Landscape-scale analysis of changes in forest land use/land cover across Puerto Rico, 1990-2020.

Abstract

Land use/land cover (LULC) change is a global phenomenon that greatly impacts the extent of forest cover, particularly in tropical countries which face increasing pressures related to agriculture, production of timber and non-timber materials, urbanization, and other anthropogenic drivers. There is a need for research focused on post-agricultural societies in tropical urbanizing landscapes to assess LULC change where forest regeneration is occurring due to both passive and active management strategies. The Caribbean archipelago of Puerto Rico is an ideal place to explore such change. It has experienced drastic changes in forest cover over the course of the past century, in response to changing socio-economic and conservation trends. This study analyzes decadal changes in forest land use (as a proxy for LULC) in Puerto Rico from 1990 to 2020 at the island scale to accomplish the following objectives: 1) explore temporal trends in forest loss and gain; 2) calculate transition probabilities between land use categories; 3) analyze the spatial distribution of forest loss and gain; and 4) examine spatial autocorrelation of changes in forest loss/gain at the block group scale in relation to select social variables. The analyses were conducted using land use data derived from 30 m x 30 m Landsat imagery and obtained from the Landscape Change Monitoring System (LCMS), together with social data representing population density and forestry/agricultural occupation density at the census block scale. Trends in forest loss/gain, transition probabilities, spatial clustering, and geographically weighted regression analyses (GWR) were performed using ArcGIS Pro software and spatial analysis tools.<br /> <br /> The results showed that forest land use in Puerto Rico during the study period remained relatively constant, representing approximately 62%-64% of total area. This result was higher than the various estimates provided by previous studies that analyzed data within similar time frames and calculated a low of 32% forest area (1990) and a high of 57% (2003). This difference between the estimates is due to differences in classification methods and forest definitions between LCMS and other approaches. Results show that forest represents the largest fraction among the land use classes and represents the lowest percentage of overall (30-year) loss of 1.4%. Prominent forest losses were observed in coastal peri-urban areas adjacent to major cities and interior mountainous areas in eastern central Puerto Rico. Among the decadal intervals evaluated, the period 1990-2000 was marked by a dominant forest loss equivalent to about 520.9 km², while the dominant period of forest gain was 2000-2010 and equivalent to 443.7 km². The results also show that Forest land use is the most resilient to change and therefore the most stable among the land use classes. This stability is evidenced by a probability of 86% (30-year) of remaining forest. The relatively small areas of Forest that did change tended to convert to Developed and Rangeland or Pasture. Additionally, areas classified as Developed, Rangeland or Pasture, and Agriculture also converted to Forest. The observed expansion of Forest in urban areas is likely due attributed to increases in tree canopy cover rather than actual land conversion, thus indicating that urban forests are a key contributor to overall forest cover.<br /> <br /> At the census block group scale, significant clusters of forest loss were found in many coastal areas and the expanding peri-urban areas around urban centers. Significant clusters of forest gain were in mountainous areas, eastern central Puerto Rico, and select certain coastal areas. Our analyses using GWR models revealed the presence of several spatially-variable relationships between Forest loss and gain with population density and to a lesser extent forest/agricultural occupation density. The strength of these relationships varied from weak to strong and in sign (positive vs negative) across geographical space and from one decadal interval to the next. Population density had a stronger influence on forest losses in densely urbanized areas. The focus on census block groups provided an innovative way to assess LULC changes in Puerto Rico but was also limiting in the variables for which data were available at that scale. The results of this study are useful to understand at the landscape scale the long-term effects of land use change and socio-economic factors on forest-related gains and losses in an urbanizing tropical environment and their implications for ecosystem function and the provision of services. Future research should include other relevant variables like road density and precipitation, and other socio-economic data which could also have very large influences on LULC change.