RESEARCH ARTICLE10.1002/2016MS000721

Projecting regional climate and cropland changes using alinked biogeophysical-socioeconomic modeling framework:2. Transient dynamicsKazi Farzan Ahmed1 , Guiling Wang1 , Liangzhi You2,3 , Richard Anyah1,4, Chuanrong Zhang5,and Amy Burnicki1

1Department of Civil and Environmental Engineering, and Center for Environmental Science and Engineering, Universityof Connecticut, Storrs, Connecticut, USA, 2Key Laboratory of Agri-informatics, Ministry of Agriculture/Institute ofAgricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing, China, 3InternationalFood Policy Research Institute, Washington, District of Columbia, USA, 4Department of Natural Resources and theEnvironment, University of Connecticut, Storrs, Connecticut, USA, 5Department of Geography, University of Connecticut,Storrs, CT, USA

Abstract Understanding climate-cropland interactions and their impact on future projections in WestAfrica motivated the recent development of a modeling framework that asynchronously couples fourmodels for regional climate, crop growth, socioeconomics, and cropland allocation. This modelingframework can be applied to a future time slice using an equilibrium approach or to a continuous projectionusing a transient approach. This paper compares the differences between these two approaches, examinesthe transient dynamics of the system, and evaluates its impact on future projections. During the course ofprojection up to mid-century, food demand is projected to increase monotonically, while the projectedcrop yield shows a high degree of temporal dynamics due to strong climate variability. Such temporaldynamics are not accounted for by the equilibrium approach. As a result, the transient approach projects agenerally faster future expansion of cropland, with the largest differences over Benin, Burkina Faso, Ghana,Senegal, and Togo. Despite the relative large differences between the two approaches in projecting landcover changes associated with cropland expansion, the projected future climate changes are fairly similar.While the additional cropland expansion in the transient approach favors a wet signal, both the transientand equilibrium approaches project a future decrease of rainfall in the western part of West Africa and anincrease in the eastern part. For quantifying climate changes, the equilibrium application of the modelingframework is likely to be sufficient; for assessing climate impact on agricultural sectors and devisingmitigation and adaptation strategies, transient dynamics is important.

1. Introduction

Anthropogenic land use and land cover changes (LULCC) affect regional or local climate through alterationsof both biogeophysical and biogeochemical processes involved in land-atmosphere interactions. Human-induced modifications of the physical land surface properties lead to changes in surface albedo and rough-ness length, partitioning between sensible and latent heat fluxes, compositions of atmospheric greenhousegases, and other key components of water, energy and carbon cycles, which in turn alter the existing climat-ic patterns [Claussen et al., 2001; Pitman et al., 2009; Pongratz et al., 2010; Sylla et al., 2016; Wang et al., 2016].While assessments of the effect of anthropogenic land use are critically important in climate change impactstudies, uncertainties are rife in quantifying the response of climate variables to land cover changes [Pitmanet al., 2009; Brovkin et al., 2013; Frieler et al., 2015]. It is challenging to understand and quantify the climateresponse to anthropogenic LULCC in the context of climate projections for two primary reasons. First, pro-jection of future land use patterns, especially on the large scale, is subject to a high degree of uncertaintiesrelated to human decision making [Rounsevell et al., 2014; Ahmed et al., 2016]. Second, because of the coun-teracting effects of various underlying processes with large regional variability, no clear directionality isexpected from the response of some climatic variables to LULCC. For example, while conversion of foreststo croplands may lead to warming when/where reduction in evapotranspiration is dominant [Brovkin et al.,

This article is a companion to Wang etal. [2017], doi:10.1002/2016MS000712.

Key Points:� Transient approach to climate-land

use projection suggests fastercropland expansion than equilibriumapproach� Transient dynamics favors a weak

wet signal� Transient dynamics is important for

impact assessment but not essentialfor climate projection

Correspondence to:G. Wang,guiling.wang@uconn.edu;L. You,youliangzhi@caas.cn

Citation:Ahmed, K. F., G. Wang, L. You,R. Anyah, C. Zhang, and A. Burnicki(2017), Projecting regional climate andcropland changes using a linkedbiogeophysical-socioeconomicmodeling framework: 2. Transientdynamics, J. Adv. Model. Earth Syst., 9,377–388, doi:10.1002/2016MS000721.

