運用相對輻射校正增強小衛星Planet NICFI偵測熱帶乾燥林植被季節性能力

Abstract

受限於技術和成本，具有良好空間解析度的陸地觀測衛星傳感器往往具有較低的時間解析度，因此在實際應用中需要不斷在空間和時間解析度間權衡，這種權衡在熱帶地區的研究中尤其具有挑戰性。濕潤季節頻繁的雲層增加了清晰影像的獲取難度。商業衛星儘管具有高性能，但其成本過高難以進行實際應用。如今以小衛星為代表的廉價商業衛星得到了發展，其中以Planet公司的產品作為代表。小衛星通過削減單個衛星的成本實現更密集的發射，並以星座組網的方式改善觀測模式，在空間解析度及獲取清晰影像能力上存在優勢。但小衛星資料質量尚未得到有效檢驗。現在來自挪威的環保組織NICFI與Planet開展了合作項目，提供了免費的熱帶地區Planet影像（Planet NICFI），藉此機會我們檢視了Planet NICFI的資料質量，以經過嚴格輻射校正的傳統衛星Landsat 8/9 OLI作為參照，發現Planet NICFI反映光譜變化能力有限，不能有效檢測乾燥季節植被生長差異。我們通過相對輻射校正（relative radiometric calibration, RRC ）的方法嘗試改善Planet NICFI資料的輻射質量，以Landsat 8/9 OLI作為參考，增強不同波段間兩源影像的反射率一致性。本研究選取三種RRC方法進行對比，分別為基於非線性原理的直方圖匹配法（histogram matching, HM）、基於線性原理的MAD法（multivariate alternation detection）與IR-MAD法（iteration re-weighted MAD）。處理結果顯示，三種方法在操作穩定性上存在差異：HM法受限於參考影像質量，在濕潤季節效果不好。MAD法在濕潤季節檢測噪聲能力不足，而IR-MAD法在濕潤季節也可正常開展，取得了較好效果。經過RRC校正後Planet NICFI在保持高空間解析度與無雲的優勢的同時提高了光譜季節性差異。該結果為在熱帶乾燥林進行低成本與高空間解析度的大範圍植被生長監測提供了參考。

Earth observation satellite with fine spatial resolution often has a low temporal resolution mainly due to technical limitations. The trade-off between spatial and temporal resolutions is particularly crucial in the tropics, for the more frequently cloud cover impedes optical satellite remote sensing instruments from obtaining cloud-free imagery. Commercial satellites may overcome the above issue, but the data may be costly. An alternative and cost-effective observation strategy based upon SmallSat has recently become an operational reality. SmallSats, represented by Planet’s products, achieve more intensive launches by reducing the cost of a single satellite, and improve the revisiting time through satellites constellations, which lead to its advantages in high spatial resolution and stronger ability to obtain clear images. Planet, through Norway’s International Climate and Forest Initiative (NICFI), now open free access to analysis-ready basemaps over the tropics, which shows a high potential in the field of vegetation metabolism monitoring. However, the radiometric data quality of SmallSat is not equivalent to a rigorously calibrated satellite such as Landsat 8. We tested the data quality of Planet NICFI by using Landsat 8/9 OLI as the reference images, and found that Planet NICFI has limited ability to reflect vegetation spectral changes in tropical dry forests (TDF), especially during the dry seasons. Therefore, we attempted to improve the radiometric quality of Planet NICFI by using relative radiometric calibration (RRC).We using Landsat 8/9 OLI as a reference data to enhance the consistency in surface reflectance of the two images. Three RRC methods were selected for comparison, namely histogram matching (HM) based on the nonlinear assumption, multivariate alternation detection (MAD) and iteration re-weighted MAD (IR-MAD) based on the linear assumption. The result showed that HM cannot be implemented when the images are contaminated by clouds , while MAD and IR-MAD are more generally available, for they can detect cloud-free pixels automatically and enhance the background no-change area. IR-MAD is more effective in removing noise and therefore can also operate normally in wet season. The calibrated data showed improvement in seasonal variation of spectral and has higher spatial details with clear boundary information compared to the reference Landsat 8/9 OLI imagery. Our findings provide a foundation for cost-effective and accurate high spatial resolution vegetation phenology monitoring in TDF.