The aim of this thesis is to evaluate the capability of radio detection and ranging (radar) data collected by the Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture radar (PALSAR), supported by field measurements obtained through ground survey, to predict and map Above Ground Woody Biomass (AGWB) in the tropical savannas of the developing country of Belize, and to understand how the forest structure may influence the backscatter observed. Firstly, an extensive inventory of the woody vegetation of the tropical savannas of Belize was created by measuring the diameter at breast height (dbh), the total height (ht) and the location of 6547 trees in plots covering a total woodland area of 30.8 hectares, located within four protected areas (the Rio Bravo Conservation and Management Area (11×1ha), Deep River (108×0.1ha) and Manatee Forest Reserve (1ha) and the Bladen Nature Reserve (1ha) and also from plots located in unprotected areas (7×1ha). These measurements of forest structure, when combined with information about forest management practices obtained from local organisations revealed that different forms of protection and management may lead to the development of pine woodlands with different structural characteristics in these savannas. Secondly, a case-study was conducted to establish the sensitivity of the ALOS PALSAR backscatter data to AGWB and determine the effect of sample plot size to their relationship. The findings of this case-study show that the L-band backscatter in these low density pine woodlands is a possible predictor of AGWB and confirm that the appropriate sample plot size for predicting AGWB is one hectare; while the sensitivity degrades significantly with decreasing sample plot size. Taken together, the findings described above were combined to assess the capability of ALOS PALSAR backscatter to predict AGWB in these woodlands. A semi-empirical Water Cloud Model (WCM) describing the interaction between the backscatter and vegetation was re-arranged to enable the prediction of AGWB. Non-linear regression analysis revealed that the ALOS PALSAR backscatter predicted AGWB with an R2=0.92; an external validation conducted with additional ground reference data estimated this AGWB prediction to have an RMSE ~13 t/ha. The form of the regression model linking backscatter to AGWB appears to be particularly influenced by sample plots with higher tree numbers and by plots in which the trees were more homogeneous. The presence of many similar sized individuals within some plots is postulated as one explanation for the elevated saturation level for predictions in this study (> 100 t/ha) compared to other models. The model developed here predicts complete saturation in the backscatter - AGWB relationship to occur primarily as a result of increases in the tree number density and often concurrently in basal area, two parameters which are usually strongly correlated with AGWB in these woodlands. Thirdly, the locally validated relationship between ALOS PALSAR backscatter and AGWB is used to map AGWB for the lowland pine savannas of Belize at a spatial resolution of 100m. The mapping estimates that over 90% of these pine woodlands have an AGWB below 60 t/ha, with the average woody biomass estimated at 23.5 t/ha. When these new predictions are mapped and aggregated over the extents of two protected areas (Rio Bravo and Deep River), the totals obtained agree closely (error ≤20%) with previous estimates of AGWB obtained from ground data and previous research. The combined evidence suggests that woodland protection may produce a small, positive effect upon AGWB, with the mean of the AGWB/ha predictions higher in areas that are protected and managed for biodiversity (29.55 ± 0.84 t/ha) than in other areas that are not protected (23.29 ± 0.19 t/ha). When the fine scale local AGWB mapping produced using ALOS PALSAR is compared cell-by-cell with global biomass products at coarser spatial resolutions (500m and 1000m), the AGWB differences observed range from 115-120%. When the coarser AGWB estimates are aggregated over the extents of Deep River and Rio Bravo, the AGWB totals obtained differ significantly (~280 – 300%) from AGWB estimates from ground data and previous research. Overall, these findings suggest that where sufficient ground data exists to build a reliable local relationship to radar backscatter, more detailed biomass mapping can be produced from ALOS and similar satellite sensor data at resolutions of ~100m. This more accurate and spatially detailed information about the distribution of woody biomass within tropical lowland savannas is more appropriate for monitoring local changes in forest cover and for supporting management decisions for forested areas of around ~10,000ha than estimates based upon previously available, but coarser scale, global biomass products.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:700063 |
| Date | January 2015 |
| Creators | Michelakis, Dimitrios |
| Contributors | Stuart, Neil ; Woodhouse, Iain |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/17879 |
Page generated in 0.0025 seconds