Online TA Help
To facilitate your learning of the subject material for this course, this page has been set up to allow you to ask the professor questions about the readings, lectures, homeworks, and other related topics. Your questions and their answers will be posted (anonymously)here, so that all students may benefit from the answer, since it is likely that several people are wondering about the same thing.
Before sending in your question, be sure to re-read the appropriate textbook section, re-visit the corresponding lecture notes, and consider any relevant handouts. Please take the time to carefully phrase your question so that it may be properly answered. The more specific and well-phrased the question is, the more direct and useful the answer will be to yourself and your classmates.
Send your question to firstname.lastname@example.org
Questions and Answers
Q: (HW#1) I was wondering what exactly number 1 means? I guess I don't understand the phrase "suite of basic metrics."
A: Good question. Whenever you get hold of data it is always wise to start off your thinking about that data with a suite of basic statistics to get a sense of what the data is all about. The word "metric" means a numerical quantity that you use to evaluate something. When you use statistics, you can learn things like the central tendency of the data (i.e. mean and median), the spread of the data relative to the central tendency, like is the data widely spread or tightly clustered (i.e. standard deviation), is the data more heavily on one side of the mean (i.e. skewness), and how much data is there. Knowing these numbers allows you to start thinking critically about the data. Does it seem to you like these data are too sensible, too high, too low, etc, I am guessing that school has mostly taught you how to mindlessly calculate this and that. Now it is time to start trying to interpret the numbers in terms of the physics of the real-world that they represent. Is 10 mm/week a lot or a little? This is your chance to start building up your understanding of what such numbers mean.
Q: (HW#1) What are the units for the field data?
A: The units are millimeters and the data are gross weekly values.
Q: (HW#2) In using the field data you provided for the Gash's model homework, by sampling only weekly and not after each storm, aren't you saying that each storm lasts for one week?
A:Yes. The data is gross weekly, consistent with the measurment strategy of that study, but the model is an "event" model. If a storm occurred over a long weekend, then it would be broken up into 2 different weekly events, which is technically wrong. Just goes to show that you can often make a model "mimic" anything, regardless if the simulation is representing the actual process.
Q: (HW#2) How exactly are we supposed to differentiate between nocturnal and diurnal storms?
A: Considering that the data is weekly observation and there is no pattern of storms starting and stopping at week breaks, then what you have uncovered is a mismatch between model assumptions and field methods/reality. This is a common occurrence. The approach is forge ahead anyway, but be mindful of the potential consequences of violating the model assumption.
Q: (HW#2) When I use the regression of T versus Pg to calculate S, I get a negative y-intercept value for S. Is this right?
A: The y-intercept value you got is right, just ignore the negative sign. You could also use the x-intercept if it is positive, which says that the canopy storage (S) is all the rain that falls before throughfall begins. The y-intercept and x-intercept values have both been used in published papers, they are very similar in values, and they mean essentially the same thing. Recall that in this assignment you are violating some of the assumptions of Gash's model, so it is understandable that "calibration" of the model will have some issues.
Q: (HW#2) How can I obtain "p" when we have no data on throughfall for storms with Pg<1.5mm?
A: If you extend the regression line back, how much throughfall would there be for that low of a gross rainfall? Think about it.
Q: (HW#2) I don't quite understand how you want us to show how the canopy parameters were obtained. Do you simply want the plots we used (which are essentially the ones from homework 1)?
A: How about showing the plots and then circling the value you picked or drawing a line to the axis to show the intercept.
Q: (HW#4) When I try to use solver, the target cell does not become a real number, but says "#DIV/0!". I also see that I/Pg has become set to zero. What gives?
A: Some PC versions of solver appear to "test" the parameter value I/Pg = 0, which violates the model and results in a failure of solver. If this is happening to you, then simply add an additional constraint to the solver dialogue box stating that I/Pg >= 0.001. This will prevent I/Pg from being 0 and then it should work
Q: (HW#4) I have a question concerning the inclusion of the difference of the summed values in the Validation worksheet. Namely, using this value as the 18th "event" in calculating average error seems wrong. Doesn't this error (sum of measured vs. sum of predicted) get covered by the average error from the 17 Storm events?
A: Averaging and summing are related in general, but the final equality is be dependent on the order of operations. For example, the average of sums does not equal the sum of averages. In this case, the error of the sum does not equal the error of the average. In earlier model runs I was getting different optimized parameters from the two error estimates, but now I get convergence toward the same values for both error estimates. When you think about it, the error of averages weighs each stom equally, regardles of storm size, such that the error in a 1 mm rainfall event is the same as the erro in a 100mm event. Field measurement error is scale dependent (it is easier to measure big storms accurately than small storms), so having an error estimate that is scale independent may not be for the best. Other optimization schemes are possible, for sure.
