OK TEDI MINING LIMITED ENVIRONMENT DEPARTMENT

Transcription

1 OK TEDI MINING LIMITED ENVIRONMENT DEPARTMENT LOWER OK TEDI AND MIDDLE FLY VEGETATION DIEBACK MONITORING A.R.MARSHALL August 2004 Andrew Marshall Pty Ltd, Geomatic and Environment Consultants 43 Warrangarree Drive, Woronora Heights, NSW, Australia Phone: amara@oz .com.au

2 EXECUTIVE SUMMARY This report details the results of a vegetation dieback mapping program in the Ok Tedi and Middle Fly floodplains of the Western Province of Papua New Guinea. The investigation was undertaken in order to determine the magnitude and spatial extent of vegetation impact for the period to 30 th June Imagery was acquired on the 12 th February 2004 from the Landsat 5 TM satellite. Vegetation dieback extent totalled 1,554 square kilometres, including 459 square kilometres of vegetation recovery. These figures represented a 2% increase in total area and a 33% increase in recovery area with respect to the 2003 epoch of monitoring. The following figure shows the dieback and recovery estimates for the Ok Tedi and Middle Fly systems. Vegetation Dieback and Recovery 2000 Area (Sq Km) Total Dieback Recovery Year Although it is clear that that rate of dieback increase has reduced significantly since 2001 and that the rate of recovery has increased in the same period it is possible that these trends will not continue into the future. Climatic conditions similar to those of the 1999 to 2000 La Nina could readily increase the extent of vegetation dieback, decrease the extent of dieback recovery and possibly convert recovered areas back into dieback or stress conditions. In the Lower Ok Tedi it evident that the maximum extent of dieback is being approached and that the future maximum extent is a conservative estimate. In the Middle Fly the measured dieback extent is following the predicted trend. While it is probable that the lower than predicted result are due to dryer climatic conditions than in earlier epochs, particularly 1999 to 2000, the maximum extent predictions allow for future La Nina events. Based on the results of the 2004 monitoring epoch the maximum dieback extent of 2,395 square kilometres is believed to be a conservative estimate. The following figures show the relationship between observed dieback and predicted maximum extents for both the Ok Tedi and the Middle Fly Page 2 of 26

3 Lower Ok Tedi Dieback 250 Extent (Sq Km) Actual Dieback Future (Likely-Dredge) Extent (248 sq km) Projected Dieback Year Middle Fly Dieback Extent (Sq Km) Actual Dieback Future (Likely-Dredge) Extent (2147 sq km) Projected Dieback Year The majority of the Ok Tedi floodplain is under some form of recovery although the backwater zones are still subject to severe dieback. Approximately 70% of the Ok Tedi floodplain is in a recovery phase. The minimal increase in recovery extent in this zone indicates that it has stabilised and the backwater zones will not recover quickly. Vegetation dieback is extensive in the lower reaches of the Upper Fly, particularly in the vicinity of D Albertis Junction. Vegetation recovery is, however, significant in the upper reaches of the system and accounts for 55% of the total impacted area. It is important to note that vegetation recovery in this zone will readily convert to a dieback or stress state should extended wet climatic conditions become evident. There are signs that zones of fringe stress and dieback are recovering. In many cases, however, the forested vegetation along the edges of the lake system has totally died out and the physical evidence of dieback no longer exists. In these cases a conversion process has been completed with the forested rim being converted into an aquatic zone with a swamp / grassland border. Page 3 of 26

4 While the total area of dieback in the lower reach of the catchment remains largely unchanged there has been significant conversion of stressed vegetation and recovered vegetation into the dieback classification. This has primarily been in the reach from Bosset to Aiambak. Forested pockets on the floodplain adjacent to the main river stem have been severely impacted although the forest along the tributaries and tributary backwaters remains largely unstressed. With the high estimate of recovering vegetation as a percentage of total area it is critical that these estimates be classified into type of recovery and the bio-diversity of the recovery be assessed. Page 4 of 26

