Advertisement

Socio-Ecological Practice Research

, Volume 1, Issue 3–4, pp 297–324 | Cite as

Use of the McHargian LUSA in agricultural research and decision-making in the age of non-stationarity and big earth observation data

  • Theodore C LimEmail author
Review Article

Abstract

In the past 50 years, there have been two major changes that are of methodological and consequential importance to the McHargian land-use suitability analysis (LUSA): increasing evidence of non-stationarity of global and regional ecological conditions and increasing availability of high-resolution spatial–temporal earth observation data. For 50 years, the McHargian LUSA has been an important analysis tool for designers and planners for both regional conservation planning and development. McHarg’s LUSA is a decision support tool that reduces the dimensions of spatial–temporal data. This makes the technique relevant beyond decision support to spatial identification and prediction of areas of socio-ecological opportunity, risk, and priority. In this article, I use a set of recent studies relating to agricultural LUSA to reveal relationships between the traditional McHargian LUSA and related spatial–temporal research methods that are adapting to more data and non-stationary ecological conditions. Using a classification based on descriptive, predictive, and prescriptive research activities, I organize these related methods and illustrate how linkages between research activities can be used to assimilate more kinds of spatial “big data,” address non-stationarity in socio-ecological systems, and suggest ways to enhance decision-making and collaboration between planners and other sciences.

Keywords

Land-use suitability analysis Non-stationarity Big earth observation data Agriculture 

1 Introduction

The land-use suitability analysis (LUSA) is a method popularized by Ian McHarg’s seminal work, Design with Nature, originally published in 1969. McHarg’s original LUSA was a spatial overlay technique in which social and environmental variables of the region of interest were gathered and spatially mapped, classified by suitability for different land-use types (e.g., Residential, Conservation, Industrial, etc.) and then overlaid on each other, creating a composite map of suitability that could be used to support planning decision-making processes. The principle is that by gathering these layers and organizing them into a composite map, better plans can be made and better outcomes achieved that meet both ecological and social values.

In Design with Nature, after introducing an example of a composite map produced through his overlay technique, McHarg pointed out “[the composite of the overlays] is not a plan. A plan includes the entire question of demand and the resolution of demand relative to supply, incorporating the capacity of the society or institution to realize its objective.” (McHarg 1969, p 105). In other words, the outcome or goal of the McHarg’s LUSA was neither prediction nor prescription, but a rather, a rational, spatially explicit, and reproducible synthesis of the data from the socio-ecological due diligence and fact-gathering stages of the planning process. The McHargian LUSA can be thought of as a dimension reduction technique where multiple criteria, or “dimensions” for the site, are combined to form one quantitative and descriptive summary of the study area. Inherent in the above quotation are two questions that are not addressed by the McHargian LUSA itself but are critical to spatial planning. The first question is that of demand—or the program that must be accommodated within the LUSA-described space. The second question is that of allocation—or the process by which conflicts due to space and programmatic limitations are resolved. These two questions—whose goals are predictive and prescriptive—are closely related to the descriptive power of the McHargian LUSA. They extend the fact-finding process to what may occur and what should occur in the study area in the future. While these two areas have not been a part of the traditional LUSA, they are major areas of socio-ecological research in the fields of land-use modeling and spatial optimization, respectively.

In the decades following the publication of Design with Nature, there have also been two major changes that are of consequence to the LUSA that will be addressed in this study. First, of methodological importance, is the increase in the amount of geospatial data now more readily available for analysis (“big earth observation data” or BEOD). Second, of consequential importance, is the increasing evidence of spatial and temporal non-stationarity in global and regional ecological conditions. “Non-stationarity” refers to when an observed pattern, such as the mean or variability, in a given variable changes over space and/or time. Long-term climatological probabilities, for example, are now considered susceptible to non-stationary conditions because of climate change.

In this paper, I show how an expanded review of the socio-ecological literature to include goals of prediction and prescription, in addition to the descriptions enabled by LUSA, illuminates opportunities for transdisciplinary collaboration, especially considering BEOD and non-stationarity. I use a literature review of agricultural LUSA toward answering two questions:
  1. 1.

    How does the McHargian LUSA relate to other quantitative spatial–temporal land suitability methods, such as land-use modeling and spatial optimization techniques?

     
  2. 2.

    How are big earth observation data (BEOD) and non-stationarity being incorporated into LUSA and related methods?

     

I selected the case of agriculture because of its relevance to LUSA, BEOD, and non-stationary processes. First, like other land uses, there are many opportunities to incorporate values and knowledge into planning areas suitable for agriculture. Expert knowledge is frequently used in determining what criteria should be included in the LUSA, how criteria should be binned to represent suitability, and how the different criteria should be weighted against each other, based on importance. Second, agricultural land is susceptible both to development, and to abandonment, naturalization, and reforestation. The trade-offs in ecosystem services that result between land-use decisions around agricultural land are therefore well-suited to LUSA (see, for example, McHarg 2014, pp 181–190, for his thinking on the susceptibility of agriculture and other ecosystem services on undevelopment land to fragmented suburbanization and Goldstein, et al. 2012, pp. 7568–7569, for analysis on different kinds of ecosystem service trade-offs involving agriculture). Comparisons between agricultural locations to identify and prioritize key agricultural lands have been part of the US Department of Agriculture Land Evaluation and Site Assessment decision support system since the 1970s (more details are in the following section). Third, more so than other land-use considerations, agricultural production is both a social and a biophysical process that is subjected to non-stationary conditions and has been shown to incorporate feedback loops that affect subsequent suitability of land for agricultural use. The selection of one particular land-use type narrowed the extent of the literature review and allowed for more focused illustration of examples.

2 The relevance of BEOD and non-stationarity to agricultural land-use suitability

The volume of BEOD readily available for public use has increased manifold since the first remote sensing programs were launched in the 1960s (Guo et al. 2015, p. 109). Much of the growth in BEOD has been driven by remotely sensed (RS) products collected through sensors mounted on orbital satellites. For example, in 2008, when NASA announced free, web-enabled access to data collected through the Landsat program, a civilian satellite program whose mission objective is to monitor and conduct scientific and exploratory studies of the Earth’s surface, there were over 2 million images collected between 1972 and 2008 (Woodcock et al. 2008, p. 1011). In 2015, there were over 5 million images in the USGS archive (Wulder et al. 2016, p. 282). NASA projects the volume of data stored and distributed through its Earth Observing System Data and Information System (EODIS) to accelerate through 2025, with the years between 2017 and 2022 having data ingestion rates projected to grow from 3.9 petabytes (PB) per year to as much as 47.7 PB per year, and by 2025, the volume of data in the EOSDIS archive is expected to be more than 246 PB. (https://earthdata.nasa.gov/about/eosdis-cloud-evolution).

Table 1 shows examples of RS-derived data products, with currently available temporal coverages and spatial resolutions. In addition to raw BEOD, there are also a host of derivative products that are made and distributed. An example of a derivative product is the popular Normalized Difference Vegetation Index (NDVI), which is a calculated ratio using near-infrared and red channels of multispectral imagery to represent vegetation density. Data products have also been developed into indexes for specific applications, such as monitoring urbanization, infrastructure, commodity stockpiles, agricultural productivity, and in emergency and disaster management and mitigation (UNGWG 2017, p. 18).
Table 1

Examples of RS data sources

(adapted from Kerr and Ostrovsky 2003, p. 301)

Data source

Spatial resolution

Temporal Res

Temporal coverage

Description

URL

MODIS Vegetation Indices

250–1000 m

16 days–monthly

2000–present

AVHRR global/continental land-cover products using six different classification schemes

https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/mod13q1_v006

SPOT/VGT composites

1000 m

1 day, 10 days

1998–present

Composite indices available

http://www.spot-vegetation.com/

University of Maryland Global Land Cover Facility

Various

Varies

Varies

Very large satellite data archive including land-cover products and processed satellite imagery with global coverage, examples include: Impervious surface cover, tree cover, flood maps, leaf area index, photosynthetically active radiation, vegetation index (NDVI)

http://www.landcover.org/data/

Global Land Cover 2000

1000 m

one time, representing state at the year 2000

2000

Global land-cover data product based on global data acquired by the VEGETATION instrument on SPOT 4 satellite

http://forobs.jrc.ec.europa.eu/products/glc2000/glc2000.php

CORINE land cover

500 m

annual

1990, 2000, 2006, 2012, 2018

Land cover inventory with 44 classes. 38 participating countries

https://land.copernicus.eu/pan-european/corine-land-cover

USGS National Land Cover Dataset

30 m

 

2001, 2006, 2011, 2016 (forthcoming)

Quantifies land cover and land-cover change in the coterminous USA

https://www.mrlc.gov/national-land-cover-database-nlcd-2016\

USGS Gap Analysis Program: National Terrestrial Ecosystems

30 m

one time, representing state at the year 2011

 

Land-cover data includes between 8 and 590 land-use classes, using Landsat 1999–2001 imagery as the base for its models

https://gapanalysis.usgs.gov/gaplandcover/

There has also been an increase in global interpolated datasets. Unlike RS-data products, which are derived from data collected via satellite at global coverages, global interpolated datasets apply spatial interpolation techniques to data collected in specific locations on Earth in order to derive a continuous surface of values with global coverage. An example of this is the ISRIC-worldgrid for soil data (https://www.isric.org/explore/soilgrids). The global ISRIC 250 m resolution gridded dataset (SoilGrids250 m) was created using machine learning algorithms trained on 150,000 soil profiles and remotely sensed soil covariates and is made available through an online interface or programmatically through REST APIs (Hengl et al. 2017, p. 3). Another example is WorldClim, a gridded high-resolution (1 km2 resolution) monthly climate data source interpolated using data from up to 60,000 weather data stations (Fick and Hijmans 2017, p. 5).

Both RS data and interpolated data are usually stored and distributed in a raster-like format—where data are stored in an equally spaced grid array corresponding to the horizontal resolution of the data. This format is very amenable to incorporation into the type of Map Algebra weighted overlay technique that is at the core of the McHargian LUSA (Tomlin 1990).

Non-stationarity can either be spatial or temporal. Spatial non-stationarity refers to situations when a “global” representation of a phenomenon would fail to capture the localized structures within the study area (Fotheringham 2009, p. 398). Temporal non-stationarity refers to situations where conditions are changing over time. Climate change is one example of temporal non-stationarity, since past climate patterns will not predict future climate patterns (Karl and Trenberth 2003, p. 1721). Social processes of technology diffusion result in spatial–temporal non-stationarity, since over time, both the locations and timing of these adoptions incorporate feedback loops that make them dynamic and subject to larger patterns that emerge from individual behaviors.

The increase in BEOD and consideration of non-stationary conditions are particularly relevant when assessing agricultural land suitability. Handbooks summarizing environmental factors influencing productivity of common crops have been available for decades and are commonly used in assessing land capability globally (see, for example, FAO 1976; USDA 1961). Factors that are commonly considered include: climate, topography (e.g., slope and aspect), and soil fertility, as there is much evidence of the impact of these environmental variables on crop growth and yields. For example, in the US Midwest, where the highest global yields of corn and soybean are observed, soil properties have been found to explain 30% of yield variability, with soil organic matter explaining the most (Kravchenko and Bullock 2000, p. 79). The role of soils in evaluating agricultural productivity has been used in the land evaluation and site assessment (LESA) method adopted by the Soil Conservation Society (now the Natural Resources Conservation Service, NRCS) of the USDA since the 1980s. This program is discussed more in the following subsection. Interannual variability in crop production is explained by weather conditions faced during the growing season, which are linked to global long-term climatologies (Kellner and Niyogi 2015, p. 18). Global BEOD datasets such as WorldClim and SoilGrids250 m can be useful in assessing land suitability for specific crops, especially in parts of the world where local datasets may be lacking. In the USA, the amount of land in agricultural (cropland and pastureland) use collected by the USDA’s Census of Agricultural is also complemented by its Natural Resources Conservation Service (NRCS) Natural Resources Inventory data product, which is based on remote sensing (USDA 2018).

