A Maximum Entropy (MaxEnt) algorithm was calibrated with ground data to generate living aboveground biomass (AGB), its associated uncertainty, and forest probability maps for Mexico. The input predictor layers were extracted from Optical and Synthetic Aperture Radar (SAR) imagery, as well as from a digital elevation model. The combination of the three spatial datasets showed superior accuracy and lower relative error (0.31 and 58%) than the use of single dataset (0.12 - 0.19, and 62% - 74%) or two datasets (0.25 - 0.28, and 58% - 59%). The AGB map showed a root mean square error (RMSE) of 17.3 t C ha-1 and R2 = 0.31 when validated with inventory plots. The total carbon stored in forests was estimated to be 1.69 Gt C ± 1%, which agrees with the total national estimations. This new map proved to have similar accuracy as previous AGB maps of Mexico, but to be more representative of the shape of the probability distribution function of AGB in the national forest inventory data. Different forest area masks with similar forest definitions but originating from different sensors are widely-used to constrain AGB retrievals. The use of different forest masks yielded differences of about 24.1 million ha in forest cover extent and 0.36 Gt C in total carbon stocks for Mexico. A forest cover mask derived from the combination of spatial datasets showed higher accuracy (κ=0.83) than alternative masks derived from SAR (0.78) or optical datasets (κ=0.66). This work found an increasing AGB trend with elevation in Mexico, and that the allometric relationship between AGB and canopy height (H) at plot level significantly varies within biomes and across the topographic gradient (p-value < 0.001). As a result, the amount of AGB per unit of H at higher altitudes is higher than at lower altitudes. This has implications in the use of generalised models across large areas such as those seen in the tropical carbon maps (TCMs) (Saatchi et al., 2011b, Baccini et al., 2012). TCMs show large discrepancies when compared to in-situ observations and regionally calibrated maps. The use of a single allometry (vs. regional allometry), and the calibration of the algorithm without taking into account regional variations are the main sources of the discrepancies. Errors up to 74% are found in this thesis when using the continental allometry from Saatchi et al. (2011b) over Mexico. The results show that the variability on forest ecosystems play a key role when mapping AGB at larger scales. Thus, approaches that take into account these regional variations, are the way forward to improve these products.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:694176 |
| Date | January 2016 |
| Creators | Rodríguez Veiga, Pedro |
| Contributors | Balzter, Heiko ; Tansey, Kevin |
| Publisher | University of Leicester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/2381/38039 |
Page generated in 0.0021 seconds