biomass (AGB) in a tropical forest. Aerial photographs and airborne Light Detection and Ranging (LiDAR) data were simultaneously acquired under leaf-on canopy conditions. A 3D point cloud was generated from aerial photographs using the SfM approach and converted to a digital surface model (DSMP). We also created a DSM from airborne LiDAR data (DSML). From each of DSMP and DSML, we constructed digital terrain models (DTM), which are DTMP and DTML, respectively. We created four CHMs, which were calculated from (1) DSMP and DTMP (CHMPP), (2) DSMP and DTML (CHMPL), (3) DSML and DTMP (CHMLP), and (4) DSML and DTML (CHMLL). Then, we estimated AGB using these CHMs. The model using CHMLL yielded the highest accuracy in four CHMs (R2 = 0.94) and was comparable to the model using CHMPL (R2 = 0.93). The model using CHMPP yielded the lowest accuracy (R2 = 0.79). In conclusion, AGB can be estimated from CHM derived from aerial photographs using the SfM approach in the tropics. However, to accurately estimate AGB, we need a more accurate DTM than the DTM derived from aerial photographs using the SfM approach.

The quality of diameter measurements for natural old-growth forest was evaluated by comparing repeated field measurements in a diameter census and tree-ring chronologies. The diameter census was repeated three times from 1973 at 10- to 19-year intervals in old-growth Cryptomeria japonica forests on Yakushima Island, Japan. The diameter growth in the three intervals between (1) first and second measurement (15–19 years), (2) second and third measurement (10–13 years), and (3) first and third measurement (28–30 years) were calculated. The diameter growth of each interval was also measured using tree-ring chronologies as a control. There were significant differences in paired t tests for diameter growth calculated from the diameter census data and from the tree-ring chronologies when the interval was 27 years. The average percentage differences (APDs) were relatively high for every interval and for all diameter classes, but the APDs for the 28- to 30-year intervals were better than the 10- to 19-year intervals. The APDs followed a rising trend with increasing diameter at breast height for every interval. In conclusion, data from an interval of >27 years is required for accurate estimation of diameter growth. Furthermore, extra care should be taken when measuring large diameter trees. The results from this study demonstrated the difficulty of using a diameter census in natural old-growth forest.

To clarify the effects of pre-logging forest type (hinoki plantation or lucidophyllous forest) and harvesting roads on tree regeneration, we examined the density of regenerated trees, physical properties of soil, and soil erosion 2 years after clear-cutting in northern Kyushu, Japan. The results showed that the density of regenerated individuals was affected by the forest type before logging and by road construction. In a clear-cut lucidophyllous forest, most of the recovering vegetation consisted of lucidophyllous species originating from advanced regeneration and seedlings of pioneer species. Conversely, since few individuals originated from advanced regeneration in the clear-cut hinoki plantation sites, recovery to climax stand conditions at these sites is likely to be delayed compared to areas of clear-cut lucidophyllous forest. Few regenerated individuals were observed on harvesting road surfaces. Similarly, vegetation recovery was limited on harvesting road surfaces and along the side-cast slopes adjacent to roads. The number of germinated seedlings of pioneer species was positively correlated with litter cover and thickness of the A horizon, and negatively correlated with the rate of apparent soil erosion. These findings indicated that the establishment of seedlings of pioneer species originating from buried seeds is affected by the thickness of the A horizon of the soil. Since the recovered vegetation and the leaf litter derived therefrom will inhibit soil erosion, we propose that advanced regeneration and soil physical properties immediately after clear-cutting are important for the rapid recovery of climax forest conditions and for the prevention of soil erosion after clear-cutting.

In this study, we test and demonstrate the utility of disturbance and recovery information derived from annual Landsat time series to predict current forest vertical structure (as compared to the more common approaches, that consider a sample of airborne Lidar and single-date Landsat derived variables). Mean Canopy Height (MCH) was estimated separately using single date, time series, and the combination of single date and time series variables in multiple regression and random forest (RF) models. The combination of single date and time series variables, which integrate disturbance history over the entire time series, overall provided better MCH prediction than using either of the two sets of variables separately. In general, the RF models resulted in improved performance in all estimates over those using multiple regression. The lowest validation error was obtained using Landsat time series variables in a RF model (R2 = 0.75 and RMSE = 2.81 m). Combining single date and time series data was more effective when the RF model was used (opposed to multiple regression). The RMSE for RF mean canopy height prediction was reduced by 13.5% when combining the two sets of variables as compared to the 3.6% RMSE decline presented by multiple regression. This study demonstrates the value of airborne Lidar and long term Landsat observations to generate estimates of forest canopy height using the random forest algorithm.

