Greiner, L.; Brandl, R. & Farwig, N. (2016): <b>Texture images as tool for predicting bird feeding guilds in a tropical montane rainforest</b> Philipps-Universität Marburg, Department of Conservation Ecology, <i>master thesis</i>
Texture images as tool for predicting bird feeding guilds in a tropical montane rainforest
License and Usage Rights:
PAK 823-825 data user agreement. (www.tropicalmountainforest.org/dataagreementp3.do)
Facing the ongoing loss of natural ecosystems, worldwide monitoring of biodiversity across different spatial scales is essential for conservation planning. Remote sensing (RS) has proven to be a cost-efficient tool to access environmental characteristics such as vegetation structure and associated distributions of animal species on a broad scale. Special emphasis is put on birds as indicators for biodiversity owing to their strong species–habitat relationship. So far, bird diversity was modeled ignoring that species–habitat relationships differ among feeding guilds. This is surprising, since habitat preferences strongly depend on diet specialization. Therefore, I investigated RS texture image based vegetation metrics to test whether the predictability of specialized avian feeding guilds including insectivores, frugivores and nectarivores is higher than of the less specialized omnivore guild and overall bird diversity. I used point count data of bird communities among 30 study sites in a complex tropical mountain forest ecosystem in south-eastern Ecuador to estimate (i) Shannon index and (ii) community composition as measures of ?-diversity and combined ?- and ?-diversity, respectively. In order to relate both diversity measures to RS metrics, I compared two high dimensional predictor sets – satellite images and airborne orthophotos – with structural indices derived from a discrete return airborne Lidar sensor. Partial least squares regression was used to unveil the predictive power of all fitted feeding guild models. For the comparability of all models, a sample size correction on species number per guild was applied. Shannon index predictability ranged between 37 % and 65 %; and best predictions were achieved for insectivores using metrics from satellite or Lidar images and nectarivores species using metrics from orthophotos. Community composition was generally better predicted than Shannon index with explained variations from 65 % to 85 %. Frugivore and nectarivore community compositions were best predicted using metrics from orthophotos, whereas the two other sensors best predicted omnivores. For both diversity measures, performance of satellite derived metrics revealed slightly better model results compared to other sensors emphasizing its applicability for the regarded study area. In conclusion, specialized feeding guilds were not consistently better predicted than omnivore or overall bird diversity; rather the study showed that model performances depended on the regarded diversity measure and RS image type. However, insectivores might be the best surrogate for overall diversity with high predictability in all compared models. In addition, the high explanatory power for community composition suggests that the measure should considered in avian diversity modeling for conservation planning.