GIS & Remote Sensing Workflow for Rainfall Network Design

Goal

This seminar develops a scientifically grounded workflow for designing physiographically stratified rainfall networks in low mountain regions (e.g. the Burgwald, Germany).

Rather than optimising station locations directly, the course emphasises:

• spatial structure before optimisation,
• explicit separation of structure, process, and context,
• reproducible spatial analysis,
• and validation through diagnostic reasoning, not model perfection.

The workflow combines literature-driven design, GIS-based spatial structuring, and selective quantitative diagnostics.

Course Structure

• Total sessions: ~10 × 3 h
  
• Core workflow modules: Sessions 1–6
  
• Advanced / toolkit modules: Sessions 7–10 (project-specific)

Each session contains: - ≤ 1 h instructor input, - guided implementation, - structured discussion and reflexion.

Introduction: From Conceptual Design to Calculation

Rainfall network design is not a purely technical optimisation problem. It is a scientific design task that links:

• landscape structure,
• physical processes,
• and measurement constraints.

In the first phase of the course (Tasks First Design and Resources & Pitch), the following were developed:

• conceptual network layouts,
• deployment rationales (physiographic, information-gain, hydrological),
• resource plans and feasibility assessments,
• an initial literature base using benchmark observatories.

The technical modules that follow do not invent completely new questions. They operationalise, adapt and test these design ideas using spatial data and reproducible workflows.

Warning🔁 Adaptive redesign

In response to the evolving projects, the module plan was adapted rather than replaced.
The workflow was reordered to ensure that spatial structure and process assumptions are made explicit prior to interpolation or optimisation.

Module 0 – Concept & Evidence Scan

Status: completed in Tasks First Design and Resources & Pitch

Goal: Establish a literature-anchored design rationale before technical work.

Core elements

• Review of benchmark rain-gauge networks and observatories.
• Classification by dominant rationale:
  • physiographic stratification,
  • information-gain optimisation,
  • hydrological coupling.
• Conceptual Burgwald network:
  • ≤ 20 stations,
  • explicit station roles (Backbone, Infill, P–Q, Open–Wood, Event-Scout),
  • data and resource plan,
  • initial literature justification.

Tip🧩 Key idea

Defines what should be built and why. All following modules answer how.

Module 1 – Project Organisation & FAIR Data Retrieval

Goal: Build a reproducible project skeleton consistent with Module 0.

Technical scope

• R project setup (renv, here, targets)
• FAIR principles (Findable, Accessible, Interoperable, Reusable)
• Targeted datasets:
  • DEM / DGM → terrain, landforms, drainage
  • CORINE / ATKIS → structural & physiographic context
  • RADOLAN, DWD gauges → precipitation fields and point data
  • Sentinel-2 → structural layers (canopy/open), not full RS modelling

Tip🧩 Key idea

Data selection is not purely technical.
It is biased by process assumptions and directly constrains the interpretation of the project results.

Module 2 – Geodata Preprocessing as Process Modelling

Goal: Derive process-relevant base variables, not just cleaned inputs.

Technical scope

• Raster/vector handling (terra, sf, stars)
• Morphometry:
  • slope, aspect, curvature / convexity
  • landform classification (ridge / slope / valley)
• Hydrological pre-structuring:
  • flow direction, flow accumulation
  • watershed and sub-catchment delineation
• Functional land-cover simplification (forest/open, canopy density)

Tip🧩 Key idea

This step reframes preprocessing as process modelling at the first order.
Consequently, all scale decisions made here become part of the scientific argument and must be revisited explicitly.

Module 3 – Wind Fields & Luv–Lee Annotation

Goal: Provide atmospheric context, not full atmospheric modelling.

Technical scope

• Wind data (DWD stations, ERA5 / COSMO)
• Dominant wind directions (seasonal / aggregated)
• Aspect × wind → Luv/Lee annotation

Tip🧩 Key idea

This module introduces key conceptual constraints.
Variables such as Luv/Lee are treated as annotation layers, not as clustering variables.
They support interpretation and diagnostics across segments, but do not define the spatial segmentation itself.

Module 4 – Spatial Structuring & Stratification (Core Module)

Goal: Reduce spatial complexity before interpolation or optimisation.

Technical scope

• Tile-based structural pre-analysis
• Adaptive spatial units:
  • supercells (Nowosad ecosystem)
  • optional simple clustering (e.g. kmeans)
• Focus on:
  • spatial coherence,
  • reproducibility,
  • process plausibility.

Tip🧩 Key idea

This module marks a shift in emphasis.
Rather than pursuing exhaustive cluster-validity optimisation (e.g. silhouette scores), the focus is placed on interpretability, stability, and process plausibility.
As a consequence, the workflow requires a renewed and more critical engagement with the literature to reassess methodological assumptions and design choices.