Non-stationarity stemming from climate change has been a major source of uncertainty for crop growth modeling, which is an important research activity for agricultural LUSA, because the prediction of how a crop will fare under future circumstances has major impacts on the long-term viability of this land use. The past 15 years has seen conflicting results of how increased atmospheric CO2 concentrations and warmer summertime temperatures might affect corn growth. There is much concern that climate change can result in increased droughts and flooding. Process-based models based on decades of field experiments on how plant physiology responds to heat unit accumulation vary in their predictions of outcomes of increased levels of atmospheric CO2, with some models foreseeing positive effects of warming and increased CO2 for plant growth processes, and others predicting decreased yields. Statistical–empirical evidence, however, tended to indicate that high temperatures would cause stress during key crop growth stages and therefore have very negative impacts on the yields of much of the world’s grains and legumes (Schlenker and Roberts 2009, p. 15594; Roberts et al. 2013, p. 236).

Agricultural productivity is not merely a function of environmental factors. Productivity is also highly dependent on technological adoption and land management practices, which are both social processes driven by market changes and technological diffusion dynamics within farming communities (Berger 2001, p. 247). Empirical research shows evidence of spatial differences in high-yielding variety adoption by farmers (Griliches 1957, p. 501; Feder and Umali 1993, p. 215; Foster and Rosenzweig 1995, p. 1195) and management practices: crop rotation (Lockie et al. 1995, p. 61), fertilizer application (Kassie et al. 2013, p. 536), and irrigation (Conley and Udry 2001, p. 668), for example. Proximity to markets and infrastructure can also affect the economic viability of agricultural land use (Fuglie and Kascak 2001, p. 386). Many of these studies highlight the processes of technological diffusion through information networks such as extension services or peer-to-peer social networks, indicating that technological adoption is not a function of static individual characteristics or social factors, but is a dynamic process influenced by neighbors and spatial proximity (Li et al. 2013, p. 632). In addition, global or local networked effects—such as global trade practices and policies, social processes of technology diffusion, or proximity to reliable water delivery infrastructure—might be imperceptible from BEOD datasets. Nevertheless, they remain important factors when assessing the viability of agricultural land use, especially under longer time ranges, when such temporal and spatial non-stationarity is likely to become more perceptible, and when past patterns are not as likely to be able to predict future patterns.

Another source of non-stationarity may come from agricultural practices themselves. Unsustainable agricultural practices can influence the long-term fertility of the land (Mueller et al. 2010, p. 604). Agricultural land use on steep slopes results in erosion from wind and water. Low yields may cause farmers to remove grass strips, hedgerows, and shelterbelts to maximize field area, leading to higher rates of soil erosion. Crop intensification and the use of heavier machines can also damage the crop ecosystem (Pimentel et al. 1995, p. 1117), which can result in a feedback where increasingly marginal lands are developed for agriculture. Studies also demonstrate that crop intensification and increased irrigation in the US Midwest have induced changes to regional climate patterns. Increased evapotranspiration of highly yielding varieties of corn increases atmospheric water vapor, cooling temperatures that might otherwise result in crop-damaging heat extremes, an example of a positive feedback loop where crop intensification leads to more favorable conditions for crop growth (Lobell and Bonfils 2008, p. 2068; Mueller et al. 2016, p. 5).

2.1 Agricultural land evaluation and site assessment (LESA)

The focus of this paper on reviewing agricultural LUSA necessitates an explanation of the highly relevant and related decision-making tool, land evaluation, and site assessment (LESA). The LESA method is a scoring system and decision support tool that was developed in 1971 by Lloyd E Wright (Steiner et al. 1994, pp. 32–34). While originally applied to determine land values for tax purposes, it was later piloted by the USDA as a way to compare across agricultural properties to quantify how federal projects might influence the nation’s supply of highly productive agricultural farmland. LESA is composed to two parts: (1) Land evaluation is the process of identifying soil limitations and farmland ratings and involves experts such as conservationists, cooperative extension representatives, soil and water conservation district representatives, farmers, planners, local agricultural officials, and others with local land resources (Steiner et al. 1994, p. 35). (2) Site assessment is the process of scoring non-soil fertility-related attributes of the land context, including parcel size, on-farm investment, surrounding land uses, zoning ordinances, and other farmland protection policies or programs. Site assessment involves local officials or a locally appointed site assessment committee (Steiner et al. 1994, p. 35).

Those opting to use the LESA decision support system can determine the points and weights allocated to the land evaluation and site assessment portions of the analysis. This process of points and weights determination makes it highly related to the McHargian LUSA, though less emphasis is placed on the spatial overlay mapping part of the process than in traditional overlay analyses (e.g., of traditional LESA scoring, see Wright et al. 1983, p. 86). However, some have also attempted to incorporate geospatial data for larger areas into the LESA framework using GIS (e.g., see Dung and Sugumaran, 2005). The LE and SA portions of LESA, each contain different criteria, can also be used separately (see Steiner et al. 1987 pp. 185–187, for example, criteria in LE and SA).

Since its original official adoption by the USDA in 1981, numerous local and state public agencies have used LESA to support land-use decision-making processes (Coughlin et al. 1994, pp. 7–8). For example, to protect against the threat of suburban sprawl and loss of agricultural lands, LESA criteria were incorporated into the Farmland Protection Policy Act (Steiner et al. 1994, pp 43–54). Land trust and transfer/purchase of development rights (TDR/PRD) programs, which compensate farmers for lost potential revenue of foregone urban development in favor of continued agricultural use, use the LESA system to determine which lands should be retained for continued agricultural use and to prioritize TDR and PDR programs (Hoobler et al. 2003, p. 110). Nonprofit land trusts use the LESA method to prioritize strategic land acquisition (Hoobler et al. pp. 6–10). Like the traditional McHargian LUSA, LESA is used as an aid in decision-making, to reduce complex factors into a more comprehensible score that is reflective of the decision-makers’ values.

To summarize, (1) BEOD has the potential to improve measurements of agriculturally relevant conditions; (2) decisions about land-use conversion between agricultural and other land-use types (including urbanization and reforestation) are important and have long-term effects; and (3) decisions to develop agricultural land, maintain it in agricultural land use, or naturalize the land are in-part based on expectations and assumptions about non-stationarity, including changes in climate, such as frequency of droughts and floods and broader social change.

3 Methods

In order to illustrate the range of goals and applications of the LUSA in agriculture, I conducted a systematic literature review. The objective of the literature review was to capture studies that utilize the McHargian overlay analysis to quantify agricultural suitability, but also to capture studies evaluating agricultural suitability using related methods. These methods have been mentioned by others and include: participatory models, fuzzy logic, linear programming and optimization, cellular automata, artificial intelligence, and machine learning methods (Collins et al. 2001, p. 616; Malczewski 2004, Chapter 4). Because land-use modeling, the prediction of land use of land-cover change, often uses similar datasets to predict land-use change as LUSA uses to assess suitability, I hoped to capture a representative slice of agricultural land-use modeling in the literature review. Theoretically, there is a potential information feedback loop between LUSA and land-use modeling, where the former identifies areas that are suitable for agriculture given input criteria, and the latter identifies which criteria have been important in explaining past agricultural land cultivation.

The selection of research articles was conducted through multiple queries through EbscoHost. Search terms included: “agriculture” or “agricultural” and each of the phrases “land-use suitability,” “land suitability,” “land suitability analysis,” and “land-use suitability analysis.” The search terms were intentionally left sufficiently broad as to capture a range of methods related to traditional LUSA, including land-use/land-cover modeling. Articles were limited to publication within the 10-year period between 2009 and 2018 to reflect the current state of the art. Only articles in English, whose full texts were available through the EbscoHost database, which were published in peer-reviewed academic journals were included. A total of 93 unique articles were returned. Each article was screened for quality (comprehensibility in English, clarity of focus, and methods) and relevance (land-use suitability analysis related and agricultural land use related), leaving a total of 63 articles. It was not required that the included studies feature LUSA as the primary motivation of the study, and therefore, the study set also included research that used LUSA as an intermediate data processing step for other research objectives.

3.1 Literature classification

After selecting the articles to be reviewed, they were classified into categories. Past reviews have included classifications to organize the LUSA literature and illustrate trends in research methods. In 2001, Collins et al. traced the historical development of LUSA in the USA, describing six eras of the LUSA. The modern (post-computer-assisted overlay) areas were: the redefinition of spatial data and multicriteria evaluation, which included methods such as fuzzy set theory and multicriteria decision-making; the replication of expert knowledge, which included artificial intelligence tools; and the “new horizon” of land-use suitability techniques, which included heuristic search process, expert systems, neurocomputing, and genetic programming (Collins, Steiner, and Rushman 2001, p. 616). Malczewski (2004, Chapter 4) categorized GIS-based LUSA into the following broad groups: computer-assisted overlay mapping, multicriteria evaluation methods, artificial intelligence, or “soft computing or geocomputation” methods (including fuzzy logic, neural networks, evolutionary/genetic algorithms, and cellular automata techniques).

The above classifications are primarily based on methods. The classification used in this research focused instead on the purpose research activities included in each study. For each study included in the literature review, I made note of the presence of whether the study included goals of description, prediction, and/or prescription, where the following definitions were used to identify these goals:
  • Description: A research activity whose goal is to summarize conditions in the area of interest for agricultural suitability. Multiple criteria, or dimensions, of the area of interest are considered and reduced into one composite map that describes the site. These studies include methods of dimension reduction or data aggregation that allow decision-makers to more easily comprehend and process sub-areas of opportunity, risk, or priority within the area of interest.

  • Prediction: A research activity whose goal uses causal relationships of a particular phenomenon (e.g., land-use or land-cover change) to project or simulate into the future what may occur under different or future circumstances (e.g., climate change). These studies may employ empirical–statistical, stochastic, optimization, or dynamic simulation (e.g., cellular automata, agent-based modeling, or physically based models, such as biophysical crop growth models and physics-based hydrological models) (Lambin, Rounsevell, and Geist 2000, p. 325). They may be based on either deductive (theoretical) or inductive (empirical) logic (Overmars et al. 2007, pp. 584–585).

  • Prescription: A research activity whose goal is to optimally allocate scarce resources (e.g., land) considering one or multiple potentially competing objectives (e.g., accommodating projected growth while minimizing environmental impacts).

Descriptive activities correspond closely to the original intent of the McHargian LUSA to gather the spatial socio-ecological dimensions of the site, organize them, and weight them according to their relevance to some proposed land use, reducing the site’s multidimensionality to a highly relevant suitability index. The LESA process (see Sect. 2.1) similarly uses a scoring system that evaluates both soil-based agricultural productivity and other socioeconomic contextual factors in decision-making, reducing sites’ multiple dimensions to a single comparative score. Prediction and prescription activities extend the descriptive process to what may occur and what should occur in the study area in the future given the objectives of the decision-makers. Others have used a similar framework to classify agricultural supply chain studies, referring to these goals as “levels of analysis” (Sharma et al. 2018, p. 105). The classification used in this paper differs in its emphasis on goals of the activities rather than “levels” of analysis and distinguishes description and prescription based on the presence of an explicit objective function and/or trade-offs considered in the analysis.

In addition to noting research activities in each study according to this framework, I also made note of the use of BEOD and attention to issues of non-stationarity. BEOD was defined as the use of either RS data or globally interpolated datasets. Non-stationarity was considered “addressed” if the authors theorized about potential spatial–temporal feedbacks, locally varying (not global) parameters, climate change, parameter uncertainty, or scenario planning and incorporated these theorized dynamics into their research methods.