5 CONTENTS EXECUTIVE SUMMARY INTRODUCTION DATA SOURCE DATA COVERAGE DATA INTERPRETATION RESULTS COMPARISON TO MAXIMUM PREDICTED EXTENT COMMENTS REFERENCES 26 Page 5 of 26

6 1.0 INTRODUCTION This report details the results of a vegetation dieback mapping program in the Ok Tedi and Middle Fly floodplains of the Western Province of Papua New Guinea. The investigation was undertaken in order to determine the magnitude and spatial extent of vegetation impact for the period to 30 th June Imagery was acquired on the 12 th February 2004 from the Landsat 5 TM satellite. Vegetation impacts have been mapped with the spatial extent and magnitude of impact defined for significant reaches within the system. The origin of impact has not been assessed. 2.0 DATA SOURCE The data used for the analysis was Landsat 5 TM satellite imagery, acquired on the 12 th February The Landsat data set offers full cloud free coverage of the catchment, excluding a zone in the immediate vicinity of Everill Junction and the Strickland River. This image has been used as the primary source of defining vegetation impact. At the time of image acquisition the Ok Tedi and Fly River floodplains were inundated. There were no other suitable satellite images covering the area of interest in the period to 30 th June The cut-off date for this monitoring epoch has been defined as 30 th June The spectral specifications for the Landsat 5 TM images are defined in Table 1. Band Wavelength (μm) Pixel Size (m) visible blue visible green visible red near infrared middle infrared Thermal infrared middle infrared 30 Table 1 Landsat 5 TM Spectral Bands 3.0 DATA COVERAGE Two consecutive Landsat scenes were acquired to cover the Lower Ok Tedi and Middle Fly from Komopkin through to Ogwa. The Landsat images were path 100; rows 64 and 65 and were acquired on the 12 th February Figure 1 shows a mosaic of the 2004 Landsat 5 TM scenes, with cloud free coverage of the Ok Tedi and Fly River floodplains in the zone of vegetation impact. At the time of image acquisition the floodplain was inundated. As a comparison the 2003 Landsat 7 TM scenes are shown in Figure 2, with the floodplain being predominantly dry at the time of acquisition. The 2004 data covers the full floodplain of the Lower Ok Tedi and Middle Fly River, excluding a minor zone in the immediate vicinity of Everill Junction. Additional imagery was not available to fill this void however this deficiency has not impacted on the 2004 definitions due to the limited area within the Fly River floodplain obscured by cloud. The accuracy of dieback and recovery mapping is in the order of 10%. This is due to the resolution of the imagery utilised, which limits detailed interpretation, but is also due to the complex vegetation regime in the catchment, particularly in the lower Middle Fly. The accuracy of assessment of recovery in the Fringe Dieback and Fringe Stress zones is poor due to the resolution of imagery. Many of these zones are less than 50 metres in width and are not represented adequately on the 30 metre Landsat imagery. Classification in these zones should be updated and verified with appropriate ground truthing. Page 6 of 26

