.. _personalProject: Personal Project ================== Introduction -------------- The personal project is done in groups of students. All members of a group will get the same mark. Create your group by self-subscribing to a Personal Project Group (Blackboard: Course Content -> Personal Project Groups). Most topics include literature study and modelling. There are also topics that only include literature review. Before starting, it is important to define research questions. This will determine what modelling work you will do and how you will report on the modelling results. For some topics I give hints for research questions (see below). Tutor support is provided during the scheduled lab hours (see the course schedule) or on request by e-mail. Research proposal --------------------------------- At the start of the personal project you need to compile an approximately 1 page research proposal. It needs to include (at least): - Topic - Problem definition - Objective and research questions - Methods - Discussion points / questions you have Upload the proposal to Blackboard before the due date. You will get feedback from your tutor(s) mainly to help you with executing the project. The research proposal is not marked. Report on personal project -------------------------- The report should be written as a scientific article with the following structure: - Abstract - Introduction (problem definition, research questions or objectives, outline of the rest of the paper) - Methods (here: description of the model and/or scenarios done) - Results (provide results of model simulations, without extensive discussion) - Discussion and conclusions (discuss results, compare with other studies, provide main findings) - References For the topics that do not include hands on modelling you can consider using a somewhat different different structure for the paper. The report should not be longer than 4 pages (12p font, single line spacing, including figures and tables). If needed you can provide tabulated data, additional figures, or details on implementation of the model as an appendix. Upload the report to Blackboard using the Blackboard Assignment called Personal Project Report. Final presentation on personal project --------------------------------------- In the last week of the course you will present your personal project in a presentation of ~10 minutes. The presentation needs to include topic, problem definition, key objective / research question(s), methods, results, and discussion. Topics (pick one) -------------------- Land degradation modelling +++++++++++++++++++++++++++++ Indication for the content of the topic: literature study 20%, modelling 40 %, writing report 40%. The output from rainfall-runoff models can be used to calculate water erosion. Extend the model that you used in the calibration computer labs (download below) with a component that calculates a map with the total amount of water erosion. Use the Morgan, Morgan and Finney model described in Morgan, R.P.C., 2005, Soil Erosion and Conservation, Blackwell. You at least need to include detachment of soil by raindrop impact and runoff, ignoring the transport capacity (i.e. assuming everything can be transported). Compare modelled soil loss with tabulated values from the literature (to see whether the order of magnitude of the modelled values is correct); this could be one of your research questions. `Download dataset and model. <_static/calibration.zip>`__ `Download additional source data in case you want to run the model for another time period. <_static/observed_meteo_and_discharge.zip>`__ The Morgan (2005) book is available in the Files section of the MS Teams team of our course. Suggestions for extending the model: it is recommended NOT to calculate erosion for each separate time step as the Morgan, Morgan and Finney model is designed to calculate total erosion over a long time span (typically one year) from inputs (e.g. precipitation) summed or averaged over the same long time span. So you need to aggregate (sum/average) values from the hydrological model and use these as input to the erosion model. At the bottom of the dynamic section, add something like the following code: :: # example how to calculate total precipitation, this statement adds precipitation in the current # time step to the total precipitation self.totPrecip = self.totPrecip + precipitation # at the last time step (change 'lastTimeStep' below to the last time step in your model) # run the Morgan, Morgan and Finney model if self.currentTimeStep() == lastTimeStep: # this part is run only at the last time step # and you can add here the calculations for the static # MMF model, for example print 'running MMF model' fractionOfTotalPrecipitation = self.totPrecip*0.2 self.report(fractionOfTotalPrecipitation,'fr') In the initial section you need to initialize the total values, for instance for precipitation this would be: :: # initialize total precipitation self.totPrecip = scalar(0) If the hydrological model is slow, an alternative approach would be to report these totals (e.g. total precipitation) calculated in the hydrological model to disk and creating a separate static model (without time steps) that reads these maps from disk and calculates the erosion. During development of the erosion model you then only need to run this static model calculating the erosion. Early warning signals of critical transitions +++++++++++++++++++++++++++++++++++++++++++++++ Indication for the content of the topic: literature study 40%, modelling 20 %, writing report 40%. In the dynamic modelling practicals you developed a vegetation model with a critical transition to lower biomass when a threshold of grazing pressure is exceeded. It is notably difficult to predict such critical transitions