4 Results

Of the 63 articles included in the literature review, 70% included descriptive research activity, 44% included predictive research activity, and 6% included prescriptive research activity. Figure 1 summarizes the number of articles including each research activity type. Table 2 summarizes all the studies included in the review, including the presence of each research activity type, the motivations for each study, locations, decision-makers included (especially in the determinations of criteria weights), the resolution of grids/units of analysis, and the geophysical and social criteria considered in the analysis. All but two studies included geophysical criteria or variables in their research activities. Thirty-three out of 63 studies included social criteria or variables.
Fig. 1

Number of articles of each type published

Table 2

Summary of studies included in review

Desc

Pred

Prescr

BEOD

Non-station.

Motivation

Location

Decision-makers

Unit of Analysis/Spatial resolution

Geophysical criteria considered

Social criteria considered

Study

X

    

Determine best land-use types

Turkey

Not specified

map units

Soil, water

Fertilizing practices, irrigation scheduling

Dengiz et al. (2010)

X

    

Integrated evaluation of coastal land use

Malaysia

28 experts

20 m × 20 m

Proximity to geo-hazard risk areas

Population density, transportation access, public health access, beach access, school access, proximity to life support systems, proximity to high-value areas, proximity to various industries, existing plan

Pourebrahim et al. (2011)

X

    

Identify less favored areas and biophysical constraints

Lithuania

Experts

Not specified

Soil texture, soil drainage, and terrain slope

 

Jarasiunas et al. (2017)

X

    

Develop geopedological approach to mapping suitability

Iran

NA

Not specified

Soil, climate

 

Mousavi et al. (2017)

X

    

Identify areas suitable for citrus cultivation

Iran

30 experts from the Iran Citrus Research Institute

80 m × 80 m

Climate, elevation, aspect, slope, water resources

Roads network, population areas

Zabihi et al. (2015)

X

    

Identify most important factors for saffron cultivation in region, assess land

Iran

20 local saffron experts

100 m × 100 m

Climate, soil, topography

 

Maleki et al. (2017)

X

    

Agricultural planning

Spain

Various sources, though not clear exactly how they were incorporated

Cluster based

Natural environment

Socioeconomic conditions, infrastructure, and legal framework “objective information,” “local experts information,” “farm information”

Cardín-Pedrosa et al. (2012)

X

    

Identify areas for forest conservation (for water quality) through multicriteria decision analysis

Brazil

University professors, researchers, landscape ecologists, forest hydrologists, conservationists (21 total)

30 m × 30 m

Land-use suitability (biogeophysical), soil erodability, erosivity, proximity to roads, proximity to surface water

 

Vettorazzi and Valente (2016)

X

    

Identifying areas “suitable” for agricultural land use according to expert weights of input layers@@@@Planned dams will inundate the study area (“an actual problem”)

Iran

Agronomists and local university faculty

30 m × 30 m

Slope, aspect and elevation, soil depth, erosion degree, land cover, groundwater levels

 

Akinci et al. (2017)

X

    

Identifying areas “suitable” for agricultural land use according to expert weights of input layers

Iran

Agronomists and local university faculty

90 m × 90 m

Soil parameters, climatic data, topographic data, land-cover data

Land use

Kazemi and Akinci (2018)

X

    

Develop systematic evaluation of land quality

India

Experts

Not specified

Soil characteristics

 

Chatterji et al. (2014)

X

    

Quantitative evaluation of local soil series using different methods

India

Participatory sessions with local farmers

NA

Soil characteristics

 

Vasu et al. (2018)

X

    

Identify criteria important to local farmers, group learning

Ghana

Local participants

NA

Changes in precipitation

Changes in subsidies, prices, and government credits

Badmos et al. (2014)

X

    

Identify areas suitable for barley cultivation

Iran

Experts

Not specified

Slope, groundwater, soil characteristics

 

Hamzeh et al. (2014)

X

    

Identify areas suitable for jute and lentil cultivation

India

10 agronomy experts

Not specified

Soil characteristics

 

Singha and Swain (2018)

X

    

Assessing land-use suitability for agriculture and grassland

Iran

Experts

Not specified

Physical, agronomic

Socioeconomic

Memarbashi et al. (2017)

X

    

Identify areas suitable for faba bean cultivation

Iran

Experts

Not specified

Soil, climate, topography

 

Kazemi et al. (2016)

X

    

Create agroecological zones identifying candidate crop types

India

Experts

Not specified

Altitude, climate

 

Sati and Wei (2018)

X

    

Design indicators and reveal system transitions in agroecology/farmland diversity

India

Literature (Reidsma et al. 2006)

NA

Biodiversity

 

Amjath-Babu and Kaechele (2015)

X

    

Identifying areas “suitable” for agricultural land use according to expert weights of input layers

Turkey

Local agronomists, local university faculty members

25 m × 25 m

Soil group, land-use capability class, soil depth, slope, elevation, erosion level, and other soil properties

 

Akıncı et al. (2013)

X

    

Identify areas of opportunity for a new cropping system involving burclover

United States

Literature and farmers’ knowledge

1 km × 1 km

Soil, environmental, and climate

 

Mbũgwa et al. (2015)

X

    

Other scoring methods cannot incorporate > 10 factors

United States

Soil professionals and previous agricultural land suitability studies

50 m × 50 m

Terrain, fertility, depth to water/bedrock, soil density, climate, economics, accessibility, management, topography, soil characteristics

 

Montgomery et al. (2016)

X

    

Find sites for industry, taking into account natural factors and excluding economic

Turkey

Seven experts

5 m × 5 m

Wildlife development, protected areas, vegetation, olive groves, pastures, agricultural lands, water, streams, land capability class, slope, erosion

Settlement, transportation,

Aktaş et al. (2018)

X

    

not specified

Turkey

Not specified

map units

Soil characteristics

Agricultural profitability index

Kiliç (2011)

X

    

Sensitivity analysis to layer inclusion/exclusion

United States

Not specified

polygons based on cluster analysis

Disturbance change, vegetation change, land polygon size, cover type max patch size, cover type total size

Land-use condition, road density,

Humphries et al. (2010)

X

  

X

 

Use farmers’ existing knowledge as a starting point for land evaluation criteria

Jordan

Not specified

Not specified

Soil characteristics, precipitation, topography

 

Ziadat and Sultan (2011)

X

  

X

 

Develop a Web-based tool (Google Earth Engine) to ingest global datasets for agricultural suitability

Ethiopia

NA

varies

Rivers/water bodies, soil characteristics, elevation

Towns, land use

Yalew et al. (2016)

X

  

X

 

Identify areas suitable for shea cultivation

Sub-Saharan Africa

Experts

Not specified

Temperature, precipitation, elevation, fire, Normalized Difference Vegetation Index (NDVI), soil type and soil drainage

Land-use

Naughton et al. (2015)

X

  

X

 

Develop new land-use plan

Bangladesh

5 experts on Detailed Area Plan

300 m × 300 m

Existing land cover, elevation, distance to fault line

Utilities, proximity to transportation and amenities

Ullah and Mansourian (2016)

X

  

X

 

Update suitable areas for oil palm (previous 1969 study’s rainfall and water deficit have changed)

Ghana

Experts

1.1 km × 1.1 km

Soil, climate, topography

 

Rhebergen et al. (2016)

X

X

   

Identify areas for rice cultivation

Pakistan

Agronomists

Not specified

Temperature, soil type, soil pH, soil drainage, soil electrical conductivity

 

Raza et al. (2018)

X

X

   

Describe agricultural suitability taking into account water quality objectives

Hypothetical

Experts

NA

Productivity and larger-scale water contaminant impacts

 

McDowell et al. (2018)

X

X

   

Develop an agroecological index that incorporates local data, knowledge, into a model that can be used for prediction

Vietnam

Local and international experts

30 m × 30 m

DEM, water bodies, soil losses by water erosion (estimated through universal soil loss equation), precipitation, slope length and steepness factor, soil erodability from topsoil texture and organic matter content, cover factor for rubber

Proximity to roads

Nguyen et al. (2015)

X

X

   

Identify areas suitable for crops

Romania

Authors

1 km × 1 km

Slope, altitude, slope orientation, density of fragmentation, probability of landslides, flooding, temperature, rainfall

 

Bilaşco et al. (2016)

X

X

   

Incorporate waterlogging risk into the National Standard for Land Suitability Analysis

China

15 experts from universities, planning, design companies, and construction

not specified

DEM, seismic, meteorology, water resources, geotechnical, nature and ecology

Human impact

Jiao et al. (2017)

X

X

 

X

 

Land-use planning at a regional level using the decision support system MicroLEIS (interactive software)

Iran

NA (model)

not specified

Soil characteristics, climate, production and ecosystem modeling, erosion, and contamination modeling, and impact and response simulation

Engineering and technology prediction, information and knowledge databases

Shahbazi and Jafarzadeh (2010)

X

X

 

X

 

Use participatory model to identify factors of land-use suitability, and spatially predict suitability for a given land use

United States

Business leaders, NGO scientists and managers, educators, scientists, policy-makers, government officials, and members of the public

30 m × 30 m

Environmental constraints, soils, slope, drainage, productivity, water source, climate, rare species, drinking water,

Zoning/planning, environmental amenities, community amenities, roads, public utilities, urban/rural, market access, proximity/diversity of nearby markets, housing density, historical/cultural features, education, property taxes, permitting, views, recreational opportunities

Meyer et al. (2014)

X

X

 

X

 

Predicting effects of land-use change (reforestation from Grain to Green Program: agriculture to forest/grass) on hydrology

China

Not specified

30 m

elevation, slope, geomorphology, soil organic matter, soil drainage condition, soil PH, mean annual rainfall, accumulated temperature of 10 °C and distance from water source

 

Yu et al. (2018)

X

X

  

X

Potential for other land uses in a forested area

India

Authors

30 m × 30 m

Topography, environment, ecology

Social, infrastructure, demographic factors

Jeganathan et al. (2011)

X

X

 

X

X

Beyond expert weights, reveal emergent patterns through CA rules

Australia

Surface-water hydrologists, groundwater hydrologists, soil scientists and irrigation specialists

100 m × 100 m

Slope, soil texture, depth to water table, electrical conductivity of groundwater, hydraulic conductivity of soil, distance to streams.

Irrigation

Yu et al. (2011)

X

X

 

X

X

Assess effects of climate change on crop (maize) production

Canada

NA (model)

soil polygons

Soil, climate, landscape

 

Gasser et al. (2016)

X

X

 

X

X

Determine how low/high cropping intensity influences yields, implications for climate change

West Africa

NA (model)

not specified

Climate, spatial resolution, crop model parameter inputs

 

Challinor et al. (2015)

X

X

 

X

X

Model how sika deer habitat suitability will change/has changed because of land-use change and climate change

Japan

NA (model)

1 km × 1 km

Topography, snow cover, climate, forest cover, land use

 

Ohashi et al. (2016)

X

X

 

X

X

Future scenarios on land suitability

India

Local experts

0.25’ × 0.25’

Aspect, elevation, slope, rainfall, temperature, soil depth, geology, distance from the river, distance from built-up, distance from cropland, distance from forest cover

Distance from the road, distance from the rail,

Sahoo et al. (2018)

 

X

   

Predict land use given long-term optimal allocation of agricultural demand (2015)–2040)

Iran

Not specified

Regional level

Annual precipitation, annual average temperature, soil fertility, average elevation, and average ground water depth.