7 4.0 DATA INTERPRETATION Lower Ok Tedi and Middle Fly Vegetation Dieback Monitoring Processing of the data was undertaken using the ENVI software package. The data was imported, corrected for atmospheric distortion, georeferenced and classified using multiple band combinations and data transformations. Data transformations utilised included: The Normalised Difference Vegetation Index (NDVI) this utilises the NIR and Red (R) region of the spectrum. LAI (Leaf Area Index) is usually positively related to an increase in the difference between NIR and R radiation. NDVI was calculated as (TM4-TM3) / (TM4+TM3). The simple vegetation ratio (NIR/R) - which is similar transformation to the NDVI without the normalisation. Brightness, Greenness and Wetness - derived as an extension to the MSS Tasselled Cap transformation. Brightness is the weighted sum of all 6 TM channels and is a response to changes in total reflectance, driven primarily by soil reflectance changes. Greenness contrasts the sum of the visible and near infrared bands and has been shown to be moderately to well correlated to percentage canopy coverage, LAI and fresh biomass. Wetness contrasts the sum of the visible and near infrared channels again the sum of the longer wavelength bands, and is named due to the sensitivity of the longer infrared channels to soil moisture. TM Band 7 - has been shown to be negatively related to leaf water content and as a result on sites of higher basal area there should be a corresponding decrease in the TM band 7 response. The ratio of this band with TM bands 5, 4 and 3 accentuates the differences between the change in leaf water content in band 7 and the relationship between the NIR (of TM bands 5 and 4) and Red (TM3) and vegetation density and biomass. The results of the image classifications were compared to the 2003 and earlier epochs of measurement. The classifications were manually vectorised with subsequent derivation of impact areas for each category within the standard reaches. The extraction of zones of vegetation dieback and recovery for the period has been undertaken without the aid of conventional or photographic ground truthing. As conventional ground truthing for the period was not undertaken spectral responses from earlier epochs of monitoring were utilised to classify the imagery of this monitoring epoch. While a system has been configured to enable the acquisition of high resolution digital airborne images for ground truthing it is still in a commissioning phase and was not been utilised to assist with this monitoring epoch. Ground truthing of the recovery zones through out the whole catchment is required in order to assess the type of recovery (recovery, conversion or re-establishment). Recovery characteristics cannot be accurately determined without additional input. In addition the ground truthing imagery is becoming increasingly important for the assessment of vegetation impact in the lower reaches of the catchment where the floodplain is undergoing complex and dynamic change. Zones of impact in the lower reaches of the catchment are often small and difficult to assess. Difficulties in image classification are do to similar spectral responses from sparse floodplain forest with underlying swamp grass cover and impacted vegetation with colonising grasses. High resolution digital imagery would resolve much of the ambiguity in this zone. Vegetation has been classified into a broad range of categories as per the classifications defined in the Marshall and Rau (1999). The categories of Fringe Dieback and Fringe Stress from Marshall (2001) have also been included. In order to determine the rate of increase of recovery, the 2004 classification has highlighted the new recovery compared to Page 7 of 26

8 Figure 1 Landsat 5 TM Coverage (12 th February 2004) Page 8 of 26

9 Figure 2 Landsat 7 TM Coverage (28 th October 2002) Page 9 of 26

10 5.0 RESULTS Vegetation impacts in the Lower Ok Tedi and Middle Fly have been mapped with spatial extents and extent magnitude defined. Figure 3 shows the coverage of mapping and the reaches for reporting. Figure 3 Lower Ok Tedi and Middle Fly - Vegetation Dieback Reach Definitions Page 10 of 26

11 Vegetation recovery has continued in the lower reaches of the Ok Tedi with a significant percentage of the Ok Tedi now in a recovery phase. Figures 4, 5, 6 and 7 highlight the impact to the Ok Tedi system over the past four epochs. Blue zones in the 2002, 2003 and 2004 figures indicate new areas of recovery for the subject epoch. Figure 4 Lower Ok Tedi 2000 Figure 5 Lower Ok Tedi 2002 Figure 6 Lower Ok Tedi 2003 Figure 7 Lower Ok Tedi 2004 Page 11 of 26

12 Figures 8 to 15 show the vegetation impact extent for the various reaches defined in Figure 3. Figure 8 Vegetation Impact Ieran to Iogi Figure 9 Vegetation Impact D Albertis Junction to Ieran Page 12 of 26

13 Figure 10 Vegetation Impact D Albertis Junction to Kiunga Figure 11 Vegetation Impact Wygerin to D Albertis Junction Page 13 of 26

14 Figure 12 Vegetation Impact Mabaduam to Wygerin Figure 13 Vegetation Impact Manda to Mabaduam Page 14 of 26

15 Lower Ok Tedi and Middle Fly Vegetation Dieback Monitoring Figure 14 Vegetation Impact Everill Junction to Manda Page 15 of 26