Q: (HW#4) Should the value for 'p' be changing with each optimization? It is not one of the tunables and it is just remaining at zero after each optimization. It was originally entered as zero from the field data. The model error values, and the values of I/Pg and pt all did change as expected.
A: No. It is fixed. In theory one could add it as a tunable, but with the constraint that it be pretty small. Not part of this assignment though.
Q: (HW#4) For the second optimization, I am not sure how to minimize the average error for only F, instead of for all three fluxes. I ran the solver with the target cell linked to the validation sheet cell K43 as instructed, but it looks like doing that includes all three fluxes.
A: K43 should point to just one cell, whereas the other one, was it K22, had a big formula in it. If K43 is not one cell for you, then delete it and point it to the stemflow average error value.
Q: (HW#5) In terms of the "best fit line" for the relation between discharge and drainage area , would a polynomial line to the order of 4 be incorrect? The R^2 value is 1 with this line so it is clearly the best fit...however, is that too high?
A: I think that is one of those cases of lying with stats in a manner of speaking. Given points, one can make any number of wiggles to exactly go through those points. In the limit, we call it a "spline"- a sufficiently smooth piecewise-polynomial function that goes through all the points. That is not the goal of the question or what we are often trying to do in hydrology- not to mimic data, but to explain it. What we are looking for is a causal relationship between X and Y. For example, for constant rainfall on land, as Area increases, then there is a linear increase in runoff- think about what that is- maybe play with your kitchen sink, too. So if you do not see a linear response, then you need to think through what is happening. Something is happening here. We don't want to hide that in a spline, we want to acknowledge the problem and solve it.
Q: (HW#5) I don't understand what you mean by drawing a stick figure diagram of the hydrology. What is that?
A: The big circles represent rainfall-derived water (i.e. storm excess after "losses") running off the land and down streams that are not your focal stream. This is the land draining to a station, ***except for*** the water coming down the main stream. That water was already accounted for by the next upstream gage. If there is no upstream gage for the station you are thinking about, then ok, all the water is in the big circle and that connects directly to the station with no stick between the two.
For downstream gages, you've got water "routing" through the channel from one station to the next PLUS the additional water coming from the land only draining to the next gage. SO every downstream gage has at least two things coming into it.
It is up to you to decide how much detail to put into your schematic. You can represent every tiny stream. You can't even represent even every medium-sized stream given the size of the Salinas. You have to be mindful and make conscious decisions. What do you think are the important contributors that should be represented explicitly and what are the ones that can just go inside a big circle.
It helps a ton if you get yourself a good map of the Salinas. One way to do that is to download Google Earth for free. Then do a web search for something like "USUS topographic maps for Google Earth" and see if you can add those in to your view. Also, you can try to see if you can download USGS or CDEC gaging station locations as points on the map, too.
Q: The date field in the wetland spreadsheet looks wrong.
A: Be careful when you download and start to use the file "tides.xls". This file has a date column that should begin at 04/12/96 10:20. Times should increment by 10 minutes per data point. Excel uses a different date scheme for different versions of its software, which can cause the dates to appear wrong on your spreadsheet. If you encounter a problem, the best solution is to write in the correct date in the first 2 rows and then autofill the date down to the bottom. If you need further help, come see me.
Q: Is the water level measured actually the rise and fall of the water level at some point in the bay or is it the water level that flowed onto the marsh areas? So the reference point would be the elevation of the marsh land which might be higher than the lowest level of the tide?
A: The water level is measured in a well on the marsh surface, not in the open water part of the estuary. Therefore, when the water surface elevation of the tide drops below the elevation of the moarsh surface at the sensor location, the sensor is no longer recording tidal processes.
Q: What water level is considered to be flooded?
A: "Flood duration" is defined as the amount of time the water level is greater than some meaningful datum. Zero in the dataset is the elevation of the marsh surface, so anything higher than that could be considered flooded. Of course it's more complicated than that in reality because the marsh surface has bedforms, roots, etc. I suggest choosing a level of 0.1 feet as a reasonable criterion. You can calculate flood duration in excel by making a cell formula of the type "If level>0.1, then 1, else 0". Next, add up the number of 1's and divide by the total number of points per day. The fraction of time >0 is the flood duration. Q: I have been donig all the graphs, and because there are so many dates, I can't really appreciate the small tidal changes when I graph it all. Are we suposed to graph all the time period in one graph or should I split it in intervals of a few days, or daily...?