Module 5 – Scientific Recursion (Re-Entry Loop)

Goal: Enforce an explicit scientific recursion:
projects must now be re-aligned, refined, and justified based on accumulated insights.

This module introduces a deliberate loop in the workflow.
Instead of progressing forward, all groups are required to re-enter their own project logic.

The intention is to make explicit what ideally should have happened iteratively from the start: the continuous adaptation of methods to concepts, not the other way around.

Scientific Recursion

Over the past weeks, students have: - implemented substantial technical workflows, - explored multiple methodological templates, - generated concrete spatial structures, strata, and candidate designs.

What has often not happened sufficiently is the recursive alignment between: - conceptual intent, - methodological choices, - and scientific justification.

This module therefore enforces that alignment retroactively and explicitly.

Warning🔁 Scientific Recursion

Each group must now refine its project by:

• revisiting the original concept and design intent,
• adapting the workflow where necessary (simplification, re-weighting, omission),
• justifying these adaptations with literature and own argumentation,
• explaining why the resulting workflow is appropriate for their specific project.

The provided workflow and examples are to be treated as exemplary scaffolding,
not as a blueprint to be reproduced.

What this module demands

• A clear statement of what was changed and why.
• Explicit references that support or contradict earlier assumptions.
• Recognition that technical completeness does not imply scientific adequacy.
• Acceptance that refinement often means removing methods, not adding more.

Tip🧩 Key idea

A scientifically grounded, internally consistent project design that: - reflects informed methodological choices, - acknowledges limitations and alternatives, - and demonstrates ownership over the workflow.

This recursion marks the transition from method execution to scientific authorship.

Module 6 – Interpolation, Mapping & Diagnostic Validation

Goal: Use interpolation and mapping as diagnostic tools to evaluate spatial structuring and process assumptions — not to construct a “true” precipitation field.

Conceptual framing

Interpolation is deliberately positioned after spatial structuring and stratification. Its role is not optimisation, but testing:

• Do the chosen spatial units and strata explain systematic deviations?
• Where do interpolated fields perform consistently well or poorly — and why?

Technical scope

Interpolation inputs

• RADOLAN precipitation products, or
• simple gauge-based approaches (IDW, basic Kriging).

Evaluation strategy

• Extract precipitation time series at:
  • existing gauges,
  • planned or hypothetical station locations.
• Analyse residuals and differences by:
  • spatial strata and segments,
  • landforms (ridge / slope / valley),
  • annotation layers (e.g. Luv/Lee),
  • catchments and sub-catchments.

Mapping & visual diagnostics

• Cartographic tools: tmap, ggplot2, mapview
• Core overlays:
  • segmentation and physiographic strata,
  • landforms and drainage networks,
  • interpolated precipitation fields,
  • spatial residual patterns.

Maps are treated as analytical artefacts, not presentation products: they are used to identify structural consistency, mismatches, and process-related patterns.

Tip🧩 Key idea

Interpolation functions as a stress test. The question is not whether an interpolated field looks smooth or accurate, but whether the chosen spatial structuring helps explain where and why interpolation fails.

Suggested Timeline

Session	Theme	Module(s)	Format
1	Concept & Evidence	Module 0	Discussion + design review
2	Project & Data Infrastructure	Module 1	Setup + guided workflow
3	Preprocessing & Landforms	Module 2	Analysis + maps
4	Spatial Structuring	Module 4	Segmentation workshop
5	Wind Context & Interpretation	Module 3	Annotation + discussion
6	Scientific Recursion	Module 5	Literature-based redesign
7	Diagnostic Validation	Module 6	Interpolation + residuals
8–10	Workshops & Refinement	Advanced modules	Project-specific deep dives

WarningOriginal Schedule vs. Adaptive Workflow

Summary Table

Original Module	Original Focus	Adaptive Role Now	Status
Module 0	Implicit literature review	Explicit concept & evidence scan	Central
Module 1	Project setup & data	FAIR, reproducible implementation base	Active
Module 2	Preprocessing	First-order process modelling	Active
Module 3	Wind fields	Contextual annotation (Luv/Lee etc.)	Active
Module 4	Structuring & clustering	Core spatial backbone (segmentation & strata)	Active
Cross-cutting	Iterative scientific reflection	Explicit Scientific Recursion (Module 5)	Central
Interpolation	Field generation	Diagnostic / validation instrument (Module 6)	Limited
Visualization	Mapping	Analytical diagnostics, not presentation	Integrated

Key Take-Home Messages

• Structure precedes interpolation.
• Interpolation serves validation, not truth.
• Spatial decisions imply scale decisions.
• Adaptive workflows require renewed literature engagement.
• Scientific recursion is part of the method, not an afterthought.