 

Mesgari and Jabalamel (2018)

 

X

   

Develop a land-use models that takes economic processes into account

European Union

NA (model)

100 m × 100 m

Supply of crops and animal outputs

Global trading model

Ustaoglu et al. (2016)

 

X

   

Determine effects of various management practices on behavior and the presence of bird (timing of mowing and land abandonment)

Germany

NA (model)

NA

 

Agricultural management practices

Arbeiter et al. (2018)

 

X

 

X

 

Develop land-use conversion model to evaluate scenarios’ environmental impacts

Canada

NA (model)

not specified

Distance to closest water body

Population density, distance to closest city, distance to closest road,

El-Khoury et al. (2014)

 

X

 

X

 

Subnational agricultural statistics too coarse for investment planning. Fill spatial gaps through spatial disaggregation

Sub-Saharan Africa

NA (model)

5 min (~ 10 km)

NA, suitability is pre-calculated

 

You et al. (2009)

 

X

 

X

 

Predicting risk of feral outcrossings of cultivated sorghum

United States

NA (model)

NA

Micro-topography, vegetation cover

Road type, road material, nearby land use (crop)

Ohadi et al. (2018)

 

X

  

X

Use ecological niche model to predict crop choice: cassava, fruit trees, heavy rice, and jasmine rice

Thailand

NA (model)

NA

Environmental, geographic

Social

Heumann et al. (2013)

 

X

  

X

Develop future agricultural land use based on habitat suitability

Iberian Peninsula

Authors expert knowledge in conservation biology, socioeconomics, and agronomy

Not specified

Elevation, slope, soil quality, mean annual temperature, and precipitation

 

Cardador et al. (2015)

 

X

  

X

Verify agents’ assumed decision-making models respond in realistic ways

Hypothetical

NA (model)

1 hectare

Slope and precipitation

Labor costs, travel time to regional markets, and purchasing power parity, prices, input costs

Magliocca et al. (2013)

 

X

  

X

Quantify the effects of major environmental and socioeconomic factors on land-use change for each conversion type, accounting for spatial autocorrelation (SAC)

Canada

NA (model)

township

Change in growing season precipitation, frost-free days, growing degree days (5′C), soil moisture, change in daily mean snowpack water equivalent, elevation, proportion of land with suitability ratings of 2 or 3

Change in population density, road density, irrigated or non-irrigated, agricultural land value

Ruan et al. (2016)

 

X

 

X

X

Planning for climate change yields 2021–2050 and 2071–2100

Hungary

NA (model)

County-level

Water, climate

Socioeconomic

Gaál et al. (2014)

 

X

 

X

X

Develop land-use prediction tool for Black Sea Catchment

Black Sea Catchment

NA (model)

1 km × 1 km

Elevation, slope, soil quality, mean annual temperature, and precipitation

 

Mancosu et al. (2015)

 

X

 

X

X

Assess effects of farmland abandonment on habitat suitability for Shrike

Italy

NA (model)

20 m × 20 m

 

Land-use types within buffer area

Brambilla et al. (2010)

 

X

 

X

X

Evaluate climate and land-use change on water budget, using hydrological and socioeconomic

Morocco

NA (model)

100 m × 100 m

Total annual precipitation, average annual temperature, soil classification, geomorphologic

Information, population, employment information, agricultural information (production, productivity, consumption), mobility patterns

Antonellini et al. (2014)

 

X

  

X

Economic theory of agricultural industry clustering

Brazil

NA (model)

250 m (MODIS), 30 m SRTM for elevation, Soils local source

Environmental

 

Richards (2018)

 

X

 

X

X

Determine driving factors of rural land prices

Chile

NA (model)

Land transactions

 

Changes in parcel size, market potential, rural land prices in 1997, crops, grassland, agroecological zones, population classification,

Foster et al. (2016)

  

X

X

X

Determine how prioritizing agricultural preservation vs natural land preservation results in different urbanization patters

Iran

Experts

30 m × 30 m

Biophysical suitability of agriculture and natural land conservation

 

Sakieh et al. (2015)

X

 

X

  

Develop land allocation model that prevents fragmentation

China

Bureaus and experts in relevant government agencies

25 m × 25 m

Biophysical suitability

Accessibility, land-use policy, and stakeholders’ preference.

Liu et al. (2016)

     

Investigate possible increase in revenue for the area from optimizing spatial allocation of crops

Australia

Not specified

5 km × 5 km

Biophysical suitability

Yields, costs, revenue

Benke et al. (2011)

The research activities were found to be associated with utilization of BEOD and consideration of non-stationarity. Table 3 summarizes whether BEOD was included or/and non-stationarity addressed by each type of research activity.
Table 3

Summary of articles considering non-stationarity and including BEOD, by type

Type

NS considered

NS not-considered

BEOD = Yes

BEOD = No

BEOD = Yes

BEOD = No

Descriptive only

0

2

6

23

Some predictive

10

6

6

8

Some prescriptive

1

0

0

1

BEOD was only used in 19% of articles that only included descriptive activities (6 out of 31), while it was used in 53% of those that had some predictive analyses (16 out of 30), and 50% of those that had some prescriptive analyses (1 out of 2). Only 6% of articles that only incorporated descriptive activities considered non-stationarity (2 out of 31), whereas 47% of those that had some predictive analysis in them did (14 out of 30), and 50% of those that had prescriptive analysis did (1 out of 2). In the following sections, more detail is given to illustrate descriptive, predictive, and prescriptive research activities and their relationships to each other.

4.1 Descriptive LUSA

Studies that utilized descriptive methods only were the most numerous. Of the studies that incorporated descriptive activities, the majority employed a technique equivalent to McHarg’s original weighted overlay method, a weighted sum of multiple criteria to form a spatially explicit composite map. These studies were often motivated to spatially identify lands suitable within a given study area to grow a specific crop. The motivation could be to explore the viability of introducing a new crop to a region, for example shea trees in Africa (Naughton et al. 2015), saffron in Iran (Maleki et al. 2017), oil palm in Ghana (Rhebergen et al. 2016), or burclover in the USA(Mbugwa et al. 2015). Criteria commonly used in the weighted sum analysis included: soil fertility characteristics, topography (elevation, slope, aspect), and climate variables, as is suggested by LUSA methods recommended by the FAO (1976) and USDA (1961). Acknowledging that economic viability of agricultural land use requires markets for products to be sold and infrastructure in transportation and production processes, several studies also consider socioeconomic conditions, infrastructure, and legal frameworks (Cardín-Pedrosa and Alvarez-López 2012, p. 89; Pourebrahim et al. 2011, p. 87; Ullah and Mansourian 2016, p. 20; Memarbashi et al. 2017, p. 4; Humphries et al. 2010, p. 229). Studies that included socioeconomic conditions were more likely to include suitability analyses for conservation land uses alongside agriculture land uses. These studies’ incorporation of socioeconomic conditions are similar to the “site assessment” portion of LESA, which is used as a decision-making tool for prioritizing agricultural land conservation in the face of development pressure, by considering both soil productivity and social factors in a weighting scheme.

Dimension reduction from multiple criteria to one composite suitability index is highly related to a large literature in agroecological zonation delineation, where methods range from matrix zonation (such as the Koppen climate classification) and unsupervised clustering techniques based on parameter space distance algorithms (van Wart et al. 2013). In crop suitability analyses, each criterion included is assigned a weight that reflects that criterion’s importance to overall suitability. In the majority of the descriptive studies, these weights were determined by “experts” (agronomists, crop specialists, farmers, extension agents, the authors themselves) through methods such as fuzzy logic, the analytical hierarchal process (AHP), or the closely related analytic network process (ANP). Although these methods are often referred to as “multicriteria decision-making” (MCDM) techniques, it should be noted that in the case of many of the crop suitability assessments, there are often no alternative land-use decisions being considered—the “decision” refers instead to a need to deduce the relative importance of physical factors under conditions of imperfect or missing knowledge of the “true” physical understanding of environmental controls on crop growth and yields. AHP allows experts to ensure the consistency of their weights through pairwise comparisons, and fuzzy logic explicitly acknowledges the gradients of transition in criteria assignments to better reflect human cognition in the layer aggregation process (e.g., Akıncı et al. 2013 p. 72; Montgomery et al. 2016, p. 341).

Another motivation for expert knowledge-driven weighting processes was to compensate for a dearth of locally relevant data in the area of interest. For example, a study evaluating multiple crops in Jordan was motivated to start with local knowledge to inform the weighted sum of criteria because FAO-like suitability criteria typically do not take into account local management practices, such as irrigation. The omission of these practices resulted in large differences between on-the-ground conditions and the actual locations of agriculture and locations that appear to be “suitable” for agriculture based on FAO criteria (Ziadat and Sultan 2011, p. 288). The inclusion of local management practice criteria again is similar to what might be considered in the site assessment portion of LESA.

Consideration of non-stationarity only occurred in two descriptive articles. In one, participation was elicited from local farmers to determine how changes in precipitation patterns and fluctuating market prices would change their choice of crop (Badmos et al. 2014, p. 19). In the other, socio-ecological feedbacks are incorporated into land suitability analyses related to intensification of agriculture in India (Amjath-Babu and Kaechele 2015, p. 174).

4.2 Predictive activities and LUSA

Following studies that utilized descriptive methods, the next most numerous were those that utilized predictive methods. Types of predictive models included both empirical–statistical-based models and process-based models. Empirical–statistical-based models use independent (or “predictor”) variables (such as climate, topography, and soil type) to explain the variability in an observed target variable (or “predictand”) (such as crop yields). Process-based models start with rules that are defined a priori to predict how the system will react under certain circumstances. Examples of this are agent-based and cellular automata models, where the rules of how a particular agent (a person, a tract of land, etc.) might react (e.g., change in state from rural to developed) are based on given transition probabilities and neighboring agents’ states; economic-based models where individuals or regions are expected to act according to theories of benefit maximization and economic rationality; and physics-based models. These two types of predictive models can be thought of as deductive (theory-based) or inductive (observation-based) (Overmars et al. 2007, p. 585). Table 4 shows the target variable (predictand) included in each study, the type of predictive model, and the method of evaluation (if evaluated) for the model.
Table 4

Summary of articles including predictive activities

Predictand

Model approach

Model type

Model evaluation criteria

Study

Daily streamflow

Deductive

Physical hydrology model

Nash–Sutcliffe efficiency, coefficient of determination, and percent bias

Yu et al. (2018)

Contaminant source loads

Deductive

Physical contaminant transport model

None

McDowell et al. (2018)

Land-use change of a cell

Inductive

Stepwise logistic regression

AUC-ROC

Sahoo et al. (2018)

Dry rubber yield

Deductive

Results of farmers’ AHP weights

Coefficient of determination (R2)

Nguyen et al. (2015)

Crop production

Deductive

Spatial allocation model (entropy approach)

Coefficient of determination (R2)

You et al. (2009)

Land-use change of a cell

Deductive

Linear programming using net present value method

None

Ustaoglu et al. (2016)

Species presence, arrival departure, and detection

Inductive

Statistical (Markov Chain Monte Carlo)

F statistic

Arbeiter et al. (2018)

Land-use change of a cell

Deductive

Cellular automata with scenario-based rules

Kappa statistic, fuzzy Kappa, and location.

