Hydroloom Overview"
In hydroloom: Utilities to Weave Hydrologic Fabrics

local <- (Sys.getenv("BUILD_VIGNETTES") == "TRUE")

knitr::opts_chunk$set(
  collapse = TRUE,
  warning = FALSE,
  comment = "#>",
  fig.width=4,
  fig.height=4,
  fig.align = "center",
  eval=local
)

library(sf)

oldoption <- options(scipen = 9999)

hydroloom

Hydroloom is designed to provide general hydrologic network functionality for any hydrographic or hydrologic data. This is accomplished with
1) the hy S3 class,
2) a collection of utility functions,
3) functions to work with a hydrologic network topology as a graph,
4) functions to create and add useful network attributes,
5) and functions to index data to a network of flow network lines and waterbody polygons.

This introduction covers the hy S3 class and the core flow network topology concepts necessary to use hydroloom effectively.

For the latest development and to open issues, please visit the package github repository.

`hy` S3 class

Hydroloom implements an hy S3 class designed to simplify integration of hydroloom with existing data and workflows. hy() converts a data.frame to an hy data.frame with attributes compatible with hydroloom functions. hy_reverse() converts a hy data.frame back to its original attribute names. You can teach hydroloom how to map your attributes to hydroloom_name_definitions() with the hydroloom_names() function.8

Most hydroloom functions will work with either a hy object or a data.frame containing names registered with hydroloom_names(). Any attributes added to the data.frame will contain names from hydroloom and must be renamed in the calling environment.

Internally, the hy S3 class has an attribute orig_names as shown below. The orig_names attribute is used to convert original attribute names back to their original values. Using the hydroloom names and the hy S3 object are note required but adopting hydroloom_names_definitions() may be helpful for people aiming for consistent, simple, and accurate attribute names.

library(hydroloom)

hy_net <- sf::read_sf(system.file("extdata/new_hope.gpkg", package = "hydroloom")) |>
  dplyr::select(COMID, REACHCODE, FromNode, ToNode, Hydroseq, TerminalFl, Divergence)

hy(hy_net[1:3,])

attr(hy(hy_net), "orig_names")

Network Representation

Representing Dendritic Network Topology

A network of flowlines can be represented as an edge-to-edge (e.g. edge list) or edge-node topology. An edge list only expresses the connectivity between edges (flowlines in the context of rivers), requiring nodes (confluences in the context of rivers) to be inferred.

print.data.frame(data.frame(id = c(1, 2, 3), 
                            toid = c(3, 3, NA),
                            fromnode = c("N1", "N2", "N3"),
                            tonode = c("N3", "N3", "N4")), 
                 row.names = FALSE)

x <- c(1, 5, 3, 3)
y <- c(5, 5, 3, 1)

oldpar <- par(mar = c(0, 0, 0, 0))
plot(x, y, col = NA)
arrows(x[1] + 0.1, y[1] - 0.1, x[3] - 0.1, y [3] + 0.1, 0.1)
arrows(x[2] - 0.1, y[2] -0.1, x[3] + 0.1, y [3] + 0.1, 0.1)
arrows(x[3], y[3] - 0.1, x[4], y [4] + 0.1, 0.1)
text(c(2, 4, 3.15), c(4.2, 4.2, 2), c("1", "2", "3"))

par(mar = c(0, 0, 0, 0))
plot(x, y)
arrows(x[1] + 0.1, y[1] - 0.1, x[3] - 0.1, y [3] + 0.1, 0.1)
arrows(x[2] - 0.1, y[2] -0.1, x[3] + 0.1, y [3] + 0.1, 0.1)
arrows(x[3], y[3] - 0.1, x[4], y [4] + 0.1, 0.1)
text(c(2, 4, 3.1), c(4.2, 4.2, 2), c("1", "2", "3"))
text(c(1, 5, 3, 3.25), c(4.8, 4.8, 3.4, 1), c("N1", "N2", "N3", "N4"))
par(oldpar)

The "toid" of a terminal flowline can be either NA or, by convention, 0 or "". Using 0 or an empty string is preferred within hydroloom but both are handled in most cases. Further, as long as 0 is not in the set of IDs, there is little practical difference.

In hydroloom, edge-to-edge topology is referred to with "id and toid" attributes.

Representing Dendritic Network Topology

As discussed in the vignette("non-dendritic") vignette, a hydrologic flow network can be represented as an edge to edge (e.g. edge list) topology or an edge-node topology. In the case of dendritic networks, an edge list can be stored as a single "toid" attribute on each feature and nodes are redundant as there would be one and only one node for each feature. In non-dendritic networks, an edge list can include multiple "toid" attributes for each feature, necessitating a one to many relationship that can be difficult to interpret. Nevertheless, the U.S. National Hydrography Dataset uses an edge-list format in its "flow table" and the format is capable of storing non-dendritic feature topology.

Using a node topology to store a flow network, each feature flows from one and only one node and flows to one and only one node. This one to one relationship between features and their from and to nodes means that the topology fits in a table with one row per feature as is common practice in spatial feature data.

For this reason, the NHDPlus data model converts the NHD "flow table" into node topology in its representation of non dendritic topology. The downside of this approach is that it requires creation of a node identifier. These node identifiers are a table deduplication device that enables a one to many relationship (the flow table) to be represented as two one to one relationships. Given this, in hydrologic flow networks, node identifiers can be created based on an edge list and discarded when no longer needed.

wzxhzdk:4

wzxhzdk:5

In this example of an edge list topology and a node topology for the same system, feature '1' flows to two edges but only one node. We can represent this in tabular form with a duplicated row for the divergence downstream of '1' or with the addition of node identifiers as shown in the following tables.

| id | fromnode | tonode | |----|----------|--------| | 1 | N1 | N2 | | 2 | N2 | N3 | | 3 | N3 | N4 | | 4 | N2 | N4 | | 5 | N4 | N5 |

| id | toid | |----|------| | 1 | 2 | | 1 | 4 | | 2 | 3 | | 3 | 5 | | 4 | 5 | | 5 | 0 |

Network Graph Representation

The make_index_ids() hydroloom function creates an adjacency matrix representation of a flow network as well as some convenient content that are useful when traversing the graph. This adjacency matrix is used heavily in hydroloom functions and may be useful to people who want to write their own graph traversal algorithms.

A companion to make_index_ids(), make_fromids() reverses the graph such that the adjacency matrix is directed opposite the direction of the provided graph. This is used for upstream flow network traversals.

In the example below we'll add a dendritic toid and explore the make_index_ids() output.

y <- add_toids(hy_net, return_dendritic = TRUE)

ind_id <- make_index_ids(y)

names(ind_id)

dim(ind_id$to)

max(lengths(ind_id$lengths))

names(ind_id$to_list)

sapply(ind_id, class)

Now we'll look at the same thing but for a non dendritic set of toids. Notice that the to element of ind_id now has three rows. This indicates that one or more of the connections in the matrix has three downstream neighbors. The lengths element indicates how many non NA values are in each column of the matrix in the to element.

y <- add_toids(st_drop_geometry(hy_net), return_dendritic = FALSE)

ind_id <- make_index_ids(y)

names(ind_id)
dim(ind_id$to)

max(ind_id$lengths)

sum(ind_id$lengths == 2)
sum(ind_id$lengths == 3)

names(ind_id$to_list)

sapply(ind_id, class)