diff --git a/Dockerfile b/Dockerfile
index 9cb4d02..1bd7e40 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,5 +1,5 @@
 # Stage 0 - Create from julia image and install OS packages
-FROM julia:1.11.6 as stage0
+FROM julia:1.12.5 as stage0
 RUN apt update && apt -y install bzip2 build-essential libxml2
 
 # STAGE 1 - Python and python packages for S3 functionality
@@ -15,8 +15,8 @@ ENV PYTHON="/usr/bin/python3"
 COPY deps.jl /app/deps.jl
 ENV JULIA_CPU_TARGET="generic;sandybridge,-xsaveopt,clone_all;haswell,-rdrnd,base(1)"
 RUN julia /app/deps.jl \
-	&& find /usr/local/bin/julia_pkgs -type d -exec chmod 755 {} \; \
-	&& find /usr/local/bin/julia_pkgs -type f -exec chmod 644 {} \;
+        && find /usr/local/bin/julia_pkgs -type d -exec chmod 755 {} \; \
+        && find /usr/local/bin/julia_pkgs -type f -exec chmod 644 {} \;
 
 # Stage 3 - Copy SWOT script
 FROM stage2 as stage3
@@ -26,7 +26,7 @@ COPY ./sos_read /app/sos_read/
 # Stage 4 - Execute algorithm
 FROM stage3 as stage4
 LABEL version="1.0" \
-	description="Containerized SAD algorithm." \
-	"confluence.contact"="ntebaldi@umass.edu" \
-	"algorithm.contact"="kandread@umass.edu"
+        description="Containerized SAD algorithm." \
+        "confluence.contact"="ntebaldi@umass.edu" \
+        "algorithm.contact"="kandread@umass.edu"
 ENTRYPOINT ["/usr/local/julia/bin/julia", "/app/swot.jl"]
diff --git a/Sad.jl b/Sad.jl
index a1f580e..dfd8bba 160000
--- a/Sad.jl
+++ b/Sad.jl
@@ -1 +1 @@
-Subproject commit a1f580e9e79054b06a9c5e564a7ef9c147405cf9
+Subproject commit dfd8bbaf5daa342d45e6c322139cac5bebf5b731
diff --git a/swot.jl b/swot.jl
index b544bbb..995a750 100644
--- a/swot.jl
+++ b/swot.jl
@@ -48,11 +48,11 @@ Load SWOT observations.
 """
 function read_swot_obs(ncfile::String, nids::Vector{Int})
     Dataset(ncfile) do ds
-        nodes = NCDatasets.group(ds, "node")
-        reaches = NCDatasets.group(ds, "reach")
-        S = permutedims(nodes["slope2"][:])
-        H = permutedims(nodes["wse"][:])
-        W = permutedims(nodes["width"][:])
+        nodes = ds.group["node"]
+        reaches = ds.group["reach"]
+        S = permutedims(nodes["slope2"][:, :])
+        H = permutedims(nodes["wse"][:, :])
+        W = permutedims(nodes["width"][:, :])
         dA = reaches["d_x_area"][:]
         dA = convert(Vector{Sad.FloatM}, dA)
         Hr = convert(Vector{Sad.FloatM}, reaches["wse"][:])
@@ -64,10 +64,8 @@ function read_swot_obs(ncfile::String, nids::Vector{Int})
         nid = nodes["node_id"][:]
         dmap = Dict(nid[k] => k for k=1:length(nid))
         i = [dmap[k] for k in nids]
-        time_str_var = reaches["time_str"].var
-        time_str_raw = permutedims(time_str_var[:])
-        time_str = [join(time_str_raw[i, :]) for i in 1:size(time_str_raw, 1)]
-
+        time = reaches["time"][:]
+        time_str = [string(t) for t in time]
 
         H[i, :], W[i, :], S[i, :], dA, Hr, Wr, Sr, time_str
     end
@@ -86,7 +84,7 @@ Retrieve information about river reach cross sections.
 """
 function river_info(id::Int, swordfile::String)
     Dataset(swordfile) do fd
-        g = NCDatasets.group(fd, "nodes")
+        g = fd.group["nodes"]
         i = findall(g["reach_id"][:] .== id)
         nid = g["node_id"][i]
         x = g["dist_out"][i]
@@ -116,9 +114,9 @@ function write_output(reachid, valid, outdir, A0, n, Qa, Qu, W, time_str)
     ridv = defVar(out, "reach_id", Int64, (), fillvalue = FILL)
     ridv[:] = reachid
     A0v = defVar(out, "A0", Float64, (), fillvalue = FILL)
-    A0v[:] = A0
+    A0v[:] = coalesce(A0, FILL)
     nv = defVar(out, "n", Float64, (), fillvalue = FILL)
-    nv[:] = n
+    nv[:]  = coalesce(n,  FILL)
     Qav = defVar(out, "Qa", Float64, ("nt",), fillvalue = FILL)
     Qav[:] = replace!(Qa, NaN=>FILL)
     Quv = defVar(out, "Q_u", Float64, ("nt",), fillvalue = FILL)
@@ -171,7 +169,7 @@ function main()
     else
         index = parsed_args["index"] + 1
     end
-    
+
     reachfile = parsed_args["reachfile"]
     bucketkey = parsed_args["bucketkey"]
     println("Index: $(index)")
@@ -187,31 +185,27 @@ function main()
     nids, x = river_info(reachid, swordfile)
     H, W, S, dA, Hr, Wr, Sr, time_str = read_swot_obs(swotfile, nids)
 
-    try
-        x, H, W, S = Sad.drop_unobserved(x, H, W, S)
-    catch e
-        if e isa MethodError
-            println("Error loading swot observation")
-            end
-        end
+    reach = Sad.preprocess(x, H, W, S)
+
     A0 = missing
     n = missing
-    Qa = Array{Missing}(missing, 1, size(W, 2))
-    Qu = Array{Missing}(missing, 1, size(W, 2))
+    Qa = Matrix{Sad.FloatM}(missing, 1, size(W, 2))
+    Qu = Matrix{Sad.FloatM}(missing, 1, size(W, 2))
     if all(ismissing, H) || all(ismissing, W) || all(ismissing, S)
         println("$(reachid): INVALID")
         write_output(reachid, 0, outdir, A0, n, Qa, Qu, W, time_str)
     else
-        Hmin = minimum(skipmissing(H[1, :]))
-        Qp, np, rp, zp = Sad.priors(sosfile, Hmin, reachid)
-        if ismissing(Qp)
+        p = Sad.priors(sosfile, reach.hmin, reachid)
+        if ismissing(p)
             println("$(reachid): INVALID, missing mean discharge")
             write_output(reachid, 0, outdir, A0, n, Qa, Qu, W, time_str)
         else
             try
-                nens = 100 # default ensemble size
-                nsamples = 1000 # default sampling size
-                Qa, Qu, A0, n = Sad.estimate(x, H, W, S, dA, Qp, np, rp, zp, nens, nsamples, Hr, Wr, Sr)
+                res = Sad.infer(p, reach, time_str=time_str)
+                A0  = Sad.compute_A0(reach, res.reach_ensemble)
+                n   = mean(res.reach_ensemble[1, :])
+                Qa[1, :]  = [isnan(q) ? missing : q for q in res.Q_post]
+                Qu[1, :] = [isnothing(res.A_post[t]) ? missing : std(res.A_post[t][1, :]) for t in 1:reach.nt]
                 println("$(reachid): VALID")
                 write_output(reachid, 1, outdir, A0, n, Qa, Qu, W, time_str)
             catch