Point Pattern

Programita, a software for spatial point pattern analysis in Ecology

Programita is a comprehensive software package for conducting spatial point pattern analysis in ecology. I tailored Programita to accommodate the needs of “real world” applications in ecology and developed the different modules in response to my own research questions and to requests of colleagues and students who have approached me with their specific research problems in mind. Originally, I developed Programita for my graduate courses “Patrones espaciales en ecología: modelos y análisis”, at the Escuela para Graduados, Facultad de Agronomia, University Buenos Aires, Argentina.

A detailed description of the methods implemented in Programita can be found in my Handbook of Spatial Point Pattern Analysis in Ecology .   

The updated Programita software has been downloaded since its launch in January 2014 (up to July 2021) by more than 1600 scientists from all over the world. Google scholar lists more than 600 articles that use the Programita software.

Recent publications using Programita include:


Requests for Programita and a user manual (with extensive examples):
www.programita.org

Data types included in Programita

Seventeen years after its launch in the 2004 Oikos Mini review (Wiegand and Moloney 2004), Programita has considerably grown and now contains a variety of statistical methods for most point pattern data types that are relevant in ecological applications, including:

  • univariate patterns
    one type of points, the most commonly analyzed data type. Null models and point process models include homogeneous and heterogeneous Poisson processes, Thomas cluster processes with one or two critical scales of clustering, and simple soft and hard-core processes.
    Wiegand et al. (2009) , Wiegand and Moloney (2014: section 4.1)
  • bivariate patterns
    two types of points such as two species of trees. Programita provides several options to test the independence null hypothesis by using the toroidal shift null model, Thomas cluster point processes, or pattern reconstruction. Additionally, you can use bivariate homogeneous and heterogeneous Poisson processes, bivariate Thomas cluster processes with one or two critical scales of clustering, and simple bivariate soft and hard-core processes.
    Getzin and Wiegand (2014), Wiegand and Moloney (2014: section 4.2)
  • qualitatively marked patterns
    one type of points that carries a qualitative mark such as surviving vs. dead. Null models include random labeling, trivariate random labeling, random labeling with a covariate, and random labeling for communities.
    Jacquemyn et al. (2009), Wiegand and Moloney (2014: section 3.16)
  • quantitatively marked patterns
    in the simplest case (i) one type of point that carries a quantitative marks such as size, but I implemented additional data types relevant for my work including patterns with (ii) one qualitative and one quantitative mark, and (iii) bivariate patterns with one quantitative mark. The random marking null models are adapted to each data structure.
    Wiegand and Moloney (2014: section 4.4)
  • multivariate patterns
    several types of points such as trees of different species in a forest community. You can conduct with Programita multivariate analyses at the community level using (previously generated) null communities, or analyses on the species level using for the focal species the toroidal shift null model, homogeneous and heterogeneous Poisson processes, or previously generated null model data.
    Wiegand et al. (2007), Wiegand and Moloney (2014: section 4.3)
  • multivariate patterns and pairwise dissimilarities
    uses additionally a matrix of pairwise dissimilarities such as functional or functional dissimilarities, but also pairwise differences in a single trait. It allows e.g., to analyze phylogenetic or functional beta diversity. Additionally to the null models for multivariate patterns you can randomize the dissimilarity matrix in different ways following Hardy (2009).
    Wang et al. (2015, 2016), Wiegand and Moloney (2014: sections 3.1.5.3 and 3.1.7.6), Wiegand et al. (2017)
  • objects with finite size and real shape
    designed for cases where the point approximation does not hold.
    Wiegand et al. (2006), Wiegand and Moloney (2014: section 3.1.8)

Additional features of Programita

Programita allows you to

  • conduct Monte Carlos simulations of null models or point process models that proved to be important for real world ecological applications
  • fit cluster point processes to the data and generate stochastic realizations of the fitted point processes
  • determine (global) simulation envelopes of the null model and point process models and conduct goodness-of-fit (GoF) tests ( Wiegand et al. 2016)
  • use irregularly shaped observation windows
  • combine the results from several replicate plots into a mean, weighted summary statistic

Programita offers for each of these data types the most appropriate summary statistics:

  • uni- and bivariate patterns
    pair correlation function, L-function, the K2 function, the distribution functions of the distances to the kth nearest neighbor, the spherical contact distribution (only univariate), the mean distance to the kth neighbor, inhomogeneous g- and L-functions (only univariate implemented yet)
  • qualitatively marked patterns
    mark connection functions and various test functions based on pair correlation or K-functions ( Jacquemyn et al. 2009)
  • quantitatively marked patterns
    various normalized and non-normalized mark correlation functions, including the mark and the r-mark correlation functions, the mark variogram, Morans I and Schlathers correlation functions, and the density correlation function ( Fedriani et al. 2015)
  • multivariate patterns
    spatially explicit Simpson index, individual species-area relationship
  • multivariate patterns and pairwise dissimilarities
    phylogenetic Simpson index, phylogenetic mark correlation function, rISAR function ( Shen et al. 2013, Wang et al. 2015, 2016, Wiegand et al. 2017)
  • objects with finite size and real shape
    uni- and bivariate O-ring statistic and L-function ( Wiegand et al. 2006)


see also Software for pattern reconstruction