IMPROVING SHADE MODELLING IN A REGIONAL RIVER TEMPERATURE MODEL USING FINE-SCALE LIDAR DATA

Abstract:

Air temperature is often considered as a proxy of the stream temperature to model the distribution areas of aquatic species water temperature is not available at a regional scale. To simulate the water temperature at a regional scale (10⁵ km²), a physically-based model using the equilibrium temperature concept and including upstream-downstream propagation of the thermal signal was developed and applied to the entire Loire basin (Beaufort et al., submitted). This model, called T-NET (Temperature-NETwork) is based on a hydrographical network topology. Computations are made hourly on 52,000 reaches which average 1.7 km long in the Loire drainage basin.

The model gives a median Root Mean Square Error of 1.8°C at hourly time step on the basis of 128 water temperature stations (2008-2012). In that version of the model, tree shadings is modelled by a constant factor proportional to the vegetation cover on 10 meters sides the river reaches. According to sensitivity analysis, improving the shade representation would enhance T-NET accuracy, especially for the maximum daily temperatures, which are currently not very well modelized.

This study evaluates the most efficient way (accuracy/computing time) to improve the shade model thanks to 1-m resolution LIDAR data available on tributary of the LoireRiver (317 km long and an area of 8280 km²).

Two methods are tested and compared: the first one is a spatially explicit computation of the cast shadow for every LIDAR pixel. The second is based on averaged vegetation cover characteristics of buffers and reaches of variable size.

Validation of the water temperature model is made against 4 temperature sensors well spread along the stream, as well as two airborne thermal infrared imageries acquired in summer 2014 and winter 2015 over a 80 km reach.

The poster will present the optimal length- and crosswise scale to characterize the vegetation from LIDAR data.