Mancosu, et al. (2015)

Species observations (presence/absence)

Inductive

Statistical

AUC-ROC, specificity, sensitivity

Cardador et al. (2015)

Land-use change of a cell

Deductive

Agent-based model using economic theory

 

Magliocca et al. (2013)

Species observations (presence/absence)

Inductive

Logistic regression

AIC, AUC-ROC

Brambilla et al. (2010)

Land-use change of a cell

Inductive

Statistical spatial autocorrelation model, simple linear regression, spatial autoregressive model, spatial error model

Coefficient of determination and AIC

Ruan et al. (2016)

Land-use change of a cell

Deductive

Results from AHP used as rules for cellular automata

None

Yu et al. (2011)

Hydrological water budget

Deductive

Catchment water balance model

None

Antonellini et al. (2014)

Species observations (presence/absence)

Inductive

Logistic regression

Chi-square statistic

Ohadi et al. (2018)

Land-use change to agriculture in a cell

Inductive

Statistical (panel regression)

Coefficient of determination (R2)

Richards (2018)

Farmers’ crop choice

Inductive

MaxEnt model (machine learning)

AUC-ROC

Heumann et al. (2013)

County yields

Inductive

Statistical regression

coefficient of determination (R2)

Gaal et al. (2014)

Land-use change of a cell

Inductive

CLUE model (logistic regression)

AUC-ROC

El-Khoury et al. (2014)

Rural land prices

Inductive

Statistical econometric method (instrumental variables)

Coefficient of determination (R2), root mean squared error, F-test

Foster et al. (2016)

Land-use change of a cell

Deductive

Future land scenarios based in suitability for agriculture and conservation, within the SLEUTH (cellular automata modeling) framework

 

Sakieh et al. (2015)

Occurrence of a landslide

Inductive

Logistic regression

None

Bilaşco et al. (2016)

Crop growth

Deductive

Biophysical crop growth model

None

Gasser et al. (2016)

Waterlogging

Deductive

Hydrological model

None

Jiao et al. (2017)

Crop yield

Deductive

Biophysical crop growth model

None

Challinor et al. (2015)

Species observations (presence/absence)

Inductive

Logistic regression

Watanabe–Akaike criteria

Ohashi et al. (2016)

Erosion and soil leaching

Deductive

Physically based model

None

Shahbazi and Jafarzadeh (2010)

Land-use change of a cell

Inductive

CLUE model (logistic regression)

Coefficient of determination (R2), standard error

Mesgari and Jabalamel (2018)

Land use

Deductive

Programmatic rules to resolve conflicts

None

Jeganathan et al. (2011)

Land use

Deductive

Rule based

None

Mesgari and Jabalamel (2018)

*AUC-ROC area under the curve of the receiver operator curve, AIC aikaike information criterion

As can be seen from Table 4, there was a wide range of uses of prediction in the agricultural LUSAs reviewed in this study, including predictions of land-use change, crop growth and yields, species observations (an indicator of habitat change), and physical phenomena, such as hydrological response or landslides. Of the 28 total studies that included predictive activities, half (14) were based on a priori assigned rules, derived from economic or behavioral theory or physical laws, and half (14) were based on empirical–statistical relationships uncovered by the studies themselves.

Model evaluation criteria among inductive predictive activities included the conventional statistical tests of goodness of fit (e.g., AUC, R2, etc.), and only one study evaluated the effects of spatial–temporal stratification on their models (W. Foster et al. 2016, p. 657). Generalizability of conclusions drawn from study samples is especially an issue with empirical–statistical models because of the inherent temporal, spatial, hierarchical, and phylogenic structure within ecological data; it has been shown that traditional statistical methods, even those that use parametric methods to account for spatial autocorrelation and other structures among data points, tend to underestimate errors associated with spatial–temporal data. This is because in addition to non-independence of residuals, overfitting to the dependence structure of data can occur when models absorb variation to the “wrong” predictor (Roberts et al. 2017, p. 915). Model evaluation techniques now commonly used with “big data” datasets; for example, cross-validation techniques that use test and training sets for model fitting do not usually take into account structured data typically found in models for spatial–temporal socio-ecological phenomena. While one reviewed paper included a discussion of spatial stratification based on agroecological zone, none of the papers reviewed implemented blocked cross-validation techniques to ensure properly estimated standard errors and generalizability to new predictive spaces (out of space, out of time). One paper did explicitly specify prior distributions (Ohashi et al. 2016, p. 7767). Capturing uncertainty and the existence of sources of unmodeled structure is important for demonstrating a model’s capability for making predictions for under non-stationary conditions, e.g., future climate predictions. These are ongoing areas of socio-ecological research related to BEOD that were not found to be represented in the articles reviewed in this study.

One reason cross-validation methods were not used in any of the studies may be for lack of data. Although 48% of studies that had predictive activities included the use of BEOD, predictands used in the studies usually represented relatively small data events. In empirically based land-use change models, for example, land-use change represents a small fraction of the overall dataset (most land uses remain unchanged between two points in time); in empirical habitat suitability models, RS data may be used to characterize environments, but the predictand is a much smaller and locally collected species count data (Ohashi et al. 2016, p. 7765; Ohadi et al. 2018, p. 3; Brambilla et al. 2010, p. 2272–2273; Arbeiter et al. 2018, p. 16). It is precisely under circumstances of data scarcity that violations of critical assumptions may make specious relationships appear stronger than they would be had the underlying structure of the data been considered in the statistical testing strategy (Roberts et al. 2017, p. 924).

Model evaluation criteria for deductive socio-ecological predictive activities were more frequently not discussed than for the inductive predictive activities in the literature reviewed. The lack of data for model calibration may be one reason that many deductive models did not report model evaluation criteria. The strength of deductive predictions is that they leverage existing theoretical knowledge on the processes affecting the dynamic being modeled. Theories have typically been accepted based on their applicability across a wide range of conditions and therefore may more successfully make predictions for a wide range of spatial–temporal conditions.

Parameters themselves may be highly uncertain in heterogeneous domains. Deductive predictive models may incorporate uncertainty into their predictions through explicitly specifying a wide distribution of values for input parameters. They may also examine uncertainty through sensitivity analysis, where input parameters are systematically perturbed to reflect a range of potential outcomes for future situations. In one land-use/land-cover change model reviewed, the authors test a range of values of a dispersion parameter around an expert-informed suitability function to assess sensitivity of the model to uncertain parameters (Yu et al. 2011, p. 138). In studies where non-stationarity due to climate change was considered, it was done through four main methods: (1) by varying the temperature and precipitation inputs to a model (either inductively or deductively specified), where the inputs were derived through regional downscaling of GCMs (Antonellini et al. 2014, p. 1842; Gaál et al. 2014, p. 600; Gasser et al. 2016, p. 258); (2) the delta method downscaling of GCM predictions (Ohashi et al. 2016, p. 7768); (3) through the use of “storylines” corresponding to IPCC climate scenarios (Mancosu et al. 2015, p. 28; Cardador et al. 2015, p. 121); or (4) through sensitivity analysis (Challinor et al. 2015, p. 1680).

Of the 29 studies that included predictive activities, 11 of them also included descriptive LUSA. The descriptive LUSA was used in two ways: (1) The descriptive LUSA was as a preprocessing step to the predictive activity, or (2) the predictive activity was used to create a criteria layer that was subsequently included in the descriptive LUSA. Where LUSA was a preprocessing step to the predictive activity, its function was to create likely scenarios for future development, which were then used to generate input parameters to the predictive model (Yu et al. 2018, p. 55). In cases where the predictive activity was used to create a criteria layer, that layer was then either represented as a continuous risk value in the descriptive LUSA (as in Bilaşco et al. 2016, p. 302), or recategorized using informed suitability thresholds (as in Jiao et al. 2017, p. 102).

4.3 Prescriptive activities and LUSA: toward prescriptive processes

In one sense, the ultimate goal toward which LUSA serves is a prescription of what land uses should be implemented and where—a plan, given the values of the stakeholders. In my classification of descriptive, predictive, and prescriptive goals, a “prescriptive” research activity is one that considers trade-offs in objectives, to suggest an optimal solution. In the present literature review, only two studies included prescriptive activities. Benke et al. (2011, p. 93) used a genetic algorithm to allocate land for different kinds of crops maximizing the revenues that could be produced by that allocation. Only one study of those reviewed attempted optimization of multiple land uses using multiple criteria (Liu et al. 2016, pp. 3–5). This study considers maximization of economic benefit, ecological benefit, and social benefit, alongside stakeholder and expert knowledge to prevent land-use fragmentation, as well as incorporate factors related to agricultural land operation, such as farming radius and local agricultural land-use policy. The allocation problem (how much and where) is solved using particle swarm optimization, a heuristic method similar to other genetic algorithms.

Although prescriptive research activities were the least represented type given the search criteria for this review, the approach for prescribing “optimal” land-use configurations given decision criteria is not uncommon. In his classification of GIS-based LUSA techniques, Malczewski (2004) discussed both linear optimization methods (p. 18), and evolutionary and genetic algorithms (pp. 40–42) that can search potential combinatory decision spaces in an effort to prescribe optimal solutions. Spatial optimization of land use for sustainability, for example, is an active area of research (see, for example, Cao et al. 2011) that may not have appeared in this literature review because of my limitation to agricultural LUSA. It was also clear that among the studies reviewed, some had identified prescriptive goals as underlying motivations for their research that were not classified as “prescriptive” because they did not explicitly elaborate on the prescriptive research activity. For example, Meyer et al. (2014, p. 43) developed a stakeholder-driven spatial modeling framework so that the resulting composite LUSA map could be used to better handle conflicting demands of all stakeholders involved and sufficiently differentiated data for policy-makers to understand biophysical and socioeconomic factors affecting land suitability. Here, the “prescription” for what land uses to adopt is arrived at through deliberation of stakeholder values. The contrast between prescriptive goals of the Liu et al. and Cao et al. flavor and the Meyer et al. flavor represents a fundamental difference in the role of technology in the creation of a “prescription” for land use. The former represents a Spatial Experts System (SES) that aims to use artificial intelligence to imitate, extend, and replace the reasoning process of experts in solving spatial problems. The latter represents a decision support system (DSS) that uses artificial intelligence to support users in achieving a better decision. It emphasizes deliberation, communication, and collaboration as part of the decision-making process and creation of the “prescription.”

5 Discussion

In this review of the literature of agricultural LUSA, I discussed the relationship between the traditional McHargian LUSA (descriptive LUSA), to related predictive and prescriptive spatial–temporal research activities. The majority of the agricultural LUSA literature is descriptive in nature, with the goal of spatially identifying areas suitable for growing specific crops. This study was limited to agricultural land uses. This means that the specific application and examples and the proportions of descriptive, predictive, and prescriptive research activities may not be generalizable to other land uses. However, many of the relationships between the research activities are likely to be generalizable. I found that studies that included LUSA did so to simplify the volume and diversity of socio-ecological data representing a particular spatial problem, effectively reducing the multidimensionality of the site to a more easily comprehensible composite map.

In studies that incorporate both descriptive and predictive activities, the descriptive LUSA can be used as a dimension reduction technique to create future land-use scenarios to predict other socio-ecological phenomena, or as a framework to organize the predictions of socio-ecological phenomena with other land suitability criteria. Among the studies reviewed, geophysical factors (such as climate and soil) were more likely to be included in the study than social factors (such as land use, infrastructure accessibility, and demographic variables). Agricultural LUSA is highly related to the LESA method adopted by the USDA NRCD, which evaluated both soil productivity for agriculture and social factors for determining priorities between competing development decisions. The literature reviewed in this study showed that the land evaluation portion of LESA (based on soil productivity) is better represented than the site assessment portion of LESA. Scientific experts, which are also more associated with the land evaluation portion of LESA, were more represented than planners, politicians, economists, or citizens/residents as decision-makers in processes, who are often more associated with the site assessment portion of LESA.

According to the literature reviewed in this study, predictive research activity made it much more likely that LUSA research addressed issues of non-stationarity. This is because (1) addressing non-stationarity requires representation of emergent/dynamic phenomena and potentially high levels of uncertainty that can often not be represented by existing data layers, and (2) spatial–temporal non-stationarity can be better identified and quantified when using modeling techniques common to predictive activities. BEOD was also more likely to be used in research if the study included predictive or prescriptive activity than if the study only included descriptive activity. In fact, studies that included only descriptive activity were most likely to neither include the use of BEOD, nor address non-stationarity within the research, while studies that included some predictive activity were more likely to include both BEOD and non-stationarity than to include neither. However, despite the use of BEOD and advanced modeling techniques there was also evidence that predictive activities included in LUSA often did not report model evaluation criteria that accounted for temporal, spatial, hierarchical, or phylogenic structure, making it difficult to assess how models might generalize over an area being evaluated through descriptive LUSA. This is especially true because despite the use of BEOD in many studies, dependent variables included in studies were often based on local, not global “small data,” including researcher-collected data such as species counts. While BEOD has certainly grown in coverage and resolution, many studies still rely on local data collection, especially for social criteria and variables, such as infrastructure service, local land uses, and demographic variables.

