-- erosion.lua -- This is the main file of terrainlib_lua. It registers the EvolutionModel object and some of the local function erode(model, time) -- Apply river erosion on the model -- Erosion model is based on the simplified version of the stream-power law Ey = K×A^m×S -- where Ey is the vertical erosion speed, A catchment area of the river, S slope along the river, m and K local constants. -- It is equivalent to considering a horizontal erosion wave travelling at Ex = K×A^m, and this latter approach allows much greather time steps so it is used here. -- For each point, instead of moving upstream and see what point the erosion wave would reach, we move downstream and see from which point the erosion wave would reach the given point, then we can set the elevation. local mmin, mmax = math.min, math.max local dem = model.dem local dirs = model.dirs local lakes = model.lakes local rivers = model.rivers local sea_level = model.params.sea_level local K = model.params.K local m = model.params.m local X, Y = dem.X, dem.Y local scalars = type(K) == "number" and type(m) == "number" local erosion_time if model.params.variable_erosion then erosion_time = {} else erosion_time = model.erosion_time or {} end if scalars then for i=1, X*Y do local etime = 1 / (K*rivers[i]^m) -- Inverse of erosion speed (Ex); time needed for the erosion wave to move through the river section. erosion_time[i] = etime lakes[i] = mmax(lakes[i], dem[i], sea_level) -- Use lake/sea surface if higher than ground level, because rivers can not erode below. end else for i=1, X*Y do local etime = 1 / (K[i]*rivers[i]^m[i]) erosion_time[i] = etime lakes[i] = mmax(lakes[i], dem[i], sea_level) end end for i=1, X*Y do local iw = i local remaining = time local new_elev while true do -- Explore downstream until we find the point 'iw' from which the erosion wave will reach 'i' local inext = iw local d = dirs[iw] -- Follow the river downstream (move 'iw') if d == 0 then -- If no flow direction, we reach the border of the map: set elevation to the latest node's elev and abort. new_elev = lakes[iw] break elseif d == 1 then inext = iw+X elseif d == 2 then inext = iw+1 elseif d == 3 then inext = iw-X elseif d == 4 then inext = iw-1 end local etime = erosion_time[iw] if remaining <= etime then -- We have found the node from which the erosion wave will take 'time' to arrive to 'i'. local c = remaining / etime new_elev = (1-c) * lakes[iw] + c * lakes[inext] -- Interpolate linearly between the two nodes break end remaining = remaining - etime -- If we still don't reach the target time, decrement time and move to next point. iw = inext end dem[i] = mmin(dem[i], new_elev) end end local function diffuse(model, time) -- Apply diffusion using finite differences methods -- Adapted for small radiuses local mmax = math.max local dem = model.dem local X, Y = dem.X, dem.Y local d = model.params.d -- 'd' is equal to 4 times the diffusion coefficient local dmax = d if type(d) == "table" then dmax = -math.huge for i=1, X*Y do dmax = mmax(dmax, d[i]) end end local diff = dmax * time -- diff should never exceed 1 per iteration. -- If needed, we will divide the process in enough iterations so that 'ddiff' is below 1. local niter = math.floor(diff) + 1 local ddiff = diff / niter local temp = {} for n=1, niter do local i = 1 for y=1, Y do local iN = (y==1) and 0 or -X local iS = (y==Y) and 0 or X for x=1, X do local iW = (x==1) and 0 or -1 local iE = (x==X) and 0 or 1 -- Laplacian Δdem × 1/4 temp[i] = (dem[i+iN]+dem[i+iE]+dem[i+iS]+dem[i+iW])*0.25 - dem[i] i = i + 1 end end for i=1, X*Y do dem[i] = dem[i] + temp[i]*ddiff end end end local modpath = "" if minetest then modpath = minetest.get_modpath(minetest.get_current_modname()) .. "terrainlib_lua/" end local rivermapper = dofile(modpath .. "rivermapper.lua") local gaussian = dofile(modpath .. "gaussian.lua") local function flow(model) model.dirs, model.lakes = rivermapper.flow_routing(model.dem, model.dirs, model.lakes, 'semirandom') model.rivers = rivermapper.accumulate(model.dirs, model.rivers) end local function uplift(model, time) -- Raises the terrain according to uplift rate (model.params.uplift) local dem = model.dem local X, Y = dem.X, dem.Y local uplift_rate = model.params.uplift if type(uplift_rate) == "number" then local uplift_total = uplift_rate * time for i=1, X*Y do dem[i] = dem[i] + uplift_total end else for i=1, X*Y do dem[i] = dem[i] + uplift_rate[i]*time end end end local function noise(model, time) -- Adds noise to the terrain according to noise depth (model.params.noise) local random = math.random local dem = model.dem local noise_depth = model.params.noise * 2 * time local X, Y = dem.X, dem.Y for i=1, X*Y do dem[i] = dem[i] + (random()-0.5) * noise_depth end end -- Isostasy -- This is the geological phenomenon that makes the lithosphere "float" over the underlying layers. -- One of the key implications is that when a very large mass is removed from the ground, the lithosphere reacts by moving upward. This compensation only occurs at large scale (as the lithosphere is not flexible enough for small scale adjustments) so the implementation is using a very large-window Gaussian blur of the elevation array. -- This implementation is quite simplistic, it does not do a mass balance of the lithosphere as this would introduce too many parameters. Instead, it defines a reference equilibrium elevation, and the ground will react toward this elevation (at the scale of the gaussian window). -- A change in reference isostasy during the run can also be used to simulate tectonic forcing, like making a new mountain range appear. local function define_isostasy(model, ref, link) ref = ref or model.dem if link then model.isostasy_ref = ref return end local X, Y = ref.X, ref.Y local ref2 = model.isostasy_ref or {X=X, Y=Y} model.isostasy_ref = ref2 for i=1, X*Y do ref2[i] = ref[i] end return ref2 end -- Apply isostasy local function isostasy(model) local dem = model.dem local X, Y = dem.X, dem.Y local temp = {X=X, Y=Y} local ref = model.isostasy_ref for i=1, X*Y do temp[i] = ref[i] - dem[i] -- Compute the difference between the ground level and the target level end -- Blur the difference map using Gaussian blur gaussian.gaussian_blur_approx(temp, model.params.compensation_radius, 4) for i=1, X*Y do dem[i] = dem[i] + temp[i] -- Apply the difference end end local evol_model_mt = { erode = erode, diffuse = diffuse, flow = flow, uplift = uplift, noise = noise, isostasy = isostasy, define_isostasy = define_isostasy, } evol_model_mt.__index = evol_model_mt local defaults = { K = 1, m = 0.5, d = 1, variable_erosion = false, sea_level = 0, uplift = 10, noise = 0.001, compensation_radius = 50, } local function EvolutionModel(params) params = params or {} local o = {params = params} for k, v in pairs(defaults) do if params[k] == nil then params[k] = v end end o.dem = params.dem return setmetatable(o, evol_model_mt) end return EvolutionModel