Identification of age groups of managed Pine plantation using remote sensing data
Ranitha L. Ratnayake
Dept. of Geography
University of Kelaniya
Tel: 0094 112914489
Identification of different age groups is an important requirement for forest management. Forest cover can be estimated using Normalised Difference Vegetation Index (NDVI) from satellite data. The objective of this study was to identify changes of the forest cover in different age groups from the spectral reflectance measured by the satellites. This paper analyses the relationship between NDVI and different ages of managed pine plantation in the Kings Forest, East Anglia, UK. The satellite data used was - SPOT HRV multi-spectral image 1996. Secondary data on forest management operations were derived from compartment records supplied by the Forestry Commission and informal discussions with forest managers.
Age is an important variable in forest growth. Therefore this study investigated changes of NDVI in different age groups <7, 8-17, 18-37, 38-57 and >58 years. The study found that NDVI range from 0.445-0.747 in the pine stands. However, in each age group slight changes can be easily recognised. NDVI increased from 0.427 to 0.578 in the < 7 years and 0.706 to 0.738 in the 8 -17 years old pine stands. The results of the correlation analysis indicate a high positive correlation coefficient of 0.99 and 0.97 respectively. However, in the stands aged 18-37 and 38-57 years the increase with age showed an irregular pattern with more variations than in the very young stands. No significant correlation was found between NDVI and age in this group, and it shows the lowest correlation coefficient of 0.62 and 0.65 respectively. The next stage the old age group indicates positive high correlation (r =0.81) due to considerable increase of NDVI at the age of 60 years suggesting that background features may have contributed to the reflected signals.
When determining the spectral responses of forest canopies, forest cover plays a key role because it controls the amount of understorey vegetation, which is visible to the sensors (Nemani et al., 1993). Therefore, identifying, detecting and monitoring vegetation cover, particularly in forest, has become a major application of satellite remote sensing (Hall et. al., 1996; Sader and Winne, 1992). Remote sensing instruments measure the radiation flux for a unit area of the Earth surface and provide a spatially referenced representation of its surface. Measurements of radiation flux in the visible and near infrared portion of the electromagnetic spectrum have been used to calculate various vegetation indices, which have been used for both qualitative and quantitative assessments of vegetation cover. Vegetation indices are highly useful for detecting the amount of vegetation within the canopy cover because these indices enhance the contrast between the ground and vegetation. The most widely used vegetation index is the Normalised Difference Vegetation Index (NDVI), which is used this study.
There is an increasing need for accurate measurement of forest cover in terms of both its spatial and temporal variation for improvements in forest management. Therefore the main aim of this paper is to analyse the spatial changes of forest cover between different age groups using satellite-derived values of the Normalised Difference Vegetation Index (NDVI). The analysis focuses on NDVI changes in different types of stands within the same site. Correlation analysis was also used to examine the relationship between NDVI and the age of trees in the different compartments of the forest stand.
2. Site Description
In order to assess the changes of forest cover in different age groups a study area has been selected in the Kings Forest. It contains 213 compartments covering 2334 ha. with mainly coniferous trees, situated to the south of Thetford forest in East Anglia, UK (Latitude 50 0. 20’ North and Longitude 00 0. 38’ East). Kings Forest is a forest managed for the production of timber by the Forestry Commission, from whom maps and records were obtained. The forest is dominated by single species, even-aged compartments mainly of Corsican Pine (Pinus nigra var. maritima) and Scots Pine (Pinus sylvestris) with a much smaller number of compartments of broad-leaved species such as Beech (Fagus sylvatica) and an understory of brambles, bracken, broom and grass.
The pine plantation examined in this study consists of 34 compartments and 95 sub compartments covering approximately 360 ha. It is located in the northwest quadrant of the Kings forest.
3. Methodology and Data Set
This study mainly concerns vegetation indices based on radiometric measurements developed in remote sensing. A number of mathematical formulae using visible and near-infrared reflectance, called radiometric vegetation indices, have been proposed for relating radiometric measurements in the visible and near-infrared wavelength intervals to the amount of vegetation present. Here attention has focused on a commonly used index, the Normalised Difference Vegetation Index (NDVI). The basis for NDVI is related to the abrupt increase of vegetation
Where NIR represents the surface reflectance measured in the near-infrared spectral band and R represents the reflectance in the red spectral band. The visible and near-infrared reflectance may be obtained by remote measurements from a number of satellites such as LANDSAT, SPOT, and the NOAA series, after correction for calibration, variations in solar incidence and atmospheric distortions. In this study, NDVI values were extracted from SPOT images. To analyse the NDVI the following two data sets were assembled.
The four SPOT images were used to calculate the NDVI. It was SPOT HRV (path 33 and row 244) recorded on 15th June 1996. Seasonal variations is minimised, allowing the investigation of longer-term trend
- Satellite data - SPOT HRV multi-spectral data for, 1996
- Secondary data - Forest compartment records from the Forestry Commission.
