Unit 8 — Landscape evolution, human impacts and research practice

How can competing explanations be tested using measurements and models?

Geomorphic interpretation becomes strongest when it makes falsifiable predictions about changes in form, material and time. A land-use change may alter infiltration, runoff routing, sediment connectivity and bank strength without producing an immediate visible landform. Similarly, a restoration project may alter local channel stability while transferring sediment impacts downstream. Landscape models range from conservation-of-mass boxes to distributed numerical frameworks; every model is a simplification with a domain, grid size, time step, calibrated parameters and unresolved processes. Model fit does not guarantee correct mechanism when different parameter combinations can reproduce the same profile. A robust study includes independent validation, sensitivity analysis and publication of assumptions. This final unit develops an end-to-end evidence audit from spatial data acquisition through field observation, sediment budgets, alternative interpretations, scenario simulation and ethical communication. Students should learn not to treat a color-coded hazard map as a precise prediction or to publish a clean curve without uncertainty bounds.

Learning outcomes and scale discipline

By the end of this unit, students should be able to describe at least five independent landform or sediment observations, demonstrate one dimensionally consistent quantitative calculation, design a comparative measurement, and distinguish an interpretation from the observations used to support it. Practice viewing the same landform in a map, a section and a material description. Keep an explicit column for phenomena that are absent, unobserved or only assumed. At the finest scale, a few grain contacts may govern transport; at larger scales, inherited valley geometry and event histories control the response. Switching scale is legitimate only after specifying which variables remain meaningful. At the start of the field exercise, make a drawing of baseline and impact sites, DEM change maps and documented parameter ranges; list which parts would be measured directly and which reconstructed from other evidence.

The unit's measurement choices should include model grid and time step, monitoring period and independently observed outcomes; the strongest report also checks held-out validation and uncertainty-aware metrics. Even a carefully derived rate does not establish its formative mechanism unless the sedimentary or stratigraphic evidence agrees. Interpret changes in the context of the process and observation interval, and use at least two sources of independent evidence. Most useful scientific uncertainties are not just numerical instrument errors: researchers must also consider incorrect boundaries, missing events, biased sample sites, mistaken ages and incomplete constitutive laws.

Process mechanism and scale

Pavement, drains and channelization can change runoff timing, sediment supply and channel boundary conditions. Draw a control volume showing forcing, material transfers and storage. Name the scale and timescale of the proposed process; landform shape alone does not reveal either.

Observable evidence

Before-after hydrographs, culvert maps, suspended sediment and repeat channel surveys constrain impacts. Record model grid and time step, monitoring period and independently observed outcomes. In the diagram of baseline and impact sites, DEM change maps and documented parameter ranges, label measured features separately from inferred histories. Provide survey dates, scale and a detection threshold.

Alternatives and field tests

A downstream channel incision trend may have begun before paving through an earlier base-level change. Compare paired catchments and record their unequal soils, slopes and rainfall exposures. Consider a visually attractive fitted model that fails out-of-sample. Identify one independently measurable outcome that would distinguish the two interpretations, and one negative result that would undermine the leading explanation.

Worked quantitative example — Urban runoff screening

Given. Runoff coefficient .60, rainfall intensity .030 m/hr, area 2 km². Calculation: 0.60 × 0.030 × 2000000 / 3600. Result: 10 m³/s. Work through the units before using the output to discuss Earth-surface history. Changing rainfall duration may invalidate the steady intensity assumption. This is a classroom rational-method screening, not a design hydrograph. A correct result from an incorrect conceptual model is not evidence that the process operated. State which measured variable would need independent verification. If the answer is a rate, indicate the averaging interval; if it is a geometric ratio, identify the scale and measurement direction; if it is a budget, give the boundary and the sign convention.

Research procedure and critical interpretation

Document independent validation; independently check reproducible inputs. Assess sensitivity to controls and counterfactuals and verify communication of uncertainty. Record each raw observation, units, uncertainty and assumptions before model fitting.

Correlation with urban fraction alone cannot isolate causal mechanisms. Use controls and event-based monitoring to attribute change cautiously. A high-resolution shape map without ages and a dated deposit without spatial coverage offer different constraints. Explain how they could be combined for urbanization and river response without overclaiming a process history. Identify one further test.