Received 24 MAY 2016

Accepted 4 JAN 2017

Accepted article online 7 JAN 2017

Published online 7 FEB 2017

VC 2017. The Authors.

This is an open access article under the

terms of the Creative Commons

Attribution-NonCommercial-NoDerivs

License, which permits use and

distribution in any medium, provided

the original work is properly cited, the

use is non-commercial and no

modifications or adaptations are

made.

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 377

Journal of Advances in Modeling Earth Systems

PUBLICATIONS

2009; Sylla et al., 2016], it may cause cooling when/where increase of surface albedo is dominant [Claussenet al., 2001; Lee et al., 2011]. The interaction between climate and LULCC needs to be evaluated comprehen-sively in regional climate projection.

Land-atmosphere coupling is strong in most of West Africa [Koster et al., 2004]. Potential impacts of landcover changes, both natural and anthropogenic, on the regional climate in West Africa have been a topic ofextensive research [Taylor et al., 2002; Hagos et al., 2014; Wang et al., 2016]. The West Africa climate, precipi-tation in particular, features strong inter-annual and inter-decadal variability including a persistent droughtin the Sahel during the second half of the 20th century. Many studies investigating this drought suggestedthat natural and anthropogenic land cover changes played a critical role [Zeng et al., 1999; Wang and Eltahir,2000; Xue et al., 2010]. Changes in land surface characteristics, both natural and human-induced, mainlyaffect the dynamics of the West African monsoon leading to a shift in precipitation pattern. Charney et al.[1975] proposed that the conversion of forest to bare soil perturbs the albedo gradient and increases atmo-spheric subsidence weakening the monsoon precipitation. Taylor et al. [2002], using a General CirculationModel (GCM) and a land use model, suggested that land use changes characterized by agricultural extensifi-cation and deforestation would delay the monsoon onset and thus reduce the overall precipitation. Numer-ous studies on this topic, mostly based on numerical model simulations, all demonstrate the strongsensitivity of West African Climate to LULCC, but the extent of the climate response involves large uncer-tainties, mainly resulting from the differences in methods of prescribing LULCC.

Agricultural land use represents one of the main factors responsible for anthropogenic LULCC. Globally,although intensive farming adopted by farmers in last few decades has slowed down the rate of crop areaexpansion, fraction of agricultural land use has still been increasing [Burney et al., 2010; Hurtt et al., 2011].This is especially the case in developing regions with poor socioeconomic infrastructure where expansionof crop area at the expense of natural vegetation (e.g., subsistence farming such as slash-and-burn practice)is a common practice. Therefore, with constantly increasing food demand across the globe, the trend ofagricultural land use should be carefully considered in analyzing and projecting regional climate change.Despite the crucial link between climate and LULCC in a region, the mechanisms of anthropogenic landuse, instead of being directly incorporated, are usually represented as an external forcing in climate models[Pielke et al., 2011; Rounsevell et al., 2014], and biogeophysical impacts of the land use change dynamics areoften ignored in future climate projection studies.

Crop productivity, which is strongly influenced by climate, represents an intrinsic link between climate andanthropogenic land use. The climate-induced crop yield loss, in addition to the rapidly increasing fooddemand in many regions, will be an important driver for future LULCC under future climate scenarios. Pro-jection of future yield under climate change scenarios needs to be accounted for explicitly in the modelingof future land use and land cover changes. Since different crops could respond differently to the samechanges of climate, species variability of crop yield also needs to be examined to better understand theland use dynamics. Therefore, comprehensive analysis of crop response to regional climate changes shouldbe included while investigating future land use changes and the resulting feedback to regional climate. Pre-vious studies usually prescribed changes in crop area and other land use types to examine the climate sen-sitivity to agricultural land use in West Africa. However, to our knowledge, no previous studies projectingregional climate change in West Africa directly addressed the climate change impact on crop yield in evalu-ating land use-climate interaction in the region.