4. Results and Discussion
The analysis concerns how forest cover changes spatially using NDVI. It is a quantitative measure, based on digital values, which provides a measure of vegetation vigour (Campbell, 1996). NDVI is determined by the difference between red and near-infrared reflection, which changes with the age of tree leaves. Pine needles become more spongy with the age, the mature leaves display less reflection in the visible band and more in the infrared. Moreover younger leaves have a thinner layer of the mesophyll and fewer cavities so that relatively less reflection occurs in the NIR region of the wave band, while red reflection occurs relatively more (Gibson and Power, 2000). NDVI is therefore expected to vary with the different ages in the forest compartments.
Figure 1 shows the density sliced NDVI image of the study site in 1996. It is displayed in different colours and emphasises differences of NDVI values range from a low of 0.235 to a high of 0.835. However in the pine compartments it ranges from 0.435- 0.535 .The areas with high NDVI values appear in green, yellow and red in the image, while low values are in magenta, dark blue and light blue. On the basis of the different NDVI values the study recognised four age groups (Figure 1).
Figure 1: Density sliced NDVI image in the study site – 1996
4.1. Comparison of NDVI by age groups
In this section analyses in detail NDVI changes occur between each age group. Figure 2 shows the NDVI image of the study site (part of Kings Forest pine plantation) in 1996. It is displayed in greyscale and emphasises the differences of NDVI values ranging from a minimum of 0.263 to a maximum of 0.831. The areas with high NDVI values appear in a lighter tone in the image, while low values are in darker tones.
According to the different tones in the main image the following three cover types were identified:
- Clear cut areas - low NDVI (0.263-0.366) with darker tone. (marked D in the image - Figure 2).
- Coniferous plantation (mainly pine trees) - moderate NDVI (0.562-0.747) with medium tone. (marked * A1, A2, A3 and A4).
- Broadleaf trees - high NDVI (0.778-0.835) with brighter tone C and D.
This study mainly concerns different ages of pine plantation. Because the canopy cover of a tree changes with its age, spatial changes of NDVI in the study site were identified on the basis of the following age groups (Figure 2).
- <7 years - Very young pine plantation - marked *
- 8 - 17 years - Young pine plantation - marked A 1,
- 18 - 37 years - Pre mature pine plantation - marked A2,
- 38 -57 years – Mature pine plantation –A3,
- >58 years – Old age pine plantation -marked A4,
C and D refer to non-pine areas.
In a one year-old plantation the spaces between the trees consist of dry grass with rough vegetation and soil. Therefore at this stage, very low values of NDVI (0.355-0.366) occurred. Until the increase of plant cover at the age of 8 years, NDVI changed very little. In contrast, in the 8-17 years age group (marked B in the images), NDVI changes with time are relatively greater. In the image this area appears in relatively brighter tones, due to the high value of NDVI. The evidence suggests the presence of a considerable amount of undergrowth, which remains until thinning treatment starts at 20 years but is not exposed to the sensor because the canopy cover is fully developed at this stage. However, high near-infrared reflectance from the canopy is responsible for the high NDVI. In the area marked C on the image, pine trees that were planted in the late 1960s are now in the 18-37 year age group and appear as a relatively brighter area with dark patches. At this stage the planted tree cover decreased with thinning operations and the canopy then expanded very slowly. In the mature stage (38-57 age group) NDVI decreased, due to felling.
Figure 2 Spatial Differences of NDVI in the study site – 1996
4.2 NDVI changes in the different age groups
4.2.1 Age < 7 year
Based on field observations, the interspaces of the planted trees (<1 years old) consist of soil or dry grass (Figure 3a) and there is incomplete canopy cover. At this stage the reflectance of newly planted trees does not seem to be a major factor affecting canopy reflectance. This is likely to become increasingly important after canopy closure, when background reflectance no longer dominates the remotely sensed signals.
Figure 3: Pine plantation in earlier stages during the field observations in 5th of June 2001
From 1-7 years, several other types of deciduous tree dominate the plantation blocks as the leaf area expands (Figure 3 b). It has been stated that in forest studies the leaf area is inversely related to red reflectance and positively related to near-infrared reflectance (Running et al., 1986, and Peterson et al., 1987). NDVI values increase rapidly (from 0.451 - 0 563) both with the increased age of the planted trees and the growth of the undergrowth vegetation cover. However, there is clear evidence that at this stage NDVI is controlled more by the presence of understory vegetation visible to the sensor than by the early stages of tree growth. As observed by Nilson and Peterson (1994), there seems to be more than one factor dominating stand reflectance during the early stages of stand growth.