Lastly, the meaning of “prescription” in the context of big data methods is dominated by application of quantitative optimization techniques. The meaning of “prescription” could also be expanded from its conventional meaning to include the processes of arriving at “suitable” land-use distributions that meet the complex requirements of diverse stakeholders. This expanded meaning would include: facilitating communication processes, negotiation, trust, collaboration, and consensus building, which are familiar contexts for planners to engage with the prescription process. Although a comprehensive discussion of the participatory modeling literature is outside the scope of this article, the ways in which LUSA intersect with descriptive, predictive, and prescriptive activities discussed above illustrate how the method can be used effectively as a “boundary object” facilitating communication among diverse stakeholders (Cash et al. 2002); as a tool for collaboratively designing scenarios that facilitate social learning, “bridging” and “stretching” (Xiang and Clarke 2003; Pahl-Wostl et al. 2007); for visual communication (Arciniegas and Janssen 2012); and for the iterative loops necessary when incorporating predictive modeled consequences of spatial or temporal non-stationarity (Voinov and Bousquet 2010; Pourebrahim et al. 2011, Laniak et al. 2013; Grove et al. 2015). Figure 2 shows a conceptual diagram of how LUSA fits into the stages of a collaborative spatial decision-making process.
Fig. 2

LUSA tasks within a collaborative spatial decision-making process

(adapted from Arciniegas and Janssen 2012)

6 Conclusion

This review of agricultural LUSA reveals areas of collaboration between planners and others involved in socio-ecological research. While the focus of McHarg’s original LUSA was on historical and present socio-ecological patterns and mapping, today, consciousness of non-stationary conditions necessitates consideration about how long input criteria will represent reality and whether system dynamics could result in negative or positive feedback loops over time and space. This review suggests that outputs of predictive modeling represented as probabilities, risks, and future pressures specifically to capture uncertainty would be useful to incorporate within descriptive LUSA. Dynamic system models that use BEOD sources and methods to ensure generalizability and account for uncertainty are appropriate tools for addressing non-stationarity. LUSA likewise can be used to generate realistic hypothetical scenarios for predictive modeling to explore variability and sensitivity that incorporate decision-making and perspectives from diverse stakeholders.