Our main objective is to understand and project the response of the West African climate to future landuse changes, and to project cropland expansion in the region considering climate-induced crop yieldchanges as one of the key drivers for agricultural land use change (in addition to the future increase offood demand). To model the interactions between regional climate and agricultural land use, wedesigned a comprehensive modeling framework that asynchronously couples a process-based cropmodel, a socioeconomic model and a cropland allocation model with a regional climate model [Wanget al., 2017]. This modeling framework was used in an equilibrium mode to project regional climatechange and cropland expansion in West Africa for the mid-century in a companion paper [Wang et al.,2017]. This current study makes use of the same modeling framework but takes a transient approach,and focuses on the transient dynamics in regional climate and land use and how they influence modelprojections.

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 378

2. Models and Methodology

2.1. The Asynchronously Coupled Modeling FrameworkThis study makes use of the asynchronous coupled modeling framework of Wang et al. [2017] that linksfour different models: a regional climate model, a crop growth model, a socioeconomic model, and a crop-land projection model. The regional climate model [Wang et al., 2015] results from the coupling of the Inter-national Center for Theoretical Physics (ICTP) regional climate model (RegCM) version 4.3.4 [Giorgi et al.,2012] with the Community Land Model (CLM) version 4.5 [Oleson et al., 2013]. The model includes multipleoptions of convection scheme, and the MIT-Emanuel cumulus convection scheme [Emanuel, 1991] was cho-sen for use in this study to optimize performance in capturing the present-day climate in West Africa. Forfuture projections, the model was driven with initial and boundary conditions from the CMIP5 RCP8.5 simu-lations of the Community Earth System Model (CESM) CCSM4 version. Output from the regional climatemodel was resampled to 0.58 resolution and corrected for model bias [Ahmed et al., 2013; Wang et al., 2017]before it was used as input to the crop growth model DSSAT.

The process-based crop model DSSAT [Jones et al., 2003] integrates crop physiology and phenotype, weath-er and soil data, and crop management strategies to simulate crop yield. Five major crops in West Africa,namely maize, millet, sorghum, peanut, and cassava, are included in this study. These five crops occupyapproximately 80% of cultivated area in West Africa. DSSAT was calibrated to simulate future yield for cerealcrops at a spatial resolution of 0.5˚ following the methodology of Ahmed et al. [2015]. For cassava and pea-nut, DSSAT calibration was not as satisfactory. Instead, to correct the model biases, the DSSAT-projectedfuture yield values of cassava and peanut were scaled by the ratio of country-level present-day observedyield to model yield. For each crop, the planting month is chosen based on the typical monsoon onset time,while the exact planting date is determined by the model to reflect the time when soil moisture reaches aspecified threshold within the planting month. Therefore, as climate forcing varies from year to year, theplanting date also varies accordingly. The simulated yearly yield values averaged over each 5 year iterationperiod were first adjusted using factors that reflect the DSSAT model bias, weight of minor crops, and preva-lence of mixed farming practices, and were then provided as inputs to the cropland projection model Land-Pro_Crop to develop cropland expansion scenarios at the end of the specific 5 year period.

LandPro_Crop develops cropland expansion scenarios based on the supply deficit for each crop. Specifi-cally, it calculates the gap between trade-adjusted demand for each crop in the next 5 years and the pro-jected supply from the present-day harvest area at the projected level of crop yield, and allocates naturallyvegetated land for conversion to cropland to close the gap following a set of rules or assumptions [Ahmedet al., 2016]. The resulting new land use land cover distribution will be used to update the surface boundaryconditions for the regional climate model RegCM-CLM. Here, the country-averaged crop-level food demandand international trade were projected by the International Model for Policy Analysis of Agricultural Com-modities and Trade (IMPACT) [Rosegrant et al., 2012]. In this study, the IMPACT was run under the SharedSocioeconomic Pathway-2 (SSP2), a moderate pathway characterized by historical trends of economicdevelopment and medium population growth, and driven by RCP8.5 future climate from four global climatemodels (GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR, and MIROC-ESM). The average of the four projectionswas used as input to the LandPro_Crop model. Further details about the modeling framework can be foundin the companion paper [Wang et al., 2017].

