Using multi-temporal high-resolution remote sensing in surface modeling

Version française par ici. The land surface models simulate the water and energy fluxes between soil, land cover and atmosphere. Their scopes of application spread from numerical weather prevision to soil water modeling. 

 However, these models are initially conceived to be applied on wide areas. Thus, they use low resolution cover parameters (>1km) derived from mid-resolution satellite observations (MODIS, VEGETATION). These parameters are mainly the Leaf Area Index (LAI), the vegetation type or the surface albedo. Yet the agricultural landscapes of Western Europe are characterized by a patchwork of plots smaller than one square kilometer. These plots have very different vegetation cycles, i.e. winter and summer crops, which could only be described at high resolution. The crop management practices like crop rotation or irrigation are also generally not taken into account. The products of the Sentinel-2 space mission, with their high spatial and temporal resolution, could bring elements to fill in this missing information. The main drawback of working at high resolution is that the computation time of these complex models is skyrocketing. Therefore, the solution we have chosen is to simulate at plot scale. Instead of a regular computation grid, we use an irregular grid where each plot is a computation cell. These plots are determined from a land cover map. We have tested this approach with the Surfex/ISBA model and the famous Formosat-2 series with its 8m resolution. This series has already been used in various activities at the Cesbio to prepare the arrival of Sentinel-2. 

 This approach is around 900 times faster than a pixel based approach at Formosat’s resolution. As the vegetation on each plot can be considered as homogeneous, a unique set of cover parameter can be attributed with reasonable uncertainty levels. The Leaf Area Index (LAI) is a particularly important parameter. It is defined as half the surface of the green leaves of the plant. In Surfex, this parameter is used to calculate the stomatal resistance. In other words, it determines the plant capacity to restrain its water consumption depending on the climatic conditions and the soil water content. Therefore, it is an important factor in the evapotranspiration (ET) estimation and by extension in both energy and water budgets. 

 We have performed two simulations, both at the plot scale. The first one was using the vegetation types, and associated cover parameters, from the ECOCLIMAP database. This database describes the land surface with a 1km resolution. In the second simulation, the vegetation type is forced by the Formosat-2 land cover map. The LAI, also derived from the Formosat-2 series (Claverie, 2012), is forced too. The comparison to in-situ measurements on the Auradé ans Lamsquère sites shows the improvement of the phenology (LAI) description when using Formosat-2 instead of ECOCLIMAP, like below in 2006. 

 When analyzing the ET flux, the comparison show very few impact on winter crops like wheat (Auradé) whereas the seasonality on summer crops like maize (Lamasquère) are profoundly modified. On this kind of crop, the ET peak is delayed by one month. It thus fits the measurements better. 

 The spatial analysis shows that this delay is common to all the summer crop plots present in the Formosat-2 images. Indeed, the LAI derived from Formosat-2 is lower than the ECOCLIMAP LAI during spring. Hence, the plant has a lower transpiration and lowers the soil water content slower. Then this water can be used later during summer, increasing the ET and creating the delay in the ET peak. 

 However, the years when maize is grown in Lamasquère, the ET remains too small compared to the measurements. This problem is not visible on the years when sunflower (which is also a summer crop) is grown in Auradé. A complementary analysis has shown that this problem is explained by the irrigation, which is not simulated by ISBA. So the main objective is now to implement an automated irrigation module in the model to represent a realistic irrigation amount on this kind of crop. Remote sensing data (Sentinel-2, Landsat, SPOT) will thus be used to determine the irrigated plots using some work in progress at Cesbio (Florian Helen’s article). If somebody wants more details about this study, it is publihed right here (Etchanchu et al., 2017).