16 Figure 15 Vegetation Impact Ogwa to Everill Junction The extent of vegetation impact for each zone is tabulated in Table 3. REACH RECOVERY STRESS FRINGE STRESS DIEBACK FRINGE DIEBACK TOTAL Ieran Iogi D Albertis Ieran D Albertis Kiunga Wygerin D Albertis Mabaduam Wygerin Manda Mabaduam Everill Manda Ogwa Everill TOTAL Table 3 Vegetation Dieback Extent - Lower Ok Tedi and Middle Fly ( ) Note: km 2 recovery and 0.2 km 2 dieback have been included as dredge impact. 2. Table 3 has been compiled from the vegetation state changes defined in Table 5. In order to understand the processes of conversion between the various vegetation impact states, the magnitude of each change has been measured and documented. Table 3 has been derived using the data of Tables 4 and 5 along with the 2003 estimates of dieback, as defined in Marshall (2003)b. Page 16 of 26

17 D Dieback FD Fringe Dieback ND New Dieback NFD New Fringe Dieback S Stress FS Fringe Stress NS New Stress NFS New Fringe Stress R Recovery DR Dredge Impact NDR New Dredge Impact Table 4 Categories of Vegetation Impact Ieran Iogi D Albertis Iogi D Albertis Kiunga Wygerin D Albertis Mabaduam Wygerin Manda Mabaduam Everill Manda D DR 0.24 D R ND S D FD R FS FD R D R S NFD 0.19 FS R S R NFS NS R DR 0.42 NDR 0.67 Table 5 Vegetation State Changes Ogwa Everill Figure 16 shows total dieback and total recovery for the catchment since monitoring commenced in There has been an approximate linear increase in both dieback and recovery extent between the 2003 to 2004 epochs. The magnitude of increase is, however, significantly different with a 2% increase in total impacted area compared to a 35% increase in the area of vegetation recovery. Page 17 of 26

18 Vegetation Dieback and Recovery 2000 Area (Sq Km) Total Dieback Recovery Year Figure 16 Vegetation Dieback and Recovery ( ) Although it is clear that that rate of dieback increase has reduced significantly since 2001 and that the rate of recovery has increased in the same period it is possible that these trends will not continue into the future. The large increase in dieback extent between 1998 and 2001 was not only due to vegetation impact moving into the large Middle Fly catchment but also due to the extended La Nina conditions which were evident during the same period. During the period, water levels in the catchment have been slightly below the long term average. Of more significance is the periodic inundation then draining of the floodplain during the period. This is evident in Figure 17, which gives the water surface elevation at Kiunga between March 2003 and February Estimated bank full elevations are also depicted. The figure shows that there were at least seven events, spaced evenly through the period, where draining of the floodplain would have occurred. While this data relates to the upper reaches of the catchment it would also be indicative of events throughout the catchment as a whole. Vegetation dieback is induced through a process of anoxia, caused by inundation for extended periods of time. Periodic floodplain draining, as is evident in Figure 17, would not only limit vegetation dieback but would allow for vegetation recovery from previously stressed vegetation. Climatic conditions similar to those of the 1999 to 2000 La Nina could readily increase the rate of vegetation dieback, decrease the rate of dieback recovery and possibly convert recovered areas back into dieback or stress conditions. Page 18 of 26

19 Kiunga Water Level March February Elevation (AHD) Daily Water Level Bank Full - Maximum Bank Full - Minimum 24/02/2004 4/02/ /01/ /12/2003 6/12/ /11/ /10/2003 7/10/ /09/ /08/2003 8/08/ /07/ /06/2003 9/06/ /05/ /04/ /04/ /03/2003 1/03/2003 Date Figure 17 Kiunga Water Levels March 2003 to February COMPARISON TO MAXIMUM PREDICTED EXTENT Figures 18 and 19 show comparison of the measured dieback extent with the predictions of Marshall (2003)a. In the Lower Ok Tedi it evident that the maximum extent of dieback is being approached and that the future likely extent is a conservative estimate. In the Middle Fly the measured dieback extent is following the predicted trend. While it is probable that the lower than predicted result is due to dryer climatic conditions than in earlier epochs, particularly 1999 to 2000, the maximum extent predictions allow for future La Nina events. Based on the results of the 2004 monitoring epoch the maximum dieback extent of 2,395 square kilometres is believed to be a conservative estimate. Page 19 of 26