2.2. Experimental DesignThe asynchronously coupled modeling framework described above can be implemented in a transientmode with an iteration period ranging from one to several years, or in an equilibrium mode (with an itera-tion period of several decades as in Wang et al. [2017]). The description in section 2.1 here pertains to atransient application with a 5 year iteration period, which is the focus of this paper. To examine the differ-ences in future projections between the equilibrium and transient approaches, and to elucidate the tran-sient dynamics of the coupled climate-agriculture system, two experiments of future climate projectionsaccounting for future land use changes were conducted here: FUTURE_EQ_LUC using the equilibriumapproach and FUTURE_TR_LUC using the transient approach. In addition, a future projection experiment(FUTURE_NO_LUC) was also conducted that does not consider the impact of potential land cover changes.

Experiment FUTURE_TR_LUC is a transient run from 2005 to 2050, in which LandPro_Crop projected tran-sient land use patterns every 5 years using the socioeconomic and crop yield data under bias-corrected

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 379

climate scenarios from RegCM-CLM, while RegCM-CLM dynamically downscaled future CESM climate withland use land cover updated every five years according to output from LandPro_Crop. Specifically, theCESM outputs were downscaled to 50 km in RegCM-CLM, which is then resampled to a 0.58 grid and cor-rected for model bias before being used as input to DSSAT to project crop yield for each year during a given5 year period; the 5 year averages of these crop yields, together with the demand for crop production pro-jected by IMPACT, are then provided to LandPro_Crop to project the cropland expansion and the associatedloss of natural vegetation cover; the updated land use land cover map was then used to drive the RegCM-CLM model for the next 5 year iteration. Bias correction of the dynamically downscaled climates follows theStatistical Downscaling and Bias Correction (SDBC) method [Ahmed et al., 2013] using the Sheffield et al.[2006] data as a present-day reference.

The FUTURE_EQ_LUC experiment takes the equilibrium approach described in Wang et al. [2016] in projec-ting future cropland distribution and future climate. It starts with the FUTURE_NO_LUC experiment, inwhich RegCM-CLM is driven with the present-day vegetation distribution to project future climate for 2040–2050 in West Africa without accounting for land use changes and their resulting feedback to the regionalclimate. The bias-corrected future climate from the FUTURE_NO_LUC experiment is used to drive the cropmodel DSSAT to project the future crop yield for 2041–2050, which is then used as input to LandPro_Cropto derive land use land cover distribution for the mid-century. This updated land use land cover map isthen used to drive a RegCM-CLM rerun for the period 2040–2050 for the mid-century climate projection.Similar to the FUTURE_TR_LUC experiment, future climate scenarios from the FUTURE_EQ_LUC experimentalso highlight the significance of incorporating information on future LULCC in projecting future climate.However, in the latter, the transient trends in agricultural land use in the region were not captured sincethe models were run only for one particular future time slice. Comparison between these two experimentswould indicate how and to what extent accounting for the transient processes of land use-climate couplingmay influence the outcome of the projections.

To define future changes, results from these experiments were compared to a present-day control simula-tion. The initial and boundary conditions for RegCM4.3.4–CLM4.5 were derived from the CMIP5 CESM runs,including the 20th century simulation during 1981–2000 for the present-day control and the RCP8.5 runwith varying time period for the future experiments. Land cover was prescribed according to remote sens-ing data [Lawrence and Chase, 2007; Lawrence et al., 2011] in the present-day control simulation, and wasdealt differently in each future experiment.

There are a few key assumptions in the proposed modeling approach, which are related to the projectionof crop yield and cropland allocation using DSSAT and LandPro_Crop respectively. In running DSSAT forfuture scenarios in West Africa, we did not consider the farmers’ adaptive potential (e.g., through increasedfertilizer and labor input, expansion of irrigation, and switch to more heat- or drought-resistant cultivars) toaddress the yield loss caused by climate and socioeconomic changes. In LandPro_Crop, the cropland alloca-tion algorithm assumes that crop production and agricultural land use will have priority over natural landcover types and does not consider local/national land use policies accounting for the economic value of for-ests and grasslands. Additional sources of uncertainties have been discussed in the companion paper[Wang et al., 2017] and will not be repeated here.