Problems with answer guidance

Problem 1. Recalculate the worked case after changing an input by 20% and show units. Problem 2. Contrast two possible landform origins using a specific field test. Answer guidance: correct units, explicit physical assumptions and a genuinely distinguishing observation are essential.

Process mechanism and scale

Excavation, topsoil exposure, reservoirs and farming affect source strength and pathways by which sediment reaches channels. Draw a control volume showing forcing, material transfers and storage. Name the scale and timescale of the proposed process; landform shape alone does not reveal either.

Observable evidence

Traceable mineral fingerprints, gully maps, deposition cores and reservoir bathymetry constrain redistribution. Record model grid and time step, monitoring period and independently observed outcomes. In the diagram of baseline and impact sites, DEM change maps and documented parameter ranges, label measured features separately from inferred histories. Provide survey dates, scale and a detection threshold.

Alternatives and field tests

A high sediment yield may reflect natural storm intensity rather than current land use alone. Build a source-to-sink conceptual network and estimate delivery ratios with sensitivity tests. Consider a visually attractive fitted model that fails out-of-sample. Identify one independently measurable outcome that would distinguish the two interpretations, and one negative result that would undermine the leading explanation.

Worked quantitative example — Sediment delivery ratio

Given. 2,700 tonnes reached the outlet out of 9,000 tonnes of estimated hillslope erosion. Calculation: 2700 / 9000. Result: 0.3 ratio. Work through the units before using the output to discuss Earth-surface history. Storage and remobilization cause delivery ratios to change with timescale. Input and output masses must cover the same date interval. A correct result from an incorrect conceptual model is not evidence that the process operated. State which measured variable would need independent verification. If the answer is a rate, indicate the averaging interval; if it is a geometric ratio, identify the scale and measurement direction; if it is a budget, give the boundary and the sign convention.

Research procedure and critical interpretation

Document independent validation; independently check reproducible inputs. Assess sensitivity to controls and counterfactuals and verify communication of uncertainty. Record each raw observation, units, uncertainty and assumptions before model fitting.

Unmeasured fine sediment bypass and legacy storage distort simple budgets. State which part of the sediment cascade the measurements actually cover. A high-resolution shape map without ages and a dated deposit without spatial coverage offer different constraints. Explain how they could be combined for mining, agriculture and sediment connectivity without overclaiming a process history. Identify one further test.

Problems with answer guidance

Problem 1. Recalculate the worked case after changing an input by 20% and show units. Problem 2. Contrast two possible landform origins using a specific field test. Answer guidance: correct units, explicit physical assumptions and a genuinely distinguishing observation are essential.

Process mechanism and scale

Evolution models combine conservation laws with constitutive transport rules; more than one parameter combination can fit a present-day profile. Draw a control volume showing forcing, material transfers and storage. Name the scale and timescale of the proposed process; landform shape alone does not reveal either.

Observable evidence

Observed topography and sediment flux may constrain coefficients but not uniquely recover histories. Record model grid and time step, monitoring period and independently observed outcomes. In the diagram of baseline and impact sites, DEM change maps and documented parameter ranges, label measured features separately from inferred histories. Provide survey dates, scale and a detection threshold.

Alternatives and field tests

A good visual model fit can be obtained with a wrong forcing history and compensating parameters. Fit an ideal hillslope profile with two alternative coefficients, then test held-out observations. Consider a visually attractive fitted model that fails out-of-sample. Identify one independently measurable outcome that would distinguish the two interpretations, and one negative result that would undermine the leading explanation.

Worked quantitative example — Diffusive smoothing length

Given. A conceptual coefficient D=.020 m²/yr operating for 500 years. Calculation: sqrt(.020 × 500). Result: 3.16228 m. Work through the units before using the output to discuss Earth-surface history. The proportionality constant depends on exact diffusion geometry and initial conditions. This is a characteristic scale, not a predicted ridge relocation. A correct result from an incorrect conceptual model is not evidence that the process operated. State which measured variable would need independent verification. If the answer is a rate, indicate the averaging interval; if it is a geometric ratio, identify the scale and measurement direction; if it is a budget, give the boundary and the sign convention.

