The object of this survey was to study the erosion traces left by a (now dry) modern stream bed and deduce a maximum of information from them such as its maximal flow, speed, regime, soil infiltration, etc.
The first step was to locate such a stream bed. We checked by aerial imagery some areas with promising features before coming to the MDRS using Google Earth and checked them during our first EVA's. We found a bed suitable for our survey on the first day and went back to the site the next day. We had to delimit the study area and chose it according to several factors: easy access on pedestrian EVA, sufficient width of the bed to minimize measuring errors, etc.
We went on EVA each morning to collect data and worked on it during the afternoons and evenings. The first data we collected were the depth, width and inclination of the slope in the main bed. We decided to work on a 670 meters long segment of the main stream bed and took measurements at 13 different points along the bed, most of the time near the intersection with a secondary bed linked to the main one. A clinometer was used to determine the slope while other measurements were taken using a meter.
The Manning Formula was applied on the data to compute the flow and average speed. We deduced the wetted perimeter from the width and depth using several different shape models for the bed: rectangular, circular and a one based on Fermi. This last one was the closest to the reality and was applied to the whole main bed. The result is an increasing flow going down the stream and variable water speed correlated with the width and depht of the bed.
The next step toward a second model was to incorporate the incoming secondary flows into the main one to check if the differences would add up correctly. We took measurements just before the intersection between the secondary and main beds for 24 secondary streams. We collected data in several points within the first ten meters in the secondary beds starting from the main one. By taking the average flow from these three points in consideration for the model we expected to limit the imprecision linked to the measurement method.
After computing the data from the secondary flows with the same method as for the main one, we obtained our second model. By adding the incoming flows upstream for each data point in the main flow we obtained a theoretical value of what it should amount to. This second model only take into account the inputs of the flow but not the outputs such as soil infiltration.
The minors incomes from surface run-off couldn't be measured either and greatly vary depending on the relief along the stream bed. At this point, we decided to take some additional data points in the main bed as we though thirteen were to little do describe the 670 meters-long bed. We used the following EVA to double this amount of data collection points.
Taking the infiltration into consideration, we expected to find a great difference between the two models. By applying a regression on the results and substracting the first model from the second one, we obtain the "missing flow" linked to the soil infiltration. The proportion of infiltration was quite higher than what we first tough as it was around 60%.
To accurately evaluate the soil infiltration proportion, we used a different approach. The inflitration flow was moduled with the surface of the stream bed in contact with the water. We evaluated this value in each data point in the main bad and obtained coherent results. In the first 200 meters upstream, the value obtained was negative and has to be related to the nature of the surrounding soil that lead to a high amount of surface run-off. As explained before, our equipment was not sufficient to estimate this incoming additional flow. However, for the next 470 meters, the value of infiltration was nearly constant. The small variations can be related to the nature of the soil inside the bed as the maximal values were located in sand-covered zones while the minimal ones were situated in silt or clay-covered ones.
We wanted to experimentally confirm those results and performed an infiltration test in the different zones. We first used a small-diameter plastic pipe and inserted it in the ground in order to test the time needed for the water to infiltrate as we poured some inside. This first test failed because the ground was too hard for our pipe to be inserted deep enough.
We had to go back the next day to perform another test using a stranger pipe. The results of this test were coherent with the theory but the values were inferior to our estimations. We explain this difference by the fact that our model estimate the value for a drenched soil while our experiment was done on completely dry soil.
All the coordinates of the measurements points were recorded using the GPS, a map of the area and photos taken from high places around the survey area. By combining these different informations, we could map the whole survey area and the resulting map is quite accurate. Using this as base, the survey area and map could be easily extended to a wider range.
To improve the precision of the model, we had also planned to go back to the main stream bed in order to take photos at short regulars intervals (around 10 meters) to be able to divide the bed in smaller sections. This would allow us to apply to each one of the sections the best shape model based on what we record on site. A statistical approach of the model was also planned but we are reaching the end of our rotation and won't have enough time to proceed to these supplementary process.
The model is using the program Matlab and the mapping of the area is done using ArcGIS.