A: I suggest plotting ~2 weeks at a time.
Q: What do you mean by "distribution type" in question 4?
A: In statistics, random populations reveal themselves when you see enough individuals and they take the shape of a "distribution". The bell curve (aka "normal" distribution") is one type of distribution. If the hump is shifted to the left then that curve becomes a "log-normal" distribution. If the curve starts high near zero and drops rapidly to the right, it is an 'exponential' distribution. A big listing of types is given at http://la.znet.com/~sdsampe/distr.htm#1
Q: Do we have to change or re-specify time units in order to make a water level vs. time graph that can be deciphered? I made a line graph based on highlighting the two columns, and unless I use the pointer to discern values, I can't tell which direction (i.e., time direction) is which as the lines go up and down.
A: Sometimes Excel chooses a strange way to plot the x-axis, such that you don't see smooth line changes. To fix this problem, in the Chart Wizard there is an axis option tab that allows you to force Excel to plot the x-axis as "categories". This will fix the problem.
Q: I am trying to determine which time I should choose for the maximum and minimum tides during a cycle. In general, there are multiple times when the tide has the same value at the maximum or minimum.
A: First ask yourself whether this is a semidiurnal, diurnal, or mixed system. Then you should know the period of the tidal cycle and have an expectation to see high and low tides every cycle. As you follow the water up and down, select consistent criteria for how you identify high and low tides and always stick with them. Many possible criteria exist, so rather than tell you one, I will just say that you should report whichever one you choose.
Q: I am trying to get the "percent rank" function to work. Every thing I have tried to enter has some part were the screen says "#value". Here is what I have been doing. I made a column of numbers 0 to 2.96 increasing in 0.01 intervals, I enter this into the "array" box of "percent rank" in the "x" box I entered the water levels.ÊThis gives me a list of numbers for a brief second, then the numbers change to the "#value" This has been really frustrating for me. So then I try and reverse the values for "x" and "array" and this gives me a column of numbers saying "9%" for all the values.
A:ÊIt is useful to use the INSERT->FUNCTION menu option in Excel when problems arise, because that can walk you through the process and show you a sample result. The "array" is the water level data, as that defined the probability distribution of water levels as they truly exist. Use the full amount of water level data provided. The "x" is the 0.01 increment number for which you want to know how often that is (or is not) exceeded. As x increases, the percent rank (i.e. percent of time not exceeded) should increase from close to zero to nearly 1.
Q: When I am trying to "autofill" a column in Excel using the PERCENTRANK funcion in the first cell, how do I prevent some of the cells from incrementing in number. For example, if I want to do PERCENTRANK(A1:A10,B1) and I want the A's to not shift?
A: Use the dollar sign symbol in front of any identifier that you do not want to shift. In your case, it would be PERCENTRANK(A$1:A$10,B1). For holding both the row and column constant you can do $A$1, e.g.
Q: I am trying to get the output file into excel to graph things. Is there a way to easily import the file into excel?
A: The easiest way is to open the output file in a word processor such as MS Word, select and copy the hydrograph data table, paste it into an Excel spreadsheet, and lastly use the "Text to Columns" command to convert the information into actual data.
MISCELLANEOUS QUESTIONS BEFORE MIDTERM
Q: I was looking over the 2nd reading in the Watershed Hydrology book, starting with 'Vegetation Manipulation' when it struck me that there seems to be a funamental difference in how this Black guy thinks about watersheds and how I do. He seems to be portraying watersheds as things that can (and should?) be tweaked to maximize water yield without sacrificing water quality. My impression is that increasing water yield is probably the last thing you want to do, since that usually leads to higher high Qs and lower low Qs, ultimately leading to flooding, lowering of the water table, and changes in the flora of the watershed, among other things. Is it just the hyper-flood-sensitive-Californian in me that makes me think Black's perspective is a bit skewed, if not downright naive? Or does his writing reflect the general attitude of watershed hydrologists?
A: Prof. Black's attitude reflects the traditional perspective of rangeland and agricultural hydrologists who see the watershed as a system to be manipulated to serve the economic interests of our modern food industry. The goal of hydrology in this view is to maximize streamflow, because that is the only term in the water balance that is a "point source" of water. Infiltration, ET, rainfall, and interception are all diffuse sources. When a watershed is deforested to maximize streamflow, the timing of flow is not continuous (since baseflow is drastically reduced), but the increase in total streamflow volume is more important than timing from a water yield perspective. Certainly, the idea that a watershed should be made to serve a single economic function is counter to what we now know about the importance of natural ecosystem functioning for both the environment and society. In the past, when a watershed has been altered to aid one societal value (e.g. hydraulic mining of uplands), there has been major degradation of other societal values and natural functions (e.g. floodplain agriculture, in-stream habitat, urban flood protection, etc).