References

  1. Akıncı H, Özalp AY, Turgut B (2013) Agricultural land use suitability analysis using GIS and AHP technique. Comput Electron Agric 97(September):71–82.  https://doi.org/10.1016/j.compag.2013.07.006 CrossRefGoogle Scholar
  2. Akinci H, Özalp YA, Özalp M (2017) Investigating impacts of large dams on agricultural lands and determining alternative arable areas using Gis and Ahp in Artvin, Turkey.  https://doi.org/10.15317/Scitech.2017.72
  3. Aktaş E, Türkyılmaz Tahta B (2018) Investigation of suitable land use potential for industrial sites: the case of Kemalpaşa. Environ Monit Assess 190:654.  https://doi.org/10.1007/s10661-018-7007-6 CrossRefGoogle Scholar
  4. Amjath-Babu TS, Kaechele H (2015) Agricultural system transitions in selected Indian states: What do the related indicators say about the underlying biodiversity changes and economic trade-offs? Ecol Ind 57:171–181.  https://doi.org/10.1016/j.ecolind.2015.04.029 CrossRefGoogle Scholar
  5. Antonellini M, Dentinho T, Khattabi A, Masson E, Mollema P, Silva V, Silveira P (2014) An integrated methodology to assess future water resources under land use and climate change: an application to the Tahadart drainage basin (Morocco). Environ Earth Sci 71(4):1839–1853.  https://doi.org/10.1007/s12665-013-2587-5 CrossRefGoogle Scholar
  6. Arbeiter S, Roth T, Helmecke A, Haferland HJ, Tanneberger F, Bellebaum J (2018) Conflict between habitat conservation and corncrake crex crex brood protection in managed floodplain meadows. Agr Ecosyst Environ 265(October):15–21.  https://doi.org/10.1016/j.agee.2018.05.030 CrossRefGoogle Scholar
  7. Arciniegas G, Janssen R (2012) Spatial decision support for collaborative land use planning workshops. Landscape Urban Plan 107(3):332–342.  https://doi.org/10.1016/j.landurbplan.2012.06.004 CrossRefGoogle Scholar
  8. Badmos BK, Villamor GB, Agodzo SK, Guug SS (2014) Examining agricultural land-use/cover change options in rural northern ghana: a participatory scenario exploration exercise approach. Int J Interdiscip Environ Stud 8(2):15–35CrossRefGoogle Scholar
  9. Benke KK, Wyatt RG, Sposito VA (2011) A discrete simulation approach to spatial allocation of commodity production for revenue optimisation over a local region. J Spatial Sci 56(1):89–101.  https://doi.org/10.1080/14498596.2011.567417 CrossRefGoogle Scholar
  10. Berger T (2001) Agent-based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agric Econ 25(2–3):245–260.  https://doi.org/10.1111/j.1574-0862.2001.tb00205.x CrossRefGoogle Scholar
  11. Bilaşco Ş, Roşca S, Păcurar I, Moldovan N, Boţ A, Negruşier C, Sestras P, Bondrea M, Naş S (2016) Identification of land suitability for agricultural use by applying morphometric and risk parameters based on GIS spatial analysis. Notulae Botanicae Horti Agrobotanici Cluj-Napoca 44(1):302–312.  https://doi.org/10.15835/nbha44110289 CrossRefGoogle Scholar
  12. Brambilla M, Casale F, Bergero V, Giuseppe Bogliani G, Crovetto M, Falco R, Roati M, Negri I (2010) Glorious past, uncertain present, bad future? Assessing effects of land-use changes on habitat suitability for a threatened farmland bird species. Biol Cons 143(11):2770–2778.  https://doi.org/10.1016/j.biocon.2010.07.025 CrossRefGoogle Scholar
  13. Cao K, Batty M, Huang B, Liu Y, Le Yu, Chen J (2011) spatial multi-objective land use optimization: extensions to the non-dominated sorting genetic algorithm-II. Int J Geogr Inf Sci 25(12):1949–1969.  https://doi.org/10.1080/13658816.2011.570269 CrossRefGoogle Scholar
  14. Cardador L, De Cáceres M, Giralt D, Bota G, Aquilué N, Arroyo B, Mougeot F et al (2015) Tools for exploring habitat suitability for steppe birds under land use change scenarios. Agr Ecosyst Environ 200:119–125.  https://doi.org/10.1016/j.agee.2014.11.013 CrossRefGoogle Scholar
  15. Cardín-Pedrosa M, Alvarez-López CJ (2012) Model for decision-making in agricultural production planning. Comput Electron Agric 82(March):87–95.  https://doi.org/10.1016/j.compag.2011.12.004 CrossRefGoogle Scholar
  16. Cash D, Clark WC, Alcock F, Dickson NM, Eckley N, Jäger J (2002) Salience, credibility, legitimacy and boundaries: linking research, assessment and decision making. SSRN Scholarly Paper ID 372280. Rochester, NY: Social Science Research Network. https://papers.ssrn.com/abstract=372280
  17. Challinor AJ, Parkes B, Ramirez-Villegas J (2015) Crop Yield response to climate change varies with cropping intensity. Glob Change Biol 21(4):1679–1688.  https://doi.org/10.1111/gcb.12808 CrossRefGoogle Scholar
  18. Chatterji S, Tiwary P, Sen TK, Prasad J, Bhattacharyya T, Sarkar D, Pal DK et al (2014) Land evaluation for major crops in the Indo-Gangetic Plains and black soil regions using fuzzy model. Current Sci 107(9):1502–1511Google Scholar
  19. Collins M, Steiner F, Rushman M (2001) Land-use suitability analysis in the united states: historical development and promising technological achievements. Environ Manag 28(5):611–621CrossRefGoogle Scholar
  20. Conley T, Udry C (2001) Social learning through networks: the adoption of new agricultural technologies in ghana. Am J Agric Econ 83(3):668–673CrossRefGoogle Scholar
  21. Coughlin RE, Pease JR, Steiner F, Papazian L, Pressley JA, Sussman A, Leach JC (1994) The Status of State and Local LESA Programs. J Soil Water Conserv 49(1):6–13Google Scholar
  22. Dengiz O, Ozcan H, Koksal ES, Baskan O, Kosker Y (2010) Sustainable natural resource management and environmental assessment in the Salt Lake (Tuz Golu) Specially Protected Area. Environ Monit Assess 161(1–4):327–342.  https://doi.org/10.1007/s10661-009-0749-4 CrossRefGoogle Scholar
  23. Dung EJ, Sugumaran R (2005) Development of an agricultural land evaluation and site assessment (LESA) decision support tool using remote sensing and geographic information system. J Soil Water Conserv 60(5):228–235Google Scholar
  24. El-Khoury A, Seidou O, Lapen DR, Sunohara M, Zhenyang Q, Mohammadian M, Daneshfar B (2014) Prediction of land-use conversions for use in watershed-scale hydrological modeling: a canadian case study prévision de la réaffectation des sols pour la modélisation hydrologique des bassins versants: une étude de cas canadienne. Can Geogr 58(4):499–516.  https://doi.org/10.1111/cag.12105 CrossRefGoogle Scholar
  25. FAO (1976) A framework for land evaluation. Food and Agriculture Organization, RomeGoogle Scholar
  26. Feder G, Umali DL (1993) The adoption of agricultural innovations: a Review. Technol Forecast Soc Change 43(3):215–239.  https://doi.org/10.1016/0040-1625(93)90053-A CrossRefGoogle Scholar
  27. Fick SE, Hijmans RJ (2017) WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas: new climate surfaces for global land areas. Int J Climatol 37(12):4302–4315.  https://doi.org/10.1002/joc.5086 CrossRefGoogle Scholar
  28. Foster AD, Rosenzweig MR (1995) Learning by doing and learning from others: human capital and technical change in agriculture. J Polit Econ 103(6):1176–1209.  https://doi.org/10.1086/601447 CrossRefGoogle Scholar
  29. Foster W, Anríquez G, Melo O, Yupanqui D, Ortega J (2016) Geographic disparities in rural land appreciation in a transforming economy: Chile, 1980 to 2007. Land Use Policy 57(November):655–668.  https://doi.org/10.1016/j.landusepol.2016.06.025 CrossRefGoogle Scholar
  30. Fotheringham AS (2009) The problem of spatial autocorrelation’ and local spatial statistics: spatial autocorrelation and local spatial statistics. Geograph Anal 41(4):398–403.  https://doi.org/10.1111/j.1538-4632.2009.00767.x CrossRefGoogle Scholar
  31. Fuglie KO, Kascak CA (2001) Adoption and diffusion of natural-resource-conserving agricultural technology. Appl Econ Perspect Policy 23(2):386–403.  https://doi.org/10.1111/1467-9353.00068 CrossRefGoogle Scholar
  32. Gaál M, Quiroga S, Fernandez-Haddad Z (2014) Potential impacts of climate change on agricultural land use suitability of the hungarian counties. Reg Environ Change 14(2):597–610.  https://doi.org/10.1007/s10113-013-0518-3 CrossRefGoogle Scholar
  33. Gasser P-Y, Smith CAS, Brierley JA, Schut PH, Neilsen D, Kenney EA, Yang (2016) The Use of the Land Suitability Rating System to Assess Climate Change Impacts on Corn Production in the Lower Fraser Valley of British Columbia. Can J Soil Sci 96(2):256–269.  https://doi.org/10.1139/cjss-2015-0108 CrossRefGoogle Scholar
  34. Goldstein JH, Caldarone G, Duarte TK, Ennaanay D, Hannahs N, Mendoza G, Polasky S, Wolny S, Daily GC (2012) Integrating ecosystem-service tradeoffs into land-use decisions. Proc Natl Acad Sci 109(19):7565–7570.  https://doi.org/10.1073/pnas.1201040109 CrossRefGoogle Scholar
  35. Griliches Z (1957) Hybrid corn: an exploration in the economics of technological change. Econometrica 25(4):501–522.  https://doi.org/10.2307/1905380 CrossRefGoogle Scholar
  36. Grove JM, Chowdhury R, Childers DL (2015) Co-design, co-production, and dissemination of social-ecological knowledge to promote sustainability and resilience: urban experiences from the U.S. long term ecological research (LTER) network. p. 6Google Scholar
  37. Guo H-D, Li Z, Zhu L-W (2015) Earth observation big data for climate change research. Adv Clim Change Res 6(2):108–117.  https://doi.org/10.1016/j.accre.2015.09.007 CrossRefGoogle Scholar
  38. Hamzeh S, Mokarram M, Alavipanah SK (2014) Combination of Fuzzy and AHP methods to assess land suitability for barley: case study of semi arid lands in the southwest of Iran. Desert 19(2):173–181Google Scholar
  39. Hengl T, Mendes J, de Jesus Gerard B M, Heuvelink MR, Gonzalez MK, Blagotić A, Shangguan W et al (2017) SoilGrids250 m: Global gridded soil information based on machine learning. PLoS ONE 12(2):e0169748.  https://doi.org/10.1371/journal.pone.0169748 CrossRefGoogle Scholar
  40. Heumann BW, Walsh SJ, Verdery AM, McDaniel PM, Rindfuss RR (2013) Land suitability modeling using a geographic socio-environmental niche-based approach: a case study from Northeastern Thailand. Ann Assoc Am Geogr 103(4):764–784.  https://doi.org/10.1080/00045608.2012.702479 CrossRefGoogle Scholar
  41. Hoobler BM, Vance GF, Hamerlinck JD, Munn LC, Hayward JA (2003) applications of land evaluation and site assessment (LESA) and a geographic information system (GIS) in East Park County, Wyoming. J Soil Water Conserv 58(2):105–112Google Scholar
  42. Humphries HC, Bourgeron PS, Reynolds KM (2010) Sensitivity analysis of land unit suitability for conservation using a knowledge-based system. Environ Manag 46(2):225–236.  https://doi.org/10.1007/s00267-010-9520-4 CrossRefGoogle Scholar
  43. Jeganathan C, Roy PS, Jha MN (2011) Multi-objective spatial decision model for land use planning in a tourism district of India. J Environ Inform 17(1):15–24.  https://doi.org/10.3808/jei.201100183 CrossRefGoogle Scholar
  44. Jarasiunas G, Kinderiene I, Bašić F (2017) Delineation Lithuanian agricultural land for agro-ecological suitability for farming using soil and terrain criteria. Ecology 36(1):88–100Google Scholar
  45. Jiao S, Zhang X, Ying X (2017) A review of Chinese land suitability assessment from the rainfall-waterlogging perspective: evidence from the Su Yu Yuan Area. J Clean Prod 144(February):100–106.  https://doi.org/10.1016/j.jclepro.2016.12.162 CrossRefGoogle Scholar
  46. Karl TR, Trenberth KE (2003) Modern global climate change. Science 302(5651):1719–1723.  https://doi.org/10.1126/science.1090228 CrossRefGoogle Scholar
  47. Kassie M, Jaleta M, Shiferaw B, Mmbando F, Mekuria M (2013) Adoption of interrelated sustainable agricultural practices in smallholder systems: evidence from rural Tanzania. Technol Forecast Soc Change Future-Oriented Technol Anal 80(3):525–540.  https://doi.org/10.1016/j.techfore.2012.08.007 CrossRefGoogle Scholar
  48. Kazemi H, Akinci H (2018) A land use suitability model for rainfed farming by Multi-criteria Decision-making Analysis (MCDA) and Geographic Information System (GIS)(GIS). Ecol Eng 116:1–6.  https://doi.org/10.1016/j.ecoleng.2018.02.021 CrossRefGoogle Scholar
  49. Kazemi H, Sadeghi S, Akinci H (2016) Developing a land evaluation model for faba bean cultivation using geographic information system and multi-criteria analysis (A case study: Gonbad-Kavous region, Iran). Ecol Indic 63:37–47.  https://doi.org/10.1016/j.ecolind.2015.11.021 CrossRefGoogle Scholar
  50. Kellner O, Niyogi D (2015) Climate variability and the U.S. Corn Belt: ENSO and AO episode-dependent hydroclimatic feedbacks to corn production at regional and local scales*. Earth Interact 19(6):1–32.  https://doi.org/10.1175/EI-D-14-0031.1 CrossRefGoogle Scholar
  51. Kerr JT, Ostrovsky M (2003) From space to species: ecological applications for remote sensing. Trends Ecol Evol 18(6):299–305.  https://doi.org/10.1016/S0169-5347(03)00071-5 CrossRefGoogle Scholar
  52. Kiliç Ş (2011) Agroecological land use potential of Amik Plain, Turkey. Turk J Agric For 35(4):433–442.  https://doi.org/10.3906/tar-1007-940 CrossRefGoogle Scholar
  53. Kravchenko AN, Bullock DG (2000) Correlation of corn and soybean grain yield with topography and soil properties. Agron J 92(1):75–83.  https://doi.org/10.2134/agronj2000.92175x CrossRefGoogle Scholar
  54. Lambin EF, Rounsevell MDA, Geist HJ (2000) Are agricultural land-use models able to predict changes in land-use intensity? Agr Ecosyst Environ 82(1–3):321–331.  https://doi.org/10.1016/S0167-8809(00)00235-8 CrossRefGoogle Scholar
  55. Laniak GF, Olchin G, Goodall J, Voinov A, Hill M, Glynn P, Whelan G et al (2013) Integrated environmental modeling: a vision and roadmap for the future. Environ Model Softw 39(January):3–23.  https://doi.org/10.1016/j.envsoft.2012.09.006 CrossRefGoogle Scholar
  56. Li M, JunJie W, Deng X (2013) Identifying drivers of land use change in China: a spatial multinomial logit model analysis. Land Econ 89(4):632–654.  https://doi.org/10.3368/le.89.4.632 CrossRefGoogle Scholar
  57. Liu Y, Peng J, Jiao L, Liu Y (2016) PSOLA: a heuristic land-use allocation model using patch-level operations and knowledge-informed rules. PLoS ONE 11(6):1–21.  https://doi.org/10.1371/journal.pone.0157728 CrossRefGoogle Scholar
  58. Lobell DB, Bonfils C (2008) The effect of irrigation on regional temperatures: a spatial and temporal analysis of trends in California, 1934–2002. J Clim 21(10):2063–2071.  https://doi.org/10.1175/2007JCLI1755.1 CrossRefGoogle Scholar
  59. Lockie S, Mead A, Vanclay F, Butler B (1995) Factors encouraging the adoption of more sustainable crop rotations in South-East Australia. J Sustain Agric 6(1):61–79.  https://doi.org/10.1300/J064v06n01_06 CrossRefGoogle Scholar