as the change in biomass is very low before the transition occurs. However early warning signals exist that show a change well ahead of a transition. These include spatial variance of biomass (you calculated this in the exercise), spatial skewness of biomass, and so-called 'flickering'. Study the literature below. Extend the vegetation growth model (the one you constructed during the practicals) with calculation of spatial skewness of biomass (calculated over the map for each timestep, just like variance). Compare the two early warning signals (variance and skewness) regarding their capability to forecast the transition. If you want you can do this for different scenarios of model parameters (e.g., spatial diffusion parameter). `Download dataset and model. <_static/vegetation.zip>`__ Literature: - Dakos V., van Nes E.H., Donangelo R., Fort H. & Scheffer M. (2009). Spatial correlation as leading indicator of catastrophic shifts. Theor. Ecol., 3, 163-175. - Guttal V. & Jayaprakash C. (2009). Spatial variance and spatial skewness: Leading indicators of regime shifts in spatial ecological systems. Theor. Ecol., 2, 3-12. - Scheffer M., Bascompte J., Brock W.A., Brovkin V., Carpenter S.R., Dakos V., Held H., Van Nes E.H., Rietkerk M. & Sugihara G. (2009). Early-warning signals for critical transitions. Nature, 461, 53-59. Hydrological model: calibration ++++++++++++++++++++++++++++++++++++++++++ Indication for the content of the topic: literature study 5%, modelling 65%, writing report 30%. In this assignment you will calibrate a hydrological and snowmelt model that you know from the practical exercises. You will run it for the Dorferbach catchment in the Austrian Alpes. You need to calibrate on the observed discharge at the outflow point. Have a look at the scripts and readme.txt files included in the zip files below and explanation of the data set. Suggestions for possible topics (the first two options can be done in the framework of this course, the other options are quite challenging and suitable only if you are sure you have the capabilities): - Improve upon the calibration by calibration of multiple (all?) parameters in the model. Use the brute force technique and code it yourself (using Python or numpy) by extending the runoff.py script. - Extend your analysis towards an independent validation period. That is, calibrate the model for a certain time span (e.g. year 1-3) and validate (test) it for another time span (‘prediction’), e.g. year 4-6. For this you need to extend the time period you run the model, this can be done with the second download below. You may want to stick to calibrating one or maybe two parameters. - (Very challenging - only do this if you are very good at programming; i.e. a level above what is expected in this course) Same as above but using SPOTPY (https://pypi.org/project/spotpy/), which provides built-in calibration routines. This will require you to write code that calls the model from SPOTPY Python functions. - (Interesting but only possible if you know a lot about statistics or machine learning) Compare the capability of the model to predict discharge with a statistical learning algorithm. An alternative to using the simulation model (and calibration) is to train a statistical learning model (e.g. regression, random forest, deep learning) on the data (it should be able to predict streamflow from precipitation and temperature alone without using any other model or the map data). Test this approach and compare the results with those from calibration of the simulation model. Note that it is preferable to separate between training (calibration) and validation (testing) time periods, like 3) above. - (Interesting but the compilation of the data may be a lot of work). The current data set used for meteorology is somewhat old. A more recent reanalysis data set with meteo data is ERA5 (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5). The idea here is to compare model runs that rely on this alternative meteo input. You need to download the ERA5 data for the correct location of the catchment and resample it to a 1 day time step (the download is in 1 h timesteps). And then run the model with the same data (and possibly calibrating it again). `Download dataset and model. <_static/calibration.zip>`__ `Download additional source data in case you want to run the model for another time period. <_static/observed_meteo_and_discharge.zip>`__ Land use change modelling using cellular automata +++++++++++++++++++++++++++++++++++++++++++++++++++ **Note**: if you work alone, only select this topic if you are able to quickly program a model as it takes quite some PCRaster Python scripting Indication for the content of the topic: literature study 10%, modelling 60%, writing report 30%. Cellular automata is a type of model that uses local neighbourhood interactions to simulate the larger scale behaviour of a spatio-temporal system. These local neighbourhood interactions are given by transition rules valid for each cell on a grid of cells, where the state of a cell changes over a timestep as a function of the state of cells of directly neighouring cells. Cellular automata are widely used in the spatial sciences, e.g. for modelling plant growth and spread, modelling forest fire spreading, modelling growth of bacteria on leaves of vegetation, modelling socio-economic systems. The aim of this topic is to learn more about cellular automata modelling in general, by studying literature. In addition you will construct a simple cellular automata model of expansion of cities (Randstad, the Netherlands). Start with the landuse situation in 2000 (as represented by the data set) and try to simulate the change in landuse over the coming decennia. Use simplified landuse change transition