Research procedure and critical interpretation

Document independent validation; independently check reproducible inputs. Assess sensitivity to controls and counterfactuals and verify communication of uncertainty. Record each raw observation, units, uncertainty and assumptions before model fitting.

Grid discretization, time step and fixed-boundary assumptions can dominate apparent precision. Report parameter bounds and predictive checks rather than one best-looking surface. A high-resolution shape map without ages and a dated deposit without spatial coverage offer different constraints. Explain how they could be combined for landscape models and parameter identifiability without overclaiming a process history. Identify one further test.

Problems with answer guidance

Problem 1. Recalculate the worked case after changing an input by 20% and show units. Problem 2. Contrast two possible landform origins using a specific field test. Answer guidance: correct units, explicit physical assumptions and a genuinely distinguishing observation are essential.

Process mechanism and scale

Satellite images, lidar and repeat photogrammetry can measure landform change after geometric coregistration and uncertainty screening. Draw a control volume showing forcing, material transfers and storage. Name the scale and timescale of the proposed process; landform shape alone does not reveal either.

Observable evidence

Stable rock control, independent ground points and scene metadata assess alignment and vertical accuracy. Record model grid and time step, monitoring period and independently observed outcomes. In the diagram of baseline and impact sites, DEM change maps and documented parameter ranges, label measured features separately from inferred histories. Provide survey dates, scale and a detection threshold.

Alternatives and field tests

A false deposition patch can result from vegetation-height change or DEM co-registration bias. Calculate minimum detectable change from combined errors before integrating volume differences. Consider a visually attractive fitted model that fails out-of-sample. Identify one independently measurable outcome that would distinguish the two interpretations, and one negative result that would undermine the leading explanation.

Worked quantitative example — Detectable vertical change

Given. Independent DEM standard deviations .20 and .25 m, multiplied by approximate 1.96 standard errors. Calculation: 1.96 × sqrt(.20² + .25²). Result: 0.62751 m. Work through the units before using the output to discuss Earth-surface history. Systematic alignment errors require additional analysis. A nominal 95% normal threshold only applies under assumed error conditions. A correct result from an incorrect conceptual model is not evidence that the process operated. State which measured variable would need independent verification. If the answer is a rate, indicate the averaging interval; if it is a geometric ratio, identify the scale and measurement direction; if it is a budget, give the boundary and the sign convention.

Research procedure and critical interpretation

Document independent validation; independently check reproducible inputs. Assess sensitivity to controls and counterfactuals and verify communication of uncertainty. Record each raw observation, units, uncertainty and assumptions before model fitting.

Clouds, shadow, water surfaces and vegetation cause spatially patterned missing data. Mask invalid areas and do not interpret every colored difference pixel as process. A high-resolution shape map without ages and a dated deposit without spatial coverage offer different constraints. Explain how they could be combined for remote sensing, change detection and uncertainty without overclaiming a process history. Identify one further test.

Problems with answer guidance

Problem 1. Recalculate the worked case after changing an input by 20% and show units. Problem 2. Contrast two possible landform origins using a specific field test. Answer guidance: correct units, explicit physical assumptions and a genuinely distinguishing observation are essential.

Process mechanism and scale

Restoration goals must specify measurable outcomes, counterfactual comparisons, sediment effects and adequate observation duration. Draw a control volume showing forcing, material transfers and storage. Name the scale and timescale of the proposed process; landform shape alone does not reveal either.

Observable evidence

Repeat cross sections, habitat indicators, channel stability and stakeholder records provide different success metrics. Record model grid and time step, monitoring period and independently observed outcomes. In the diagram of baseline and impact sites, DEM change maps and documented parameter ranges, label measured features separately from inferred histories. Provide survey dates, scale and a detection threshold.

Alternatives and field tests

Vegetated banks can appear stable temporarily while channel energy shifts erosion downstream. Design before-after-control-impact sampling, preregister evaluation thresholds and publish reproducible data transformations. Consider a visually attractive fitted model that fails out-of-sample. Identify one independently measurable outcome that would distinguish the two interpretations, and one negative result that would undermine the leading explanation.