20 Lower Ok Tedi Dieback 250 Extent (Sq Km) Actual Dieback Future (Likely-Dredge) Extent (248 sq km) Projected Dieback Year Figure 18 Lower Ok Tedi Actual Dieback v Estimated Maximum Extent Middle Fly Dieback 2000 Extent (Sq Km) Actual Dieback Future (Likely-Dredge) Extent (2147 sq km) Projected Dieback Year Figure 19 Middle Fly Actual Dieback v Estimated Maximum Extent In terms of spatial distribution of dieback with respect to maximum predicted extent it is clear that the upper reaches of the catchment (Iogi to Wygerin) are approaching maximum extent. There is, however, considerable scope for increase in extent in the lower reaches with Figure 20 clearly showing the potential area for impact downstream of Wygerin. Page 20 of 26

21 Figure 20 Comparison of Estimated Maximum Dieback Extent and 2004 Dieback Extent 7.0 COMMENTS Vegetation dieback extent has increased by 2% across the catchment compared to the previous monitoring epoch and is consistent with the 4% increase in This is against a predicted increase of 7 to 8%. The increase in recovery extent is approximately 35% and the magnitude of change in recovery area is consistent with 2003 estimates. Approximately 30% of the impacted vegetation is under recovery. a) The majority of the Ok Tedi floodplain is under some form of recovery although the backwater zones are still subject to severe dieback. Approximately 70% of the Ok Tedi floodplain is in a recovery phase. The minimal increase in recovery extent in this zone indicates that it has stabilised and the backwater zones will not recover quickly. This is due to the topography in the backwater zones being lower than that adjacent to the main river stem and hence more conducive to the retention of water without adequate transport to the main river stem. Figure 21 shows an example of a blocked backwater zone on the Ok Tedi floodplain. Figure 22 shows the vegetation recovery on the Ok Tedi floodplain in the vicinity of Konkonda. Page 21 of 26

22 Figure 21 Blocked Backwater Lower Ok Tedi Figure 22 Vegetation Recovery on the Ok Tedi at Konkonda Page 22 of 26

23 The total area of vegetation impact within the Lower Ok Tedi remains essentially unchanged compared to earlier epochs. The floodplain in this region is topographically constrained and the overall area of impact is therefore constrained to the current limits. b) Vegetation dieback is extensive in the lower reaches of the Upper Fly, particularly in the vicinity of D Albertis Junction. Vegetation recovery is, however, significant in the upper reaches of the system and accounts for 55% of the total impacted area. Figure 23 shows the location and direction of vegetation recovery along with the zones of severe impact near D Albertis Junction. It is possible that the onset of extensive recovery in this zone is due to lower than average water levels over the monitoring period however it is also possible that bed reduction in the vicinity of D Albertis Junction, due to dredging operations at Bige, is also having a beneficial impact. Bed level reductions would directly reduce the backwater impact induced up stream along the Fly River from D Albertis Junction and significantly reduce the incidence of overbank flooding. It is important to note that vegetation recovery in this zone will readily convert to a dieback or stress state should extended wet climatic conditions become evident. Figure 23 Upper Fly Vegetation Recovery and Dieback at D Albertis Junction c) There are signs that zones of fringe stress and dieback are recovering. In many cases, however, the forested vegetation along the edges of the lake system has totally died out and the physical evidence of dieback no longer exists. In these cases a conversion process has been completed with the forested rim being converted into an aquatic zone with a swamp / grassland border. Although such a process is strictly recovery these zones have been retained in the fringe dieback class. Figure 24 shows a typical inland lake system with zones of fringe stress and dieback as well as locations of vegetation conversion. It is planned to separate recovery types during the 2005 monitoring epoch, and with the aid of a structured ground truthing program, accurate classification of these zones will be undertaken. Page 23 of 26