rules - the approach is more important than the outcome! Alternatively, there is a possibility of using an existing (large) land use change model that you can use for a case study area in Mozambique (e.g. for a scenario analysis). Literature (available in the Files section of the course MS Teams team): Torrens, P.M., 2000, How cellular models of urban systems work (1. theory). Centre for advanced spatial analysis, working paper series. Paper 28. Batty M., Xie Y., Sun Z., 1999. Modeling urban dynamics through GIS-based cellular automata. Computers, Environment and Urban Systems 23:205-233. White, R. 1998. Cities and Cellular Automata. Discrete dynamics in Nature and Society 2:111-125. `Download dataset and model. <_static/randstad.zip>`__ `Download information regarding the data set. <_static/appendix.pdf>`__ Validation of models in the earth sciences ++++++++++++++++++++++++++++++++++++++++++++ Indication for the content of the topic: literature study 50%, writing report 50%. In 1994 Oreskes et al published a paper discussing validation of numerical models in the earth sciences. Their main message was that validation of models is not possible. This raised a lot of dicussion in the earth science community. You can find hundreds of papers citing the Oreskes paper. Read the paper by Oreskes and collect a number of other papers on the same topic (e.g. those that cite Oreskes). In your report either provide a review of these papers or provide a discussion on validation of models in the earth sciences. This topic is a good choice if you are interested in philosophy of science. Literature to start with: - Oreskes N., Shrader-Frechette K. & Belitz K. (1994). Verification, validation, and confirmation of numerical models in the earth sciences. Verification, validation, and confirmation of numerical models in the earth sciences, Science, 263, 641-646. Social factors in model building +++++++++++++++++++++++++++++++++++ Indication for the content of the topic: literature study 50%, writing report 50%. An important step in modelling is the identification of the model equations, that is, the choice the modeller makes in selecting the methods to represent processes simulated by a model. This is, amongst other factors, driven by theory, the goal of modelling, the available data, the software tools used, and the hardware available. Recently, researchers have paid attention to the role of social factors in this model identification step. The choices a modeller makes while building a model are partly driven by his or her social network, e.g. colleagues in a research team, supervisors, or fellow students. In this topic you will read some articles on social science & modelling and write a paper on this topic, for instance aiming at describing the mechanisms of how social connections influence model building. Literature: - Babel, L., Vinck, D., & Karssenberg, D. (2019). Decision-making in model construction: unveiling habits. Environmental Modelling & Software. https://doi.org/10.1016/j.envsoft.2019.07.015 - Packett, E., Grigg, N., Wu, J., Cuddy, S., Wallbrink, P., Jakeman, A., Mainstreaming gender into water management modelling processes, Environmental Modelling & Software, 2020, 104683, https://doi.org/10.1016/j.envsoft.2020.104683. - Melsen, L.A., Teuling, A.J., Torfs, P.J.J.F., Zappa, M., Mizukami, N., Mendoza, P.A., Clark, M.P., Uijlenhoet, R., Subjective modeling decisions can significantly impact the simulation of flood and drought events, Journal of Hydrology, Volume 568, 2019, Pages 1093-1104, https://doi.org/10.1016/j.jhydrol.2018.11.046 - Melsen, L.A., Vos, J., Boelens, R. (2018). What is the role of the model in socio-hydrology? Discussion of “Prediction in a socio-hydrological world”, Hydrological Sciences Journal, 63:9, 1435-1443, http://doi.org/10.1080/02626667.2018.1499025 - Addor, N., & Melsen, L. A. (2019). Legacy, rather than adequacy, drives the selection of hydrological models. Water Resources Research, 55, 378– 390. https://doi.org/10.1029/2018WR022958 - Jantiene E.M. Baartman, J.E.M., Melsen, L.A., Moore, D., van der Ploeg, M.J., On the complexity of model complexity: Viewpoints across the geosciences, CATENA, Volume 186, 2020, 104261, https://doi.org/10.1016/j.catena.2019.104261 Model comparison +++++++++++++++++++ Indication for the content of the topic: literature study 50%, writing report 50%. There are almost as many models as there are researchers or research teams. Model comparison studies are quite often done, e.g. to see what model performans best. In this topic you will compare three different models that have been built with approximately the same approach. You can define the criteria for the comparison yourself; it might be most interesting to compare them regarding the model structure, that is, the equations used to represent the processes modelled. You will see that quite many different approaches exist for more or less the same system. First choose a particular subsystem, e.g. global hydrology, water erosion at catchment scale, debris flows, or land use change. Then search for three papers that use a model to simulate this system of your choice. Then compare them based on the criteria. You do not really need to aim for selecting 'the best model', it is more about learning and describing how the same system can be modelled in various ways. Writing a short paper: checklist, misc. recommendations ---------------------------------------------------------- - Do not hand in a report that is longer than the maximum number of pages allowed (see above). - Hand-in your report in time. See the course schedule for the deadline. - Use Italics ('cursief') for all symbols in equations or in the text. However a vector ('list of values') is mostly given in bold. The style in the equations