1
Estimation and upscaling of above-ground biomass in a tropical forest landscape using field, airborne LiDAR and multiresolution satellite data | |
Author | Jha, Nidhi |
Call Number | AIT Diss. no.RS-20-01 |
Subject(s) | Optical radar Rain forest ecology Landscape ecology--Remote sensing Remote-sensing images--Interactive multimedia |
Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Remote Sensing and Geographic Information Systems |
Publisher | Asian Institute of Technology |
Abstract | Tropical forests store half of all the carbon stored in the world's forests and thus play critical role in the global carbon cycle. Accurate estimations offorests carbon stock and its recovery are thus extremely important, especially to meet the guidelines of international programs aimed at reducing greenhouse gas emissions in tropical areas, such as the United Nations Reducing Emissions from Deforestation and Forest Degradation (REDD+). The study is concentrated to a forest landscape of area of ca. 6,400 ha, which is part ofKhao Yai National Park in central Thailand (latitude: 14° 25' 20.4" N, longitude: 101° 22' 36). Khao Yai is the first national park in Thailand, established in 1962 that is home to numerous endangered plant and animal species. Combining recent field and concomitant airborne LiDAR system (ALS) derived data, a high-resolution AGB map was built over 64 km2 of the forest landscape. Random forest (RF) algorithm and Landsat time-series (L TS) data was used to classify landscape patches as non-forested versus forested on an almost annual basis from 1972 to 2017. ALS-AGB model predicted AGB with an error of 14% at 0.5-ha resolution (RMSE = 45 Mg ha-1) using the mean top-of-canopy height as a single predictor. The mean AGB over the landscape was of291 Mg ha-1 showing a high level of carbon storage despite past disturbance history. The result shows that the AGB recovery of the landscape varies non-linearly in the first 42 years of the succession, with an increasing rate of accumulation through time. The predicted mean AGB recovery rate is 6.9 Mg ha lyr', with a mean AGB gain of 143 and 273 Mg ha-1 after 20 and 40 years, respectively. The derived AGB rate (20 year) estimates are similar to the prescribed IPCC 2006 rates for young secondary tropical rainforest (:s 20years) in Asian ecozone which is about 50% larger than the ne,w 2019 refinement to the 2006 IPCC guidelines. The estimated results are within the range of those reported for the well-studied Latin American secondary forests under similar climatic conditions. For estimating the uncertainties in AGB map, AGB errors were propagated from tree level measurements to the LiDAR landscape estimate using the Monte Carlo scheme. With this method, the errors are propagated at different spatial scales, range from 0.1 to 2 ha and with a constant number of calibration plot (n=15). The result shows that at a 0.5-ha spatial resolution, the overall error was of 15% and 23% considering the "site-independent" and "site-dependent" biases respectively. Also with the increase in plot size from 0.1 to 2 ha, the decrease in error is almost 37% to 20% in scenario of "site-dependent biases" while in the scenario of "site-independent" case the decrease in error is almost 33% to 9% showing the phenomenon of averaging out of error more rapidly in later case. The result showed the incongruity with many studies that underestimated the effect field estimate errors in the final AGB map. Using the high-resolution AGB maps obtained through LiDAR data, the assessment of the extent of different satellite products, such as Landsat 8 (L8;30-m resolution), Sentinel 2 (S2; lO-m) and Worldview 3 (WV3; ca 2-m) to predict AGB was done and test if these products are complementary or not. RF algorithm was used along with forward step selection of variables with a cross-validation framework to assess model accuracy independently. The approach provided information on the saturation point of passive optic products, i.e. the AGB value above which the products fail to capture AGB variation and provide a recommendation on the best products to be used. The result show that the predictive power of Sentinel-2 is as good as Landsat 8 and is less accurate than that of Worldview 3. Overall, the predicted AGB saturated at ~ 190 Mg ha-I for Landsat 8 image and Sentinel 2 and at 247 ha-I for Worldview 3 images. Here, using a robust and blind automatized procedure, we show that four to five variables in each sensor were retained from a set of20-28 variables. The selected predictors strongly varied from one sensor to another one. The best predictor of AGB was Green band from Landsat 8, SWIR band from Sentinel 2, and NDRE variable from the Worldview 3. Only the green band and the CVI metrics were commonly retained for two different sensors. When predictors were selected according to their inter-sensor potential the three best predictors were the raw Green, Blue and Red bands. The prediction of AGB for full Khao Yai using the predictors from the selected sensors shows that the AGB in the landscape ranged from 17-489 Mg ha-I with a mean of the landscape of 249 Mg h -I a . Overall, the study illustrates the potential of ALS data not only for scaling up field AGB measurements but also for predicting AGB recovery dynamics when combined with long-term satellite data. The potential use of LIDARderived AGB estimates for upscaling biomass estimates using several passive optical remote sensing data was demonstrated in the study. The result illustrates that tropical forest landscapes that were disturbed in the past are of utmost importance for the regional carbon budget and thus for implementing international programs such as REDD+. |
Year | 2020 |
Type | Dissertation |
School | School of Engineering and Technology (SET) |
Department | Department of Information and Communications Technologies (DICT) |
Academic Program/FoS | Remote Sensing (RS) |
Chairperson(s) | Tripathi, Nitin Kumar |
Examination Committee(s) | Virdis, Salvatore G.P.;Sasaki, Nophea, |
Scholarship Donor(s) | Asian Institute of Technology Fellowship |
Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2020 |