4.2.2 Ages 8-17 years
During the early stage of this age group canopies are still open or only partially closed and most of the stands still have understorey vegetation. At this stage, the understorey dominated the changes of NDVI, so that it is very difficult to predict the forest structural situation from remotely sensed data.
Figure 4: 8-17 years old pine plantation
However during the later stages the contribution of reflectance from the ground surface is negligible because of the closed condition of the canopy (Figure 4). It can be stated that this age group is the most relevant for remote sensing application for forest management. Because the ground is not visible to the sensors, the effect of surface cover is dominant. High values (0.736, 0.741 in 1997 image) of NDVI are influenced by the significant contribution of reflectance from the top layer of the pine canopy. The undergrowth consists of a thin layer of grass cover, but the effect of this layer no longer dominates the remote sensed signals.
4.2.3 Ages 18-37 years
In this age group only 32-37 year old pine stands were found in the study area and most of the compartments had been thinned twice. The normal age of first thinning is around 25-30 years of age, second thinning is around 30-35 years of age but it varies depending on the species, yield class and the initial spacing of the stand. Initial canopy closure is disturbed by these operations, and canopy gaps are exposed to the sensors. Apart from thinning, some sub-compartments were significantly damaged by windthrow, so that second thinning was delayed (information collected from forest managers). Changes of NDVI with the increase of age therefore showed an irregular pattern. NDVI ranged from 0.725 in a 32 year stand to 0.732 in a 33 year stand, decreasing to around 0.729 in 37 year stand and 0.715 in a 38 year-old stand
Figure 5: 18-37 years old pine plantation
Figure 5 shows a picture of a pine plantation of 36 years age. It looks different from the other age group stands because of the thinning operation in which the number of planted pine trees is reduced, and the undergrowth becomes very thin, with most of the area covered with soil. The first thinning removes two adjacent rows in every six rows where the original spacing is comparatively close. The second thinning then removes two other adjacent rows between two rows. Where the stand was regularly thinned, there was little evidence of crown expansion of the pine trees, and leaf area tended to reduce. Therefore NDVI changes vary with canopy structure.
4.2.4 Ages 38-57 years
In mature pine stands canopy structure is relatively simple with a single canopy layer containing trees of similar size (Figure 6). It appears as a significantly brighter tone and is similar to undisturbed forest. During the field visit it was determined that these are fully-grown, and healthy plantations with few gaps between rows of planted trees. NDVI is comparatively high (0.713 to 0.27) with a slight peak at the age of 56 due to the effect of continuous canopy cover and the understorey grass cover.
Figure 6: 38-57 years old pine plantation
A sudden decline of NDVI values recorded at the age of 42 (NDVI = 0.709) is most likely due to the disturbance of canopy cover following partial felling. However variations of NDVI in this group are very small
4.2.5 Ages > 58 years
There are few compartments in the old age group in the study area. According to the Forestry Commission records these compartments were marked for clear or partial felling.
Figure7: > 58 years old pine plantation
Some compartments in the northeast corner of the study site were partially felled or wind damaged. Therefore those areas appear as trees with open canopy with increasing maturity the roughness of the canopy increases, but this stage shows a greater frequency of scattered gaps between the trees. Based on field observations, the old growth stands typically have a higher proportion of dead wood, especially in the upper canopy and the ground cover has less green vegetation. Therefore NDVI decreased to a relatively low NDVI ranging from 0.682-0.706.
4.3 The relationship between NDVI and age
Age is the most important variable in forest growth. Many other structural parameters change with the growth of a tree. For example the tree height of a stand increases with the growth stage. There is also an increase in the roughness of the forest canopy as a surface. However, the sub-compartments of the forest stands consist of even-aged trees.
Figure 8: Relationship between NDVI and age
This section of the analysis concerns how NDVI values vary with age in the 1996 image. The results of the correlation analysis show considerable low correlation between NDVI and age with r value of, 0.36 (Figure 8). More detailed examination of these scatter diagram shows that the rate of NDVI increase with age is greater in the younger (less than 18 year) age groups in all four images. Also the 18-37 year age group shows a more complex relationship between NDVI and age.
Therefore more detailed analysis was done in each age group. When age increases NDVI also increases in < 20 years old trees. However, NDVI did not increase continuously with the age, and the pattern changes during the mature stage, probably due to the reduction in stand density when thinning operations start at the age of 20 years. To find out the influences of thinning operations during the mature stage and the NDVI changes the study carefully considered the relationship between mean NDVI in selected year (1996) and the age of stands within each age group (Figure 9).