24 Figure 24 Fringe Stress and Dieback along with zones of vegetation conversion. The conversion process being observed at the fringe dieback sites has little impact on both the ecological and human environments. The impacted land is always along the rim of an inland lake system and would suffer no ecological damage with the change. A new ecological class is not being introduced; rather a small proportion of land has been transferred between two adjoining ecological communities. There would be no discernable impact to the biodiversity of the remaining forest zone. In terms of human impact, the conversion process would again have no discernable impact with the loss of hunting and gathering land being minimal. During the period of active fringe dieback, canoe access to many of the impacted sites would have been difficult if not impossible; however this will now have reverted to pre-impact status. d) While the total area of dieback in the lower reach of the catchment remains largely unchanged there has been significant conversion of stressed vegetation and recovered vegetation into the dieback classification. This has primarily been in the reach from Bosset to Aiambak. Forested pockets on the floodplain adjacent to the main river stem have been severely impacted although the forest along the tributaries and tributary backwaters is largely unstressed. Vegetation recovery extent throughout the entire lower catchment has increased but is predominantly upstream of Manda and in the immediate vicinity of Everill Junction. Figure 25 shows the change in vegetation status between 2003 and 2004 for the Bosset to Aiambak reach while Figures 26 and 27 show the Landsat TM imagery from 2003 and 2004 respectively. Dieback in the sparse floodplain forest zones can be readily seen. Page 24 of 26

25 Figure 25 Dieback Classifications (Bosset Aiambak) 2003 to 2004 Comparison Figure 26 Landsat 2003 Figure 27 Landsat 2004 e) With the high estimate of recovering vegetation as a percentage of total area it is critical that these estimates be classified into type of recovery and the bio-diversity of the recovery be assessed. Although this assessment is not crucial to the spatial monitoring of recovery it is believed that incorporation of this assessment into the classification process will enable more robust estimates of recovery, both extent and type. In many of the fringe dieback and fringe stress zones between Mabaduam and Ogwa it has not been possible to identify recovery due to the resolution of the satellite imagery. For this monitoring epoch uncertain classifications have remained unchanged at the classification. A directed and extensive ground truthing program will enable update and accurate classification of this data set. There has been no effective ground truthing of the remotely sensed classifications in this report, although spectral signatures from earlier epochs have been used as a guide. With the complex changes in vegetation regimes within the catchment, particularly in the grassed floodplains downstream of Mabaduam, it is critical that an extensive ground truthing program be implemented in conjunction with the next monitoring epoch. Page 25 of 26

26 8.0 REFERENCES Coops, N.C. (1995) Marshall A.R. and Rau M.T. (1999) Carroll D.G, Marshall A.R. & Moi A.S (1999) Kauth, R.J., Thomas, G.S. (1976) Marshall A.R. (2000) Spectral and Textural Remotely Sensed Responses to Predict Forest Standing Volume at Batemans Bay. Report on the Batemans Bay Vegetation Dataset CD-ROM. CSIRO Australia. Canberra Lower Ok Tedi and Middle Fly Estimate of Current Vegetation Dieback and Classification of Floodplain Vegetation, Ok Tedi Mining Limited, Internal Report The Results of Dieback Modelling on the Ok Tedi and Fly River Floodplains, Ok Tedi Mining Limited, Internal Report The Tasseled Cap - A Graphic Description of the Spectral-Temporal Development of Agricultural Crops as seen by Landsat. Proc. Symposium on Machine Processing of Remotely Sensed Data. Purdue University. Indiana. 4B-41-4B-50. Lower Ok Tedi and Middle Fly AirSAR Ground Truthing Program September 2000, Ok Tedi Mining Limited, Internal Report Marshall A.R (2001) Lower Ok Tedi and Middle Fly, Vegetation Impact September 2000, Ok Tedi Mining Limited, Internal Report Marshall A.R (2002) Marshall A.R. (2003) Marshall A.R. (2004) Pickup G. & Cui Y. (2003) Lower Ok Tedi and Middle Fly, Vegetation Dieback Monitoring , Ok Tedi Mining Limited, Internal Report Future Vegetation Dieback Extent in the Lower Ok Tedi and Middle Fly Floodplains, Ok Tedi Mining Ltd, Internal Report Lower Ok Tedi and Middle Fly, Vegetation Dieback Monitoring , Ok Tedi Mining Limited, Internal Report Effects of Mine Life Extensions and Rates of Dredging on the Ok Tedi and Fly River A Sediment Transport Model Study using OKGRAV6 and HEC-6, Ok Tedi Mining Ltd, Internal Report Page 26 of 26