and in the main text should correspond. - Avoid multi symbol variables or parameters in equations. E.g. Evap = Soilwater / SoilP . Better: E = s/a. Use subscribts when you have many parameters and variables. Note however that in programming, the use of long variable names (that describe the content of the variable) names is recommended. - Do not write like in a diary ('First we did this,.... Then we started to realize.. and we did this and that...'). - Put larger blocks of computer code (say, more than 2 lines) in a table instead of inserting it in the main text. Whole programs should be given in an appendix. - Use a main (cover) title that makes sense. - Provide quantitative data in figures (bar graphs, line graphs, scatter plots, use e.g. Excel, Splus), not tables. It is allowed using a table but it is almost always better using figures. - If you write the report with Microsoft Word, use Microsoft Equation editor (available in Word) for equations. Do not copy paste bitmaps (gif or tifs) of equations from other docs into your paper. - Number equations - always (provide the number after the equation, e.g, (3)). - Check out an article from a scientific journal (e.g. from your reader) and use that as an example for formatting, layout, use of figure captions, literature references, etc. - Do not use language as if you are talking (spreektaal) - When you submit your report by email, put everything in one file (word or pdf). Do not send a whole bunch of files (it is too much work printing everything). - Describe content in a logical order, instead of describing content in the order you dealt with it while modelling. So, do not use sentences like 'eerst deden we dit, toen zijn we dat gaan doen, etc..). - Use a spellchecker (always) - A caption of a figure or table should at least explain all symbols used in the figure or table. The same holds for an appendix. In principle, the table/figure should be understandable without reading the main text (although there are exceptions to this rule) - Provide figure legends (always) - If a figure contains a map, provide a scale (scale bar) - Do not hand in black and white prints of color figures (never.., even not when emailing originals..) - Use the same format for each reference in your literature list and refer to the references in the text. - Number the sections in your report, provide these numbers also in the contents Preferably use some kind of hierarchical numbering, for instance 1 1.1 1.1.1 1.1.2 2 2.1 2.2 etc - Number figures and tables. In the main text, refer to figures or tables by using these numbers. - All literature refered to in the main text should appear in the literature/references section at the end of the report. Check this in detail before handing in! - Do not mix past and present tense. Sometimes it is possible, but in many cases it is better to stick to past or present tense. - Avoid the use of 'I' or 'we' (1e persoon). However you can use it sometimes (if you really want and think it is apprpriate). - If you use a figure from a book or another report, always provide the reference. - Do not use English terminology in a report written in Dutch when correct Dutch terms are available (e.g. 'catchment' = 'stroomgebied') - Have a look at comments on earlier papers you wrote. Take them into account when writing your next paper. - Provide units (all variables in equations)! - Avoid the use of abbreviations. - Do not use 'etc.' - Provide page numbers. - Write concise. Also, do not add figures that could be left out. And, if possible combine to figures in one figure (e.g. two lines (of the same attribute) in a graph is better than two graphs each with one line). You can also use panels, Fig 1A, Fig 1B, etc. - Use every page from top to bottom (apart from last pages of very long sections), do not include too much whitespace! - Do not just copy-paste figures (maps) from screen. Adjust colors, add a legend, remove MS Windows bars, buttons, check size of text or modify text, etc. Use a graphics package (e.g. Freehand, Paintshop or whatever). - Try to come up with interesting results (do not just list all results from your model, but try to emphasize the most interesting results). But note that this should always fit with the goals of your research (if needed adjust these goals). - Use courier font for computer code, PCRaster scripts, or filenames. Also in the main text (not just in tables). - In a paper reporting research in the geosciences, you should avoid the use of computer code to explain calculations. Many people will not know the programming language you used, and they won't understand the code. Instead, explain calculations using mathematical equations. If you write a report on a geoinformatics related topic (e.g. how you construct a piece of software) you *can* however include code as code is the topic of your research. - Use a good dictionary. If you do not have one, buy one. I could recommend Longman Dictionary of Contemporary English (http://www.longman.com/ldoce). Or use the online version at http://www.ldoceonline.com. Google translate is also useful. - Read through the text and correct all small (or large..) errors (typos for instance) before handing in! - Do not use a title that ends with a ':'. For instance, do not use the title 'Discussie:' - Do not come up with things in the Conclusions section that have not been described earlier in the paper. - 'Introduce' an equation. Do not just put your equations somewhere between the text. You need to introduce it by stating e.g. 'Evapotranspiration is calculated as: '. Below the equation be sure to explain ALL symbols (except if they were explained earlier, however it does not hurt explaining a symbol again if it was explained 10 pages back..). - Do not write 'wouldn't', 'doesn't', etc.