Figure 9: Relationship between mean NDVI in each age group - 1996
The results indicate that the very strong positive correlation coefficient of 0.98 0.97 in the <7 years and 8-17 years age groups respectively. This shows that when age increases green canopy cover increases. This is the general trend for initial growing stage of plantation crop. However, in the pre-mature and mature stages the correlation between age and NDVI also shows a positive relationship but the correlation coefficient is relatively less. It is 0.62 and 0.66 in 18-37 year age group and 38-57 year age group respectively. The result is probably due to the reduction of canopy cover during the thinning period. It is clearly shown in figure 5.10 that at the age of 36 years NDVI decreases and this trend continues during the later pre-mature stage until values stabilise in the mature stage. During the mature stage NDVI changes are much smaller; this may be effect of a gradual increase in the shade factor. Sabol (2002) stated that in the mature coniferous plantation during this stage shade and shadowing both increase and it affected green vegetation factor.
Variations within this age group are comparatively large, which may be due to shadowing effects. Further investigations would be needed to confirm this. However, Danson (1995) has stated that pine stands of this age consist of trees whose crowns are less compact with relatively low needle density, so that as a consequence shadows are less pronounced. Results of the correlation analysis indicated a strong (r =0.81) relationship between NDVI and age. It can be stated that there is a gradual decline of NDVI until 60 years. However, after 60 years there is a considerable increase of NDVI most probably due to higher proportion of scattered regrowth in the partially felled areas.
Slight changes in NDVI can be identified due to differences in age of the pine trees. For example moderately low NDVI of 0.594 are found in very young pine compartments and moderately high NDVI of 0.710 in mature pine compartments. The results indicate that NDVI is significantly affected by the background influences of the forested region, including the soil, and the undergrowth.
This paper also demonstrates changes of NDVI with different age groups in 1996. NDVI increased from 0.427 to 0.578 in the < 7 years and 0.706 to 0.738 in the 8 -17 years old pine stands The results of the correlation analysis indicate a high positive correlation coefficient of 0.99 and 0.97 respectively. However, in the stands aged 18-37 and 38-57 years the increase with age showed an irregular pattern with more variations than in the very young stands. No significant correlation was found between NDVI and age in this group, and it shows the lowest correlation coefficient of 0.62 and 0.65 respectively. The next stage the old age group indicates positive high correlation (r =0.81) due to considerable increase of NDVI at the age of 60 years suggesting that background features may have contributed to the reflected signals.
Pine plantations at the mature stage show relatively high NDVI values of 0.763 and 0.781 due to the canopy roughness and also shadowing effects. After clear felling it decreased dramatically to a range of 0.235- 0.319. Finally, it can be noted that temporal changes of NDVI with in four years period is very little. However, significantly high positive correlation occurs between NDVI and age in every year.
I wish to express my gratitude to my research supervisors Prof. Michael Steven and Prof. Charles Watkins (School of Geography, University of Nottingham UK) for their suggestions and criticisms to this paper. I also wish to thank Mr Martin Johnson and Mr. Giles Drake-Brockman (Forest Enterprise, East Anglian region UK) for providing the stock maps and management records from Kings Forest.
- Campbell, J.B. (1996) Remote Sensing applications in the plant sciences Introduction to Remote Sensing: Ch. 13 pp .366 – 403, The Guilford Press.
- Danson, F.M. (1995) “Developments in the Remote Sensing of forest canopy structure.” In Advances in Environmental Remote Sensing. Edited by F.M. Danson and S.E. Plummer, pp. 52-69, John Wiley & sons Ltd.
- Gibson, P.J. and Power, C.H. (2000) Introductory Remote Sensing Digital Image Processing and application. St. Edmunds bury Press, Suffolk.
- Hall, F.G., D.R. Peddel and E.F. Drew (1996) “Remote sensing of Biophysical Variables in Boreal Forest Stands of Picea mariana.” Int. J. Remote Sensing 17: pp. 3077-3081.
- Nemani, R.R.; L. Pierce; S. Running and L. Band (1993) “Forest ecosystem process at the water shed scale: Sensitivity to remote sensing leaf area index estimate.” Int. J. Remote Sensing 14: pp. 2519-2534.
- Nilson, T. and U. Peterson (1994) “Age dependence of forest reflectance: Analysis of Main Driving factors.” Remote Sens. Environ. 48: pp. 319-331.
- Peterson, D.L.; M.A. Spanner; S.W. Running and K.B. Tenber (1987) “Relationship of Thematic Mapper Simulator data to Leaf Area
- Running, S.W.; D.L. Peterson; M.A. Spanner and K.B. Teuber (1986) “Remote Sensing coniferous forest leaf area.” Ecology 67 (1): pp. 1273-1276.
- Sader, S.A. and J.C. Winne (1992) “RGB-NDVI colour composite for visualising forest change dynamics.” Int. J. Remote Sensing 13: pp. 3055-3607
- Sabol, D.E.; A.R. Gillespie; J.B. Adams; M.O. Smith and C.T. Tucker (2002) “Structural stage in Pacific North West forest estimates using simple mixing model of multi-spectral images.” Remote Sens. Environ. 80,: pp. 1-16.