| 研究課題/領域番号 |
22K21332
|
|---|
| 研究種目 |
研究活動スタート支援
|
| 配分区分 | 基金 |
|---|
| 審査区分 |
1101:環境解析評価、環境保全対策およびその関連分野
|
|---|
| 研究機関 | 沖縄科学技術大学院大学 |
|---|
研究代表者 |
ROSS Samuel 沖縄科学技術大学院大学, 統合群集生態学ユニット, ポストドクトラルスカラー (60961795)
|
|---|
| 研究期間 (年度) |
2022-08-31 – 2025-03-31
|
| 研究課題ステータス |
交付 (2023年度)
|
| 配分額 *注記 |
2,860千円 (直接経費: 2,200千円、間接経費: 660千円)
2023年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
2022年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
|
|---|
| キーワード | ecology / stability / response diversity / resilience / duckweed |
|---|
| 研究開始時の研究の概要 |
I will conduct 3 experiments on floating plants. 1: a simple test asking does response diversity produce less variable biomass. 2: an extension to different dimensions of stability (resistance, resilience etc.). 3: an extension to spatially-linked communities where dispersal/competition are possible
|
| 研究実績の概要 |
We conducted an experiment in Okinawa, using 75 100L buckets of water with floating aquatic plants. We used 4 species of plants in different combinations, and exposed them to Nitrate (to simulate agricultural runoff). Each week we measured water chemistry and took photographs of the plant communities, which reproduced clonally. We are developing a machine learning algorithm to automatically classify our plant species from photos. When finished, we will have data on plant growth and community composition changes, which we can use to understand how different species respond to Nitrogen, measure response diversity, and finally to relate response diversity to resilience of the communities. I also published a literature review about the topic of response diversity, part-supported by this grant
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
Turning photographs of the plants into useable ecological data has proven more difficult than anticipated. Instead of using an existing machine learning classifier to automatically identify our different plant species from photographs, we must develop our own tool to do this. We are currently working with a deep learning algorithm which is proving promising, but still requires several validation steps before we can use the data.
|
| 今後の研究の推進方策 |
The next step is to continue validation of our deep learning algorithm to identify plants and measure changes in abundance and composition in each mesocosm (100L bucket). To do this, I am currently manually labelling thousands of image segments with species identity to aid the automated identification pipeline. I anticipate several stages of retraining based on additional manual labelling.
Then, when I have data on species abundances, I will measure growth rates of different species under different nitrogen conditions, and when species are in mixture, I can measure the diversity of their growth responses. I will finally use the diversity of these responses to predict resilience (measured here as the inverse of the temporal variability of total biomass output from the combined community).
|