3. Results and Discussion

3.1. Projected Cropland ExpansionThe FUTURE_TR_LUC experiment projects large increases in crop area at the expense of natural vegetationin many parts of the region. The substantial crop area expansion is caused by climate-induced loss in cropyield and increase in food demand (Figure 1). In the present-day crop area distribution, agricultural land useis more dominant in the eastern part of the region; extensive presence of grassland is noticeable in the cen-tral and the western parts of the region, whereas forest area largely dominates the coastal region in theSouth. According to the model projection, future changes in climate and socio-economic factors would leadto almost complete exhaustion of natural vegetation in the eastern part. More than 90% of land area is pro-jected to be occupied by cropland in Nigeria, Benin and Togo. In the eastern part of the region, fraction ofcropland in Gambia would also comprise almost 95% of total land. Compared to results from theequilibrium-mode projections FUTURE_EQ_LUC [Ahmed et al., 2016; Wang et al., 2017], country-total crop

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 380

area expansion by the mid-century projected by the transient runs is substantially higher in many of thecountries (Figure 2). However, the projected increase in crop area is slightly smaller in the transient modethan in the equilibrium mode for Guinea and Sierra Leone, while for Niger and Ivory Coast the projectedincrease is similar between the two runs. Comparison in the spatial distributions of fraction of crop areabetween the two different experiments indicates that the larger cropland expansion in the transient run ismore evident in the central part of the region, while over the southwest of West Africa the transient run pro-duces less or a similar level of cropland expansion.

The projected transient changes of total cropland coverage in each of the West African countries are pre-sented in Figure 3. The projected trend of cropland expansion in different countries follows very differentpatterns, and in any given country the rate of projected expansion from one 5 year period to another canlargely vary over the entire simulation period (2005–2050). A sharp increase in the rate of crop area expan-sion during a particular 5 year window is common for many of the countries, and the spikes in the rate ofcrop area expansion cannot be fully explained by the trend of the IMPACT-projected food demands whichusually follow a smooth increasing trend for most of the crops in the West African countries (Figure 4).

Figure 1. Spatial distribution of crop, forest and grass coverage (%) in 14 West African countries from present-day (year 2005) observation (top row), projected future changes by theLandPro_Crop algorithm driven in the transient mode by mid-21st century under the CESM-climate (middle row), and the differences in projections from the equilibrium run (describedin Ahmed et al. [2016])

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 381

Although socioeconomic factors (characterized by changes in food demand) are projected to dominatefuture changes in agricultural land use for most of West Africa [Ahmed et al., 2016], they cannot account forthe strong temporal dynamics shown in Figure 3. Instead, productivity of the major crops, which is animportant pathway for climatic change to affect the agricultural land use, is the dominant cause for thestrong dynamics of cropland expansion. The DSSAT-projected annual country-average crop yields showlarge inter-annual variability caused by climate variability [Ahmed et al., 2015]. This strong variability couldlead to abnormally low yield over a 5 year period, which would result in a sharp increase in the rate of croparea expansion during that period as shown in Figure 3. For example, the projected country-average crop-land fractional coverage in Ghana increases by 13.7% between 2030 and 2035. This increase is noticeablyhigher than the percentages of cropland expansion during the previous 5 year periods from 2005 to 2030.Similar increase (13.5%) in cropland fractional coverage would also occur between 2040 and 2045. Time-series of the projected food demand in Ghana show a smooth rate of increase for all the crops throughoutthe whole study period (2005–2050) (Figure 4). However, for each country, although the overall country-average yield is projected to decease for all five crops, the trend pattern is not consistent across differentcrops. For example, the country-average maize yield in 2035 is less than 2030 in Ghana by almost 8.9%,which could lead to a higher deficit during that period resulting in the higher rate of crop area expansion.Similarly, the country-average cassava yield in Ghana is projected to decrease by 16.6% from 2040 to 2045,which is behind the larger crop area expansion during that 5 year period. In Guinea-Bissau as another exam-ple, although the cropland expansion is minimal during 2010–2035, a large decrease in maize yield (by

almost 30%) during 2035–2040 anda large decrease in cassava yield (byabout 25%) during 2040–2045 lead to asharp increase in the projected country-average crop area over the course of adecade.