  60. Magliocca NR, Brown DG, Ellis EC (2013) Exploring agricultural livelihood transitions with an agent-based virtual laboratory: global forces to local decision-making. PLoS ONE 8(9):1–11.  https://doi.org/10.1371/journal.pone.0073241 CrossRefGoogle Scholar
  61. Malczewski J (2004) GIS-based land-use suitability analysis: a critical overview. Prog Plan 63:3–65CrossRefGoogle Scholar
  62. Maleki F, Kazemi H, Siahmarguee A, Kamkar B (2017) Development of a land use suitability model for saffron (Crocus Sativus L.) cultivation by multi-criteria evaluation and spatial analysis. Ecol Eng 106(September):140–153.  https://doi.org/10.1016/j.ecoleng.2017.05.050 CrossRefGoogle Scholar
  63. Mancosu E, Gago-Silva A, Barbosa A, de Bono A, Ivanov E, Lehmann A, Fons J (2015) Future Land-use change scenarios for the black sea catchment. Environ Sci Policy 46(February):26–36.  https://doi.org/10.1016/j.envsci.2014.02.008 CrossRefGoogle Scholar
  64. Mbũgwa G, Prager Steven D, Krall James M (2015) Utilization of spatial decision support systems decision-making in dryland agriculture: a tifton burclover case study. Comput Electron Agric 118(October):215–224.  https://doi.org/10.1016/j.compag.2015.09.008 CrossRefGoogle Scholar
  65. McDowell RW, Snelder T, Harris S, Lilburne L, Larned ST, Scarsbrook M, Curtis A, Holgate B, Phillips J, Taylor K (2018) The land use suitability concept: introduction and an application of the concept to inform sustainable productivity within environmental constraints. Ecol Ind 91(August):212–219.  https://doi.org/10.1016/j.ecolind.2018.03.067 CrossRefGoogle Scholar
  66. McHarg I (1969) Design with nature. Natural History Press, Garden CityGoogle Scholar
  67. McHarg IL (2014) Open space from natural processes. In: Ndubisi FO (ed) The ecological design and planning reader. Island Press/Center for Resource Economics, Washington, pp 181–90.  https://doi.org/10.5822/978-1-61091-491-8_18
  68. Memarbashi E, Azadi H, Barati A, Mohajeri F, Passel S, Witlox F (2017) Land-Use suitability in northeast Iran: application of AHP-GIS hybrid model. ISPRS Int J Geo-Inf 6(12):396.  https://doi.org/10.3390/ijgi6120396 CrossRefGoogle Scholar
  69. Mesgari I, Jabalameli MS (2018) An integrated and dynamic approach to agricultural land-use change modeling at country-level to regional scale: application to Iran. Syst Eng 21(1):16–29.  https://doi.org/10.1002/sys.21411 CrossRefGoogle Scholar
  70. Meyer SR, Johnson ML, Lilieholm RJ, Cronan CS (2014) Development of a stakeholder-driven spatial modeling framework for strategic landscape planning using bayesian networks across two urban-rural gradients in maine, USA. Ecol Model 291(November):42–57.  https://doi.org/10.1016/j.ecolmodel.2014.06.023 CrossRefGoogle Scholar
  71. Montgomery B, Dragićević S, Dujmović J, Schmidt M (2016) A GIS-based logic scoring of preference method for evaluation of land capability and suitability for agriculture. Comput Electron Agric 124(June):340–353.  https://doi.org/10.1016/j.compag.2016.04.013 CrossRefGoogle Scholar
  72. Mousavi SR, Sarmadian F, Alijani Z, Taati A (2017) Land suitability evaluation for irrigating wheat by Geopedological approach and Geographic Information System: a case study of Qazvin plain, Iran. Eurasian J Soil Sci 6(3):275.  https://doi.org/10.18393/ejss.297251 CrossRefGoogle Scholar
  73. Mueller L, Schindler U, Wilfried Mirschel T, Shepherd G, Ball BC, Helming K, Rogasik J, Eulenstein F, Wiggering H (2010) Assessing the productivity function of soils. A review. Agron Sustain Dev 30(3):601–614.  https://doi.org/10.1051/agro/2009057 CrossRefGoogle Scholar
  74. Mueller ND, Butler EE, McKinnon KA, Rhines A, Martin Tingley N, Holbrook M, Huybers P (2016) Cooling of US midwest summer temperature extremes from cropland intensification. Nat Clim Change 6(3):317–322.  https://doi.org/10.1038/nclimate2825 CrossRefGoogle Scholar
  75. Naughton CC, Lovett PN, Mihelcic JR (2015) Land suitability modeling of shea (Vitellaria paradoxa) distribution across sub-Saharan Africa. Appl Geogr 58(March):217–227.  https://doi.org/10.1016/j.apgeog.2015.02.007 CrossRefGoogle Scholar
  76. Nguyen TT, Verdoodt A, Van Tran Y, Delbecque N, Tran TC, Van Ranst E (2015) Design of a GIS and Multi-criteria based land evaluation procedure for sustainable land-use planning at the regional level. Agric Ecosyst Environ 200(February):1–11.  https://doi.org/10.1016/j.agee.2014.10.015 CrossRefGoogle Scholar
  77. Ohadi S, Littlejohn M, Mesgaran M, Rooney W, Bagavathiannan M (2018) Surveying the spatial distribution of feral sorghum (Sorghum Bicolor L.) and Its sympatry with johnsongrass (S. Halepense) in south texas. PLoS ONE 13(4):1–14.  https://doi.org/10.1371/journal.pone.0195511 CrossRefGoogle Scholar
  78. Ohashi H, Kominami Y, Higa M, Koide D, Nakao K, Tsuyama I, Matsui T, Tanaka N (2016) Land abandonment and changes in snow cover period accelerate range expansions of sika deer. Ecol Evol 6(21):7763–7775.  https://doi.org/10.1002/ece3.2514 CrossRefGoogle Scholar
  79. Overmars KP, Verburg PH, Veldkamp. T (2007) Comparison of a deductive and an inductive approach to specify land suitability in a spatially explicit land use model. Land Use Policy 24(3):584–599.  https://doi.org/10.1016/j.landusepol.2005.09.008 CrossRefGoogle Scholar
  80. Pahl-Wostl C, Craps M, Dewulf A, Mostert E, Tabara D, Taillieu T (2007) Social learning and water resources management. Ecol Soc 12(2). http://repository.tudelft.nl/islandora/object/uuid:9bcb1311-74ac-4400-88e2-e6b816397dfd?collection=research
  81. Pimentel D, Harvey C, Resosudarmo P, Sinclair K, Kurz D, McNair M, Crist S et al (1995) Environmental and Economic Costs of Soil Erosion and Conservation Benefits. Science 267(5201):1117–1123.  https://doi.org/10.1126/science.267.5201.1117 CrossRefGoogle Scholar
  82. Pourebrahim S, Hadipour M, Mokhtar MB (2011) Integration of spatial suitability analysis for land use planning in coastal areas; case of Kuala Langat district, Selangor, Malaysia. Landsc Urban Plan 101(1):84–97.  https://doi.org/10.1016/j.landurbplan.2011.01.007 CrossRefGoogle Scholar
  83. Raza SMH, Mahmood SA, Khan AA, Liesenberg V (2018) Delineation of potential sites for rice cultivation through multi-criteria evaluation (MCE) using remote sensing and GIS. Int J Plant Prod 12(1):1–11.  https://doi.org/10.1007/s42106-017-0001-z CrossRefGoogle Scholar
  84. Rhebergen T, Fairhurst T, Zingore S, Fisher M, Oberthür T, Whitbread A (2016) Climate, soil and land-use based land suitability evaluation for oil palm production in Ghana. Eur J Agron 81(November):1–14.  https://doi.org/10.1016/j.eja.2016.08.004 CrossRefGoogle Scholar
  85. Richards P (2018) It’s not just where you farm; it’s whether your neighbor does too: How agglomeration economies are shaping new agricultural landscapes. J Econ Geogr 18(1):87–110CrossRefGoogle Scholar
  86. Roberts MJ, Schlenker W, Eyer J (2013) Agronomic weather measures in econometric models of crop yield with implications for climate change. Am J Agric Econ 95(2):236–243.  https://doi.org/10.1093/ajae/aas047 CrossRefGoogle Scholar
  87. Roberts DR, Bahn V, Ciuti S, Boyce MS, Elith J, Guillera-Arroita G, Hauenstein S et al (2017) Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 40(8):913–929.  https://doi.org/10.1111/ecog.02881 CrossRefGoogle Scholar
  88. Ruan X, Qiu F, Dyck M (2016) The effects of environmental and socioeconomic factors on land-use changes: a study of Alberta, Canada. Environ Monit Assess 188(8):1–31.  https://doi.org/10.1007/s10661-016-5450-9 CrossRefGoogle Scholar
  89. Sahoo S, Sil I, Dhar A, Debsarkar A, Das P, Kar A (2018) Future scenarios of land-use suitability modeling for agricultural sustainability in a river basin. J Clean Prod 205(December):313–328.  https://doi.org/10.1016/j.jclepro.2018.09.099 CrossRefGoogle Scholar
  90. Sakieh Y, Salmanmahiny A, Jafarnezhad J, Mehri A, Kamyab H, Galdavi S (2015) Evaluating the strategy of decentralized urban land-use planning in a developing region. Land Use Policy 48(November):534–551.  https://doi.org/10.1016/j.landusepol.2015.07.004 CrossRefGoogle Scholar
  91. Sati VP, Wei D (2018) Crop productivity and suitability analysis for land-use planning in Himalayan ecosystem of Uttarakhand, India. Curr Sci 115(4):767–772.  https://doi.org/10.18520/cs/v115/i4/767-772 CrossRefGoogle Scholar
  92. Schlenker W, Roberts MJ (2009) Nonlinear temperature effects indicate severe damages to U.S. crop yields under climate change. Proc Natl Acad Sci 106(37):15594–15598.  https://doi.org/10.1073/pnas.0906865106 CrossRefGoogle Scholar
  93. Shahbazi F, Jafarzadeh AA (2010) Integrated assessment of rural lands for sustainable development using MicroLEIS DSS in West Azerbaijan, Iran. Geoderma 157(3–4):175–184.  https://doi.org/10.1016/j.geoderma.2010.04.010 CrossRefGoogle Scholar
  94. Sharma R, Kamble SS, Gunasekaran A (2018) Big GIS analytics framework for agriculture supply chains: a literature review identifying the current trends and future perspectives. Comput Electron Agric 155(December):103–120.  https://doi.org/10.1016/j.compag.2018.10.001 CrossRefGoogle Scholar
  95. Singha C, Swain KC (2018) Soil profile based land suitability study for jute and lentil using AHP ranking. Int J Bio-Res Stress Manage 9(3):323–329.  https://doi.org/10.23910/IJBSM/2018.9.3.1869 CrossRefGoogle Scholar
  96. Steiner F, Dunford R, Dosdall N (1987) The use of the agricultural land evaluation and site assessment system in the United States. Landscape and Urban Planning 14(January):183–199.  https://doi.org/10.1016/0169-2046(87)90028-4 CrossRefGoogle Scholar
  97. Steiner FR, Pease JR, Coughlin RE (eds) (1994) A decade with Lesa: the evolution of land evaluation and site assessment. Soil & Water Conservation Society, AnkenyGoogle Scholar
  98. Tomlin CD (1990) Geographic information systems and cartographic modeling. Prentice Hall, Upper Saddle RiverGoogle Scholar
  99. Ullah KM, Mansourian A (2016) Evaluation of land suitability for urban land-use planning: case study Dhaka City. Trans GIS 20(1):20–37.  https://doi.org/10.1111/tgis.12137 CrossRefGoogle Scholar
  100. UNGWG (2017) Earth observations for official statistics: satellite imagery and geospatial data Tast Team Report. United Nations Global Working Group, QueenslandGoogle Scholar
  101. U.S. Department of Agriculture (1961) Land capability classification. Handbook 210. USDA Soil Conservation Service, Washington DCGoogle Scholar
  102. U.S. Department of Agriculture (2018) Summary report: 2015 national resources inventory, natural resources conservation service, Washington, DC, and Center for Survey Statistics and Methodology, Iowa State University, Ames, Iowa. http://www.nrcs.usda.gov/technical/nri/15summary
  103. Ustaoglu E, Perpiña Castillo C, Jacobs-Crisioni C, Lavalle C (2016) Economic evaluation of agricultural land to assess land use changes. Land Use Policy 56(November):125–146.  https://doi.org/10.1016/j.landusepol.2016.04.020 CrossRefGoogle Scholar
  104. Vasu D, Srivastava R, Patil NG, Tiwary P, Chandran P, Singh SK (2018) A comparative assessment of land suitability evaluation methods for agricultural land use planning at village level. Land Use Policy 79:146–163.  https://doi.org/10.1016/j.landusepol.2018.08.007 CrossRefGoogle Scholar
  105. Vettorazzi CA, Valente RA (2016) Priority areas for forest restoration aiming at the conservation of water resources. Ecol Eng 94:255–267.  https://doi.org/10.1016/j.ecoleng.2016.05.069 CrossRefGoogle Scholar
  106. van Wart J, van Bussel LG, Wolf J, Licker R, Grassini P, Nelson A, Boogaard H et al (2013) Use of agro-climatic zones to upscale simulated crop yield potential. Field Crops Res 143(March):44–55.  https://doi.org/10.1016/j.fcr.2012.11.023 CrossRefGoogle Scholar
  107. Voinov A, Bousquet F (2010) Modelling with stakeholders. Environ Model Softw 25(11):1268–1281.  https://doi.org/10.1016/j.envsoft.2010.03.007 CrossRefGoogle Scholar
  108. Woodcock CE, Allen R, Ansderson M, Belward A, Bindschadler R, Cohen W, Gao F et al (2008) Free access to landsat imagery. Science 320:1011CrossRefGoogle Scholar
  109. Wright LE, Zitzmann W, Young K, Googins R (1983) LESA—agricultural land evaluation and site assessment. J Soil Water Conserv 38(2):82–86Google Scholar
  110. Wulder MA, White JC, Loveland TR, Woodcock CE, Belward AS, Cohen WB, Fosnight EA, Shaw J, Masek JG, Roy DP (2016) The global landsat archive: status, consolidation, and direction. Remote Sens Environ 185(November):271–283.  https://doi.org/10.1016/j.rse.2015.11.032 CrossRefGoogle Scholar
  111. Xiang W-N, Clarke KC (2003) The use of scenarios in land-use planning. Environ Plan 30(6):885–909.  https://doi.org/10.1068/b2945 CrossRefGoogle Scholar
  112. Yalew SG, van Griensven A, van der Zaag P (2016) AgriSuit: a web-based GIS-MCDA framework for agricultural land suitability assessment. Comput Electron Agric 128:1–8.  https://doi.org/10.1016/j.compag.2016.08.008 CrossRefGoogle Scholar
  113. You L, Wood S, Wood-Sichra U (2009) Generating plausible crop distribution maps for sub-saharan africa using a spatially disaggregated data fusion and optimization approach. Agric Syst 99(2–3):126–140.  https://doi.org/10.1016/j.agsy.2008.11.003 CrossRefGoogle Scholar
  114. Yu J, Chen Y, Jianping W, Khan S (2011) Cellular Automata-based spatial multi-criteria land suitability simulation for irrigated agriculture. Int J Geogr Inf Sci 25(1):131–148.  https://doi.org/10.1080/13658811003785571 CrossRefGoogle Scholar
  115. Yu D, Xie P, Dong X, Bob S, Xiaonong H, Wang K, Shijin X (2018) The development of land use planning scenarios based on land suitability and its influences on eco-hydrological responses in the upstream of the Huaihe River Basin. Ecol Model 373(April):53–67.  https://doi.org/10.1016/j.ecolmodel.2018.01.010 CrossRefGoogle Scholar
  116. Zabihi H, Ahmad A, Vogeler I, Said MN, Golmohammadi M, Golein B, Nilashi M (2015) Land suitability procedure for sustainable citrus planning using the application of the analytical network process approach and GIS. Comput Electron Agr 117:114–126.  https://doi.org/10.1016/j.compag.2015.07.014 CrossRefGoogle Scholar
  117. Ziadat FM, Sultan KA (2011) Combining current land use and farmers’ knowledge to design land-use requirements and improve land suitability evaluation. Renewable Agric Food Syst 26(04):287–296.  https://doi.org/10.1017/S1742170511000093 CrossRefGoogle Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2019

Authors and Affiliations

  1. 1.Department of Urban Affairs and Planning, School of Policy and International AffairsVirginia TechBlacksburgUSA

Personalised recommendations