Worked quantitative example — Restoration score

Given. Normalized progress from 20 baseline to 90 target when current score is 62. Calculation: (62 − 20) / (90 − 20). Result: 0.6 fraction. Work through the units before using the output to discuss Earth-surface history. Changing an arbitrary index definition changes apparent progress. An index cannot substitute for ecological outcome monitoring. A correct result from an incorrect conceptual model is not evidence that the process operated. State which measured variable would need independent verification. If the answer is a rate, indicate the averaging interval; if it is a geometric ratio, identify the scale and measurement direction; if it is a budget, give the boundary and the sign convention.

Research procedure and critical interpretation

Document independent validation; independently check reproducible inputs. Assess sensitivity to controls and counterfactuals and verify communication of uncertainty. Record each raw observation, units, uncertainty and assumptions before model fitting.

Extreme-event variability and changing upstream land use confound simple causal claims. Judge interventions against declared goals and uncertainty rather than visual attractiveness. A high-resolution shape map without ages and a dated deposit without spatial coverage offer different constraints. Explain how they could be combined for restoration, monitoring and reproducible interpretation without overclaiming a process history. Identify one further test.

Problems with answer guidance

Problem 1. Recalculate the worked case after changing an input by 20% and show units. Problem 2. Contrast two possible landform origins using a specific field test. Answer guidance: correct units, explicit physical assumptions and a genuinely distinguishing observation are essential.

A restored riparian corridor is surrounded by urban runoff, construction and an unmodified control stream. Compare cross-section survey error, sediment delivery ratios and before-after habitat observations. Fit an idealized model to one part of a time series; test it on held-out observations. Explain why monitoring failures or confounders prevent a simple success narrative.

Protocol and evidence allocation

Write a project proposal that names the study question, a defined sampling region, hypotheses H1 and H2 and three observations that could discriminate them. Build an input table with at least twelve synthetic measurements; record a unit, identifier, uncertainty class and data source for every row. Design one transect, one temporal repeat survey and one test using a different kind of evidence. State how the field team would avoid confirmation bias: preregister predicted outcomes, assign a second mapper to identify landforms independently and blind the initial data processing to the favored model. Any unmeasured quantity must be marked as an assumption. Base the arithmetic on the numerical cases in the unit, but do not present the invented data as actual measurements.

In the analysis, start with a mass or elevation balance rather than a purely visual explanation. If the balance does not close, label the residual and list processes that may lie outside the boundary, including sediment stored in a channel, transport at an unmonitored exit and error in estimated density or surface area. Separate interpretation of rate from reconstruction of age and chronology. An apparently consistent average rate can conceal long inactive periods punctuated by large storms, landslides or channel switching. Explore a parameter range wide enough to expose this sensitivity. Explain what observation would tell you whether the residual is an actual physical loss or a measurement artifact.

Reporting template and peer review

The final report should contain: a title that describes an inquiry rather than a conclusion; data source and acquisition notes; a map or section with scale; a table of raw observations; a short derivation with units; alternative interpretations; uncertainty and detection limits; and an honest conclusion identifying what the exercise cannot show. A peer reviewer should trace one plotted value back to an original input without assistance. Report uncertainty in both physical data and causal attribution. Describe how sampling locations could bias the result, and identify one independent validation datum that was not used when choosing the model. Treat a good fit as a starting point for further testing, not proof of a unique historical trajectory.

Assessment prompts and rubric

Interpretation. Provide two competing process histories consistent with the invented evidence, predict one differentiating field measurement, and identify which uncertainties could reverse the preferred interpretation. Quantitative reasoning. Recompute the selected unit case in SI units, vary two inputs independently, and interpret signs, magnitudes and nonlinearities. Evidence ethics. Identify all fabricated datasets and explain why they must never be described as real hazard, glacial or coastal monitoring records. Review standard. Full credit requires correct units, reproducible methods, a mass or geometry balance, explicit chronology limits and an alternative interpretation; numerical correctness alone is insufficient.

Return to the guiding question: How can competing explanations be tested using measurements and models? Make a process diagram contrasting an observable pattern with a process that could produce it. Now add a second process capable of leaving a similar pattern. Underline measurements that would distinguish the two. Ask whether the lesson's calculations predict the type of change measured by those instruments or only provide a scale for discussing it. The emphasis throughout geomorphology is not on naming spectacular landforms but on reconstructing defensible histories, with uncertainties, from repeatable measurements and physical consistency.