Comparison between Figures 2 and 3indicates that for countries that areprojected to experience strong fluc-tuations of crop yield (and thereforesharp expansion of cropland duringsome of the 5 year segments), thetransient approach produces a muchlarger cropland expansion by 2050than the equilibrium approach; forcountries where such strong tempo-ral dynamics is absent (e.g., Guinea,

0

10

20

30

40

50

60

70

80

90

100

Perc

enta

ge o

f cou

ntry

-ave

rage

crop

are

a

SPAM 2005

FUTURE_EQ_LUC

FUTURE_TR_LUC

Figure 2. Present-day (SPAM 2005) and the LandPro_Crop-projected future (mid-21st century) average crop area coverage in the WestAfrican countries.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

benin

burkinafaso

gambia

ghana

guinea

guineabissau

ivorycoast

mali

niger

nigeria

senegal

sierraleone

togo

Figure 3. Time series (2005–2050) of changes in fraction of crop area in each ofthe West African countries.

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 382

Ivory Coast, Niger), the two approaches produce very similar results. Indeed the substantially faster croplandexpansion projected by the transient approach is attributed to the climate-induced fluctuations of crop yield inthe model. The abnormally low crop yield during some segment(s) of the transient duration causes rapidexpansion of cropland to include areas that are relatively productive in that specific period but may not be pro-ductive under climate of subsequent segment(s). Cropland in subsequent years would therefore include a cer-tain fraction that is no longer productive, because in the LandPro_Crop model, once a piece of land isconverted to cropland it stays as cropland. This inclusion of historically productive land that is no longer pro-ductive directly causes the larger cropland expansion in the transient run than in the equilibrium run. The larg-est differences between the two approaches are projected over Senegal, Burkina Faso, Ghana, Togo, Benin,and part of Nigeria, most of which are located in the central part of West Africa. As this is the transition zonebetween the strong dry signal in the west and strong wet signal in the east based on projected future precipi-tation changes (as shown in section 3.2), it is not surprising that future crop yield in this region is projected toexperience stronger fluctuations than elsewhere.

3.2. Projected Future Climate ChangesBased on experimental designs described in section 2.2, differences between FUTURE_EQ_LUC and the con-trol experiment quantify future climate changes with land use scenarios projected using an equilibriumapproach, while the differences between FUTURE_TR_LUC and the control experiment reflect the changesaccounting for transient processes in LULCC dynamics. Additionally, the differences between FUTURE_EQ_LUC and FUTURE_TR_LUC highlight the impact of transient interactions between land use and climate onregional climate projections in West Africa. Changes in land use and cover affect regional climate primarilyvia changing albedo, Bowen ratio, and surface roughness. Because of the projected LULCC in West Africacharacterized by conversion of natural vegetation into cropland, albedo would increase in many parts ofthe region (Figure 5, top row). Magnitudes of albedo changes projected by FUTURE_EQ_LUC and FUTUR-E_TR_LUC experiments are generally similar, although the latter projects noticeably larger increase in somearea (e.g., the central-west part of Nigeria) because of the faster crop area expansion replacing natural vege-tation. Crop area expansion would also lead to reduced leaf area index (LAI) across the region except forseveral scattered areas where cropland would expand over currently bare land due to projected increase ofprecipitation (Figure 5, bottom row). Consistent with the difference in cropland expansion shown in Figure

0

2

4

6

8

10

12

14

16

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Dem

and

(100

0 to

n)Ghana

Maize Millet Sorghum Cassava Peanut

0

0.04

0.08

0.12

0.16

0.2

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Guinea Bissau

Maize Millet Sorghum Cassava Peanut

0

1000

2000

3000

4000

5000

6000

0

200

400

600

800

1000

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Yiel

d (k

g/ha

)

Maize Millet Sorghum Peanut Cassava

0

2000

4000

6000

8000

10000

0

200

400

600

800

1000

1200

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Maize Millet Sorghum Peanut Cassava

Figure 4. Time-series (2005–2050) of the IMPACT-projected food demand (upper row) and the DSSAT-Projected yield (lower row) for five major crops in Ghana (left column) and Guinea-Bissau (right column): maize, millet, sorghum, and peanut on the primary y axis, and cassava on the secondary y-axis due to its large magnitude.

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 383

2, FUTURE_TR_LUC projects smaller decrease of LAI than FUTURE_EQ_LUC along the southwest coast, andlarger decrease in the north.

