Viewshed and Hydrology

A viewshed is the set of all cells visible from one or more observer points, computed by tracing a straight line of sight from the observer to every other cell and checking whether the intervening terrain rises high enough to block it. Cells the land hides are 'not visible'; the rest form the viewshed. It is a global operation in Tomlin's sense - each output cell depends on the whole run of terrain between it and the observer, not just its neighbours.

The inputs are more than the DEM. Observer height matters (eye level versus a 60 m tower changes everything), as does target height, a radius (you rarely care beyond a few kilometres), and whether you account for Earth curvature and refraction at distance. Run it from many points and you get cumulative visibility - how often each cell is seen - which is how planners map visual exposure of a ridge or the visual catchment of a landmark.

The design uses are direct. A reverse viewshed - what can see a given place - maps how exposed a proposed tower or quarry will be from the surrounding town, the basis of visual-impact assessment. A forward viewshed from a proposed apartment tells you which floors actually earn the 'sea view' in the brochure. Between them they turn an argument about visual amenity into a mapped, contestable claim.

Viewsheds are seductive and easily wrong. Most DEMs are bare-earth (a DTM): they contain no buildings and no trees, so a viewshed over one assumes an empty landscape - it will happily 'see' straight through a forest or a built-up block. If you need real-world visibility in a city you need a surface model (DSM) that includes structures and canopy, which is rarer and usually lidar-derived.

Resolution bites again: a 30 m DEM misses the very ridgeline or embankment that would block a view. And the algorithm's assumptions (curvature, refraction, observer/target heights) change the answer materially. Treat a viewshed as a rigorous hypothesis to verify on site, not a courtroom certainty - and always state the DEM type and heights you used.

Terrain hydrology is a fixed pipeline, and it starts with a clean-up. Real DEMs contain sinks - single cells or dips with no lower neighbour, mostly noise - and water would get trapped in them. So the first step is always fill sinks, raising those pits until water can escape.

Then comes flow direction: for every cell, the algorithm decides which way water leaves it. The classic D8 method sends all of a cell's water to whichever of its eight neighbours lies steepest downhill - one direction per cell. (More sophisticated methods split flow across several neighbours, better for gentle terrain.) The result is a grid of little arrows, the skeleton on which everything downstream is built.

Once every cell knows where its water goes, flow accumulation counts, for each cell, how many upstream cells ultimately drain through it. Cells with tiny counts are hillslopes; cells with large counts are where water concentrates - and if you threshold the accumulation grid (say, keep cells with more than 500 upstream cells) you extract the stream network straight from the terrain. That threshold is a judgement, not a law: lower it and you grow a denser network of minor gullies, raise it and you keep only the major channels.

Finally, a watershed (catchment / basin) is every cell that drains to a chosen outlet or 'pour point'. Drop a pour point at a culvert, a drain inlet or a river gauge and the GIS traces the whole area contributing runoff to it. One caution that trips up beginners: the pour point must sit exactly on a high-accumulation stream cell, so most tools offer a 'snap pour point' step that nudges it onto the channel first. For a designer this is gold: it tells you how much land sheds water onto your site, where a retention pond should sit, and which catchment a proposed development actually lies in.

India's urban flooding is, at root, a hydrology-ignored-at-the-plot-scale problem: buildings placed in natural drainage lines, tanks and nalas built over, catchments sealed with concrete so runoff spikes. A quick DEM-based flow-accumulation run - even on free 30 m CartoDEM - reveals the natural drainage lines crossing a site before you place a building on one. It will not size a storm drain (that needs rainfall, soil and finer terrain), but it will stop the gross mistakes.

Use it early: overlay the derived stream network and the site's catchment on your plot at concept stage, and let water veto the obviously bad footprints. Then hand the numbers to a hydraulic engineer for the design storm.