Globalization and urbanization have significantly increased the food and non-food commodity demand for the last century, and it is vital to consider a business strategy with economical and ecological sustainability. The objective of this study was to project the contribution of wood to the financial performance of rubber plantations. We adopted cost and revenue data of rubber plantations in Cambodia and utilized land expectation value (LEV) as the criterion for profitability analysis. Among the top-ten rubber-producing countries in pan-tropics, the areas of rubber plantation were equivalent to from 1% to 90% of forest plantations and 0.3%–10.2% of total forest areas. Rubberwood revenue accounts for about 4%–10% of the 30th year LEV in rubber plantations at discount rates of 2% y−1–10% y−1, and this was sufficient to cover the cost of re-establishing the plantations. The proportion of the 30th year LEV contributed by wood revenue increased under conditions normally associated with a more difficult business environment, i.e., at higher wage costs, and lower latex revenue. We found that the wood revenue can improve the profitability of rubber plantations by up to 40% depending on the price of the rubberwood. We assert that timber from wood producing commodity plantations should be encouragingly utilized as industrial timber by linking the wood production in the management strategy of the plantations.

The attitudes and perceptions of local people are important for long-term survival of protected areas. Some studies have shown that people's perceptions and attitudes are shaped by knowledge about protected area goals and characteristics and related to socioeconomic factors. In this study, socioeconomic status, knowledge of protected area characteristics, perceptions of benefits and costs of protected areas, and attitudes toward two management programs were examined for people in 208 households from 14 villages around Popa Mountain Park, Central Myanmar. Approximately 50% of respondents had basic knowledge about the park; 38% perceived benefits and 45% perceived losses due to the park. Approximately half held positive attitudes toward buffer zone establishment and replacement of banana plantations. Logistic regression revealed that perceptions of benefits and positive attitudes toward management were correlated with both sociodemographic characteristics and knowledge, while perceptions of losses and negat...

Deforestation and forest degradation in the tropics have importance to the global carbon budget and biodiversity conservation disproportionately greater than the area concerned. Many studies have examined the patterns and processes of deforestation, but information about factors influencing forest degradation is still limited. In this study, we examined the factors influencing both deforestation and forest degradation in production forests of the central Bago Mountain area, Myanmar, which have been managed under the Myanmar Selection System (MSS) since 1856. We used forest canopy density maps derived from 1989 and 2006 satellite imagery together with environmental factors, locational attributes and selective logging records in logistic regression models. Our results showed that elevation and distance to the nearest town strongly influenced the likelihood of both deforestation and forest degradation, while logging intensity and distance to the nearest village were correlated only with the likelihood of forest degradation and not deforestation. We conclude that selective logging in this study area does not cause forest degradation provided that logging intensity is below the allowable cut prescribed under the MSS, but above that intensity, the likelihood of forest degradation increases markedly.

We evaluated the influence of texture information from remote sensed data on the accuracy of forest type classification at different spatial resolutions. We used 4-m spatial resolution imagery to create five different sets of imagery with lower spatial resolutions down to 30 m. We classified forest type using spectral information alone, texture information alone, and spectral and texture information combined at each spatial resolution, and compared the classification accuracy at each resolution. The classification and regression tree method was used for classification. The accuracy of all three tests decreased slightly with lower spatial resolution. The accuracy with the combined data was generally higher than with either the spectral or texture information alone. At most resolutions, the lowest accuracy was with texture information alone. However, there was no clear difference in accuracy between the combined data and spectral data alone at 25- and 30-m spatial resolution. These results indicate that adding texture information to spatial information improves the accuracy of forest type classification from very high resolution (4-m spatial resolution) to medium resolution imagery (20-m spatial resolution), but this accuracy improvement does not appear to hold for relatively coarse resolution imagery (25- to 30-m spatial resolution).

Although a number of image classification approaches are available to estimate forest canopy density FCD using satellite data, assessment of their relative performances with tropical mixed deciduous vegetation is lacking. This study compared three image classification approaches – maximum likelihood classification MLC, multiple linear regression MLR and FCD Mapper – in estimating the FCD of mixed deciduous forest in Myanmar. The application of MLC and MLR was based on spectral reflectance of vegetation, whereas FCD Mapper was operated on integrating the biophysical indices derived from the reflectance of the vegetation. The FCD was classified into four categories: closed canopy forest CCF; FCD ≥ 70%, medium canopy forest MCF; 40% ≥ FCD < 70%, open canopy forest OCF; 10% ≥ FCD < 40% and non-forest NF; FCD < 10%. In the three classification approaches, producer's and user's accuracies were higher for more homogeneous vegetation such as NF and CCF than for heterogeneous vegetation density VD such as OCF and MCF. FCD Mapper produced the best overall accuracy and kappa coefficient. This study revealed that only spectral reflectance is not enough to get good results in estimating FCD in tropical mixed deciduous vegetation. This study indicates that FCD Mapper, an inexpensive approach because it requires only validation data and thus saves time, can be applied to monitor tropical mixed deciduous vegetation over time at lower cost than alternative methods.