paging water quality experts

[bagtagley]

I'm looking for some assistance from anybody with the time/willingness to help out. I'm in the process of writing my thesis proposal, and I need to pin down my question.

I'm doing a GIS analysis of water quality data in a region of West Virginia. I've got almost 6 years worth of water quality data that includes TSS, fecal coli, total P, N02-N, N03-N, Dissolved O2, turbidity, pH, etc...no TKN. The one thing I do not have is flow. Out of 99 mapped sampling sites, only 3 of them are at USGS flow guages. All the literature I've read uses flow data to determine nutrient loading, so I'm stuck on that. I know this is pretty open ended, but I'm getting desperate. Ideally, I'd like to defend by mid-November...gotta graduate by December.

[meatdrink9]

I think powho measures amounts of doodie in water (and not just as a hobby unlike myself). He might could mebbe potentially offer some info.

[MacDaddy]

Yea Powho is the man

[homerjay]

Can you extrapolate/interpolate the flow data from the given streamflow measurements? You could find the watershed area associated to the streamflow data points and adjust volumes up/downstream based on smaller/larger representative watershed areas?

Or can you do what we always do in engineering: make assumptions?

You could go out and get flow measurements - measure velocity at different places in the stream and assume a cross-section, Q (flow) = VA where V = velocity, A = x-sectional area.

[Telenater]

Does the network you're working in have more USGS flow meters than just the three that overlap with your sampling stations? If so, you should be able to interpolate a flow regime for most of the network. It wouldn't be at a level of accuracy that I'd like to see for a thesis though.

With only three flow guages assuming that they're on the mainstem of your watershed, you should be able to build a model for the mainstem alone to build a function to relate the flow accumulation to those sample points and at least have a statistically derived regression (with the all important R^2 values though they'll be large). The more flow guages you have the better as far as the regression goes.

The flood event time series is going to be a real challenge though. You may or may not need to take that into account. You know what your region and needs are and will have to make that call. If you do need it, it will take a lot of work (second thesis?).

I've done a fair amount of GIS modeling on streams for TMDL purposes, so drop me a line <username>@hotmail.com

[Buzzworthy]

You go to WVU, just write "I am in West By God Virginia, the water sucks".

Look, now you will be out here in no time.



Actually, I wish I could help, but trees are my thing, your water helps my trees grow, that is about it.

[bagtagley]

powho, apparently my box was full, so I missed your message...sorry.

Box now douched.

[ak_powder_monkey]

What about turbitity readings? not exactally sure whatyou are looking for, but it seems knowing suspended load would help with nutrient load.

Are you looking at marcoinvertebants, BOD, and other things thathave to do with bugs?


Just Throwing out Ideas, good luck

[Lane Meyer]

Coming from the GIS end of things, being a geologist, and working around a bunch of engineers all day, I think you might have a problem. GIS uses several means of interpolation, which you get to choose when you create your GRID which you will use to rasterize the 3 stations you do have. 3/99 = shite no matter how you look at it. What is your spatial resolution if you rasterized your study area? I think you either have to a) make more measurements or b) find a different parameter instead of water flow.

If you get stuck in the GIS with interpolations when you decide how to proceed, PM me, I might have some tricks in my "GIS Bible" that I've been building over the years. I did lots of interopolations of earthquakes and tectonic processes for my MSc thesis last year. What are proxies for water flow that you could more easily measure and substitute? Precipitation? Soil type? Soil thickness? Bueller?

You might try block interpolation, when you can bolster youyr sample set:

From ArcGIS Help:

Block interpolation is used to create an ArcGIS Spatial Analyst Grid from a geostatistical layer. Block interpolation is an interpolation method that predicts the average value of a phenomena within a specified area. The specified area can be an irregularly shaped polygon like a forest stand, a more regularly shaped polygon like a field, or can be a series of equally shaped squares as in the case of an ArcGIS Spatial Analyst Grid.


How block interpolation works is that for an identified area, you calculate predictions for a number of specified locations with in the area, average their values, and then assign the average as the prediction for the entire area. For example, if you specify to predict at 10 locations within an area, then predictions will be made for each of the ten locations using an identified interpolation model and the original sample points collected for the phenomena. The location of these ten points can be either specified by the user or can be randomly or systematically assigned. Based on the distribution of the prediction points, each prediction can be weighted in the averaging process based on the area of influence of the prediction location. Those prediction locations without any other prediction locations nearby will be weighted greater than those locations that have several identified prediction locations surrounding it.


When converting a geostatistical layer to a Grid, the shapes of the blocks to perform the averaging is within individual grid cells. The image below shows the sample points with the block configuration (the Grid) that will be used in the block interpolation.

[dipstik]

If it's brown drink it down,

If it's black send it back!

[Telenater]

Actually, the method I'd recommend for flow interpolation is not to use any interpolation. I don't think that is a feasible plan.

What might work is if you have other USGS guages that don't match with your sampling sites (If you don't you're pretty much up shit creek figuring out flow values). Calculate the flow accumulation and then SAMPLE the flow sites. From this you should be able to develop a regression between drained area and flow with a reasonable r^2.

[Lane Meyer]

Good idea, TN, calculating flow accumulation is an easy thing to do with ArcHydro 1.0 for ArcGIS if you have a good 10m or less DEM of the study area. You can springboard to lots of useful places from that data if you are trying to model the flow of particulates through your field site. I can think of several nasty equations from Groundwater Hydrology you could plung into a GRID showing flow accumulation to make any sort of raster surface model of different water quality parameters.

[GeoMatt]

I think some of the suggestions are good that others have tossed out there. But..... you will need a good answer to the following question when you defend:

Why didn't you take a simple flow rate measurement at each site during sample collection?

I think some of the interpolation methods would be excellent if you had more hard data, but statistics are going to kill you. I wish you luck, but don't envy your position.

[bagtagley]

Originally posted by GeoMatt
Why didn't you take a simple flow rate measurement at each site during sample collection?

The data was collected by the WVDA, so it's not my question to answer. My understanding is that it was a matter of cost.

Nater, I sent you an email yesterday. It's a bit long winded, but let me know what you think.

I'm not extremely concerned with accuracy, what I mostly want is a consistent, baseline comparison of waterquality between sub-sheds. A professor suggested that I determine the drainage area for a downstream flow guage and then use the drainage area of each of my subsheds to determine their percent of the total drainage that feeds into the guage. That will allow me to get a percent of the total downstream flow, and therefore a flow at each location. It's very rudimentary, but he seemed convinced that it would at least pass the "laugh test".


Here's the study area. Sample sites are yellow, USGS guages are red, and delineated sub-sheds are tan.

[bagtagley]

Originally posted by Telenater
What might work is if you have other USGS guages that don't match with your sampling sites (If you don't you're pretty much up shit creek figuring out flow values). Calculate the flow accumulation and then SAMPLE the flow sites. From this you should be able to develop a regression between drained area and flow with a reasonable r^2.

Nater, can you elaborate?