Q: I have a question related to practice question #2 in the forest hydrology section. This also has to do with HW#2. In HW#2, we were supposed to find the trunk storage capacity (St) from the y intercept of the graph of stemflow vs. precip. However, when I look at the graph in Practice Question #2, it seems that the x-intercept should be used as the trunk storage capacity, since when stemflow equals zero (x-intercept), that means all of the rainfall is stored in the trunk, right? Another way to put it is that the x-intercept represents the maximum amount of Pg at which there is no stemflow, while the y-intercept represents the stemflow when there is no Pg, which ought to be zero all the time, but in reality is not, due mostly to regression methods and not reality. In Practice Question #2, I would say that the trunk storage capacity is about 0.2mm.
A: People have used both the x-intercept and the y-intercept in peer-reviewed journal papers that I have talked about in class. One author used both of those approaches and even a direct measurement, and then compared the different values. Conceptually, it makes sense to me to use the x-intercept for the reason you describe in your question.
Q: To obtain the parameters for the LS card, do we simply calculate a weighted average of CN values which accounts for the percentage of each land use/soil type combinations in each sub-basin?
Q: To obtain the parameters for the UD card, I'm assuming "n" is Manning's n, "L" is only the length of the main stem of that subbasin, "Lca" is the number you gave us in the box (but I don't know what it represents) and "S" is the slope of the main stem? But this seems to leave out the tributaries, so perhaps we just lump them all togther or maybe calculate a time lag for each segment of stream and either average that or create a separate RD card for each segment? The manual doesn't really seem to explain this much...
A: For the UD card, L is not necessarily the length of the main watercourse, but rather the length of the *longest* watercourse in that subbasin. S is the slope of the longest watercourse. Lca is the length along the longest watercourse to the centroid of the subbasin. You can figure this out by determining the center of the subbasin and then projecting from that point over to the longest watercourse. All of the other parameters are just as you guessed. Please use n=0.07 for questions 1 and 2 of the homework for uniformity. As for the tributaries issue, the point in rainfall-runoff estimation is to figure out how long it will take water to move to the mouth of the subbasin, and that is generally governed by the most distant drop that must travel down the longest watercourse. There are more precise approaches to figure out the time lag, but in this class the simple equation of the Contra Costa County Flood Control District (given in the homework handout) will do.
Q: To figure RD, again do we just calulate the slope and length of the mainstem for each subbasin or incorporate the tributaries in somehow? I could see if the tributaries were included in UD, then maybe just using the mainstem for RD would make sense, but I don't really know. Again, the manual is a bit vague...
A: In this case, the tributaries really don't matter because you have already collected the water from them in the rainflal-runoff subroutine. For routing you are just moving water downa channel from point X to point Y. So now you need to use the length of the main channel.
Q: Can the hydric soil be considered to be a type A soil? Also, what is the appropriate land use for the wetland? I put it as a meadow to find the CN number from the table you gave us in the handout the first time, but now I'm not sure that is right!
A:Think about how wetlands function hydrologically. Do they absorb a lot of water like a type A soil or do they lock water out like a type D soil? The table I gave you with CN numbers does not include wetlands, so think about where a wetland should be relative to some of the other land surface types listed.
Q: On the RD cards, you emailed us earlier about what to take for the values of n, SHAPE, WD, and Z, and that was for routing from D to C, C to B, and B to A. What about routing from E to D? And does each of those routings go with the sub-basin in which the routing is being carried out?
A: I don't think you are understanding the concept behind how the model is working. In a given subbasin the rainfall-runoff computation determines the hydrograph for the outlet of that subbasin based on all of the water that runs off in it. So for subbasin 1, all of the rainfall produces the hydrograph at point D. The time lag term in the UD card accounts for the rainfall-runoff travel time through that subbasin for the rainfall-runoff water. Hence there is no hydrograph to route from point E to point D. Once you have the water at point D it is no longer in the rainfall-runoff mode. Now it is in the routing mode. To move that water from point D to point C you use the RD card. In the meantime water has been raining in subbasin 2. The rainfall-runoff for subbasin 2 is calculated independently of the water from subbasin 1 that happens to be moving through subbasin 2. After the calculations for the routing from D to C and the rainfall-runoff for subbasin 2 are computed, it is necessary to add the 2 hydrographs together because both of those sources of water must pass through point C. This procedure is continued through the entire basin.