Impact of plot size on precision of lidar-assisted estimation of aboveground biomass in a submontane rain forest in Tanzania

摘要: Airborne laser scanning (ALS) has been used operationally to support forest management inventories in boreal forests for more than a decade (NA¦sset 2014). Because metrics such as canopy height and density derived from the three-dimensional ALS point clouds are highly correlated with parameters like stand height, volume biomass, can greatly improve accuracy of estimates parameters. also proven be economically superior alternative inventory approaches (Eid et al. 2004). Typically, georeferenced field sample plots size 200-400 m2 fit models that relate ground observations metrics. These subsequently predict parameter interest every grid cell (“pixel”) area interest. A map is often an intermediate product this process. Such predictions (maps) either (1) aggregated local planning purposes or (2) statistical estimation resources entire survey area. The two main areas application equally relevant tropical countries. With increased payments storing carbon ecosystem service under Clean Development Mechanism (CDM) REDD+ mechanism (Reducing Emissions Deforestation Forest Degradation Developing Countries) United Nations Framework Convention on Climate Change, there need cost-effective methods produce precise stocks changes over time at various geographical levels. Designing CDM become prioritized issue. In conventional based surveys plots, example national inventories, important design considerations include among other factors stratification area, cluster designs, random versus systematic selection clusters, plot sizes, size, permanent temporary simplicity robustness implementation maintenance time. However, it evident remotely sensed data will play crucial role future tropics. To date, specific challenges associated use have large extent neglected. Thus, many existing sample-based monitoring programs around world designed way prevents most effective estimation. One size. When small relative tree crowns, so-called edge effect may substantial because discrepancies between same location. particular, portions trees whose stems outside extend into vertical extensions boundaries and, similarly, inside beyond (Mascaro 2011, NA¦sset 2013). Discrepancies arise assessments biophysical variables biomass generally stem locations, whereas boundaries. decrease quality relationships measurements (McRoberts Larger tend reduce influence effect. Likewise, GPS position errors severe consequences smaller larger plots. The both examples optimality sizes when but which irrelevant alone. An ALS-assisted was conducted high-biomass submontane rain Tanzania explore impact uncertainty aboveground estimates. study Amani Nature Reserve located East Usambara Mountains eastern Tanzania, part Eastern Arc Mountains. global biodiversity hotspot (Myers 2000) stretch Udzungwa south Taita Hills Kenya north. terrain characterized by steep slopes elevation range 1000 m. reserve 8360 ha. A 173 rectangular initial 0.1 ha 450 900 m sampling established non-governmental conservation development organization, Frontier 1999-2000. were remeasured during several campaigns 2008-2012. Plot corner coordinates recorded differential Global Positioning System (GPS) Navigation Satellite (GLONASS) using 40-channel dual frequency grad receivers. Due dense canopies, we unable obtain 20 For remaining 153 total (AGB) estimated measurement individual prediction summation recently developed allometric (Masota destructive samples Reserve. AGB ranged 43.2 1147.1 Mg/ha. ALS acquired January-February 2012 Leica ALS70 sensor operated average flying altitude 800 above level. pulse repetition 339 kHz, resulting 10 points/m2. Calculation actual GPS+GLONASS recordings revealed mean 0.0914 rather (range 0.0639-0.1239 ha; standard deviation 0.011 ha). discrepancy mainly due difficulties work caused complex topography, well tessellated cells. Canopy distributions created cell. deciles distributions. Regression relating variable algorithms. model explanatory (a decile metric) selected estimation. A pure field-based estimate obtained estimator assuming simple sampling. error (SE) estimated. model-assisted SE regression estimator. 461.8 Mg/ha (SE=17.4 Mg/ha) while 435.9 (SE=11.3 Mg/ha). efficiency estimate, i.e., ratio variances, 2.4 [(17.4/11.3)**2], factor indicating approximately how would needed precision case. The gain very modest compared similar experiences considerably Relative efficiencies 5 seems good rule thumb (cf. McRoberts studies indicated increasing reduced effects geolocation 2011). We second subsequent analysis. 30 circular 0.19 measured 22 these even had 0.3 ha. distance center recorded, 0.02 steps 0.01 Standard obtained, For maximum ha, continuously 2.0, 5.7. At 0.09 2.5, those up 6.3 size. A fundamental conclusion drawn study. extremely assisted ALS. Great improvements realized if adequate. Rapidly about 0.2 further moderate increases least suggest should considered forests. Further investigations analyze types temperate