Since October 2015, the TOMCAR-Sat project aims to use high resolution satellite products to monitor Dissolved Organic Carbon (DOC) fluxes on the Arctic river Yenisei. Specifically, its objectives consist to calibrate a DOC retrieval model from time-series of satellite images and to characterize dynamics of the factors that may affect DOC concentrations, i.e. the snow cover and vegetation.
Spot 5 Take 5 (03/06/2015) - Yenisei river (Igarka, Russia)
How it works?
The Dissolved Organic Matter (DOM) has particular optical properties in the visible spectrum of the white light. Its colored fraction, called CDOM (Chromophoric Dissolved Organic Matter), absorbs intensely in the blue part of the spectrum and then the absorption decreases with increasing of wavelengths. CDOM is often well correlated to DOC, which makes it an interesting proxy to supplement, in time and space, DOC in-situ measurements. Relationships between CDOM and satellite optical signal have been widely explored to retrieve DOC in various aquatic systems such as oceans, lakes or rivers. Arctic rivers could take advantage of these works since DOC in-situ measurements are constrained by difficult logistical conditions, namely because of the accessibility to the river. Moreover, 80% of DOC fluxes occur during the peak-flow period, i.e. a few weeks between May and June, after the ice-melt period. Until now, little work focused on Arctic river systems, mainly due to existing difficulties to get field samples but also because of inadequate satellite data. I mean...
Temporal evolution of DOC concentrations in the Yenisei river in 2015
Sentinel 2, we're waiting for you!
As I said, 80% of DOC fluxes occur on a few weeks in May and June which need to have a maximum of satellite acquisitions in a very short-time period to observe rapids and abrupt changes. High-temporal resolution satellites are crucial to drive these works but high-spatial resolution data also allow optimizing the model calibration. They allow extracting leaving water reflectances between clouds (and there are quite a few...) and between ice-breaks just after the ice-melt, which is not always possible with coarser resolution data. Moreover, if combined with high-temporal resolution, atmospheric corrections are often better which may be precious in order to make a link with in-situ absorbance measurements. Yes but here, until now, no optical satellite sensors combine both aspects. Satellites such as MERIS or MODIS have a high temporal resolution (1 to 8 days) but their spatial resolution is weak (250m, 500m, 1km). High-spatial resolution satellite such as Landsat or Spot (respectively 30m and 10 in multi-spectral mode) could be suited but their low temporal revisit time (> 16 days) is too limited to provide sufficiently useful data in the peak-flow period. So, how to do?
Sentinel 2A and 2B should fill these gaps since they will observe the totality of land areas every 5 days with a spatial resolution of 10 to 60m in 13 spectral bands from visible to mid-infrared. Nonetheless, we will have to wait longer to have exploitable Sentinel 2 time-series in the region of Igarka. Acquisitions are still not fully operational for various reasons (technical, programming..) which delays the availability of high-temporal resolution data. However, and it is a rather a good piece of news, Igarka was selected by THEIA for the distribution of atmospherically corrected satellite scenes starting in Autumn 2016. About this topic, don't hesitate to read the different posts published in the Sentinel 2 section that allow well understanding the Sentinel 2 programming as well as the different processes carried out by THEIA on S2 tiles.
Meanwhile, we're training with Take 5...
Last year, the orbit of Spot 5 was lowered in order to increase its temporal resolution to simulate Sentinel 2 and training future S2 users. Called “Spot 5Take 5”, this short experiment allowed us to have Spot 5 scenes (Green-Red-Near Infrared-Short Wave Infrared, 10m) all 5 days from 09/04/2015 to 06/09/2015. DOC/CDOM in-situ measurements carried out by the TOMCAR team were also synchronized with ST5 acquisitions (the same day) to calibrate a DOC model retrieval (see below).
(at left) Relationship between predicted CDOM (a440) and measured CDOM (R2 = 0.76); (at right) Relationship between predicted DOC and measured DOC (R2 = 0.79)
By combining these data with Landsat 8 data, we could produce a first spatialization of DOC concentrations along the Yenisei river stream at a high-spatial and a high temporal resolution that you can see below. Despite an important cloud cover, peaks of DOC concentrations are well visible in May and June, i.e. during the peak-flow period, just after the ice-break. It is also possible to observe a steady decrease of DOC from mid-June to September, which is consistent with field observations. But this time, we can see it at large scales!
Mapped DOC on the Yenisei river during a theoretical open-water season
All details of this work were not described here but if you're interested, our talk presented in the Living Planet Symposium of ESA in Prague on May 2016 is available here. Note the diversity of the applications of Take 5, it's fascinating! In short, this work shows that Sentinel 2 mission is promising for DOC monitoring in Arctic rivers and should allow improving it. I will talk later about these works and also how we plan to explore S2 satellite images to monitor snow cover and vegetation in the Yenisei catchment.