Under future greenhouse gas emission and land use scenarios, the model generally projects a drier sum-mer over West Africa (Figure 6, top row). Average summer precipitation in many parts of the regionwould decrease by more than 1 mm/d. However, the model projects increased precipitation mainly inthe northern Nigeria, southern Niger, southern Chad, and Central Africa, leading to a west-east dry-wetcontrast in the projected changes. Differences between FUTURE_NO_LUC and FUTURE_TR_LUC indicatethat that land cover changes contribute to a significant portion of the projected dry-wet dipole of pre-cipitation changes, with the rest resulting from future GHGs concentration changes as discussed exten-sively in the companion paper [Wang et al., 2017]. The impact of land cover changes is especiallyimportant for the projected wet signal in the eastern part of the domain. This rather unique spatial pat-tern of precipitation changes related to land cover degradation results from the interaction of surfaceroughness changes with the low-level southeasterly flow during the summer monsoon [Wang et al.,2017, companion paper].

The large-scale spatial pattern of precipitation changes does not differ much between FUTURE_EQ_LUCand FUTURE_TR_LUC, although the wet signal is much stronger and spatially more extensive in FUTUR-E_TR_LUC. At the individual country level however, transient processes in land use-climate interactionscould considerably impact both the magnitude and direction of future changes in summer precipitation.For example, the projected increase in precipitation in Nigeria by the FUTURE_EQ_LUC experimentbecomes more noticeable in the FUTURE_TR_LUC experiment, and the two experiments largely differ inprojecting the future precipitation changes in Niger too. The transient experiment simulates substantiallylarger increase in precipitation which extends to the northern part of Niger. On the contrary, the equilibriumrun projects large increase in precipitation in the central Burkina Faso contrasting the transient run whichdoes not produce a wet signal.

Figure 5. Future changes in LAI and surface albedo averaged over summer (JJA) months projected by future-climate runs (equilibrium: left column and transient: middle column) usingtwo different land use change scenarios in West Africa, and the difference between two projections (right column).

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 384

A significant decrease of evapotranspiration (ET) is projected over most of the region by both the equilibri-um and transient experiments (Figure 6, middle row). As indicated by the differences between FUTURE_TR_-LUC and FUTURE_NO_LUC, part of the decrease in ET is caused by the degradation of vegetation coverleading to a decrease in LAI (as shown in Figure 4), which limits moisture supply to the atmosphere andthus contributes to the projected decrease of precipitation. The rest of the ET decrease might be a result ofprecipitation decrease leading to less water available for ET. In the northeastern part of West Africa whereLAI is projected to decrease in some areas and increase in others, both the equilibrium and the transientexperiments project an increase of ET accompanied by an increase in summer precipitation. The increase isgreater in both the magnitude and spatial extent in the transient run, which projects an increase of summeraverage ET by �0.25 mm/d across northern Nigeria and southern Niger coinciding with the strongestincrease of precipitation.

The model-projected increase in average summer temperature is generally similar in both magnitudes andspatial patterns between the two future climate experiments (Figure 6, bottom row). The lack of substantialdifference between FUTURE_EQ_LUC and FUTURE_TR_LUC in temperature projection indicates that landuse feedback, given the magnitude and location as projected in this study, does not particularly influencethe projection of future warming across the region. Apart from coastal regions, the warming signals fromboth the equilibrium and transient runs follow a generally similar spatial pattern, with a stronger warmingin the north than in the south. The conversion of forest to cropland generally reduces ET and thereforereduces evaporative cooling. The resulting warming effects are mostly compensated by the cooling effects

Figure 6. Future changes in average summer (JJA) precipitation (mm/d), ET and temperature projected by future-climate runs (equilibrium: first column and transient: second column)using two different land use change scenarios in West Africa, and the differences between two future projections (third and fourth columns).

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 385

of increased albedo causing decrease in absorbed solar radiation (Figures 5 and 7). The spatial pattern ofprojected changes in surface incident solar radiation and absorbed solar radiation from both the FUTUR-E_EQ _LUC and FUTURE_TR_LUC experiments resemble that of precipitation changes, with increase(decrease) of solar radiation corresponding to decrease (increase) of precipitation. This correspondence canbe attributed to the cloud feedback associated with precipitation. However, the projected decrease ofabsorbed solar radiation is larger than the projected decrease of insolation, due to the impact of albedoincrease associated with land cover changes. The projected changes of surface radiation budget in FUTUR-E_TR_LUC are substantially larger in magnitude than in FUTURE_EQ_LUC across the domain. The coolingeffects of albedo increase are partially compensated by the warming effects of reduced ET during the mon-soon season. As a result, land use land cover changes would cause a rather minor decrease of surface tem-perature, offsetting a small fraction of GHGs warming in the region.

4. Summary and Conclusion

Based on numerical experiments using an asynchronously coupled climate-land use modeling framework,this study investigates the potential impact of climate change on agricultural land use and the resulting feed-back to regional climate in West Africa using a transient approach, and compares results with those from anequilibrium approach. The model, without accounting for agricultural intensification, projects substantial crop-land expansion by the mid-century because of climate-induced losses in crop yield and increases in fooddemand. By mid-century, during the summer monsoon season (June, July, August), the projected increase ofsurface air temperature ranges from less than 1.5 degrees along the coastal region to more than 3 degreesover the desert further north inland, while precipitation is projected to decrease by more than 1 mm/d overmost of the region, with some increase primarily in the eastern part over Nigeria, Niger, and areas further east.Majority of the projected rainfall decrease in the model is caused by greenhouse gas concentration changes,while cropland expansion is responsible for majority of the projected increase of rainfall in the eastern part.Cropland expansion also tends to reduce the magnitude of projected warming.

Compared with projections for the same region in an application of the same model using an equilibriumapproach [Wang et al., 2017, companion paper], the transient approach used in this study projects fastercropland expansion for most of the West African countries. Correspondingly, the transient approach

Figure 7. Future changes in average summer (JJA) daily incident and absorbed solar radiation (W/m2) projected by future-climate runs (equilibrium: first column and transient: secondcolumn) using two different land use change scenarios in West Africa, and the differences between two future projections (third and fourth columns).

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 386

produces a larger and spatially more extensive increase of the monsoon rainfall than the equilibriumapproach, primarily over Niger, Nigeria, and Chad. For other countries within the region, the projectedfuture climate changes show no qualitative differences between the two approaches. The quantitative dif-ference in climate between the FUTURE_TR_LUC and the FUTURE_EQ_LUC projections is generally small.The future changes of climate projected using these two approaches agree in directionality even for precip-itation, which implies that the equilibrium approach captures the first order impact while the transientapproach accounts for additional but mostly minor changes.

The substantially faster cropland expansion projected by the transient approach can be attributed to thetemporal dynamics of climate-induced yield loss and how well the impact of yield fluctuations on land useis represented in the LandPro_Crop model. Once a piece of land is converted to cropland, it stays as crop-land in the model. As a result, historically productive land that is no longer productive is still retained in thetransient projection. In reality, only some of the less productive (but still arable) land may be retained, whileother less productive land (and land with repeated crop failures) will be abandoned. Not representing thismechanism amplifies the differences in future projections using the two different approaches, and is amajor limitation for the transient application of LandPro_Crop (and therefore a main cause of uncertainty inthe transient projections). In follow-up research, the LandPro_Crop model will be further developed toaccount for cropland abandonment driven by low yield and its conversion back to natural vegetation orbare soil depending on climate conditions. It is expected that with the resulting improved LandPro_Cropmodel, the transient approach would project a level of cropland expansion closer to the equilibriumapproach, although their natural vegetation cover would still be different.

Another limitation of this study is the lack of consideration for potential adjustment of national import poli-cy in response to temporary extremely low crop yield. Under such circumstance, a country can choose toimport more to alleviate the food demand pressure (as opposed to rapidly expanding its arable land). Assome governments in West Africa aim for self-sufficiency while others lack the means to suddenly increaseimport, the impact of this model limitation might be moderate.

Results from this study provide a basis for future research efforts for comprehensive assessments and robustprojections of regional climate and agriculture systems. Specifically, the consideration for land use–climateinteractions in regional climate projections can be facilitated by asynchronously coupling crop models, landuse models, and agricultural economics models with climate models. The equilibrium approach of runningthe asynchronously coupled modeling framework can sufficiently capture the primary impact of the interac-tions between climate and agricultural systems for climate projection purposes, and the more cumbersomeapplication of the transient approach is likely only necessary for land use projections. For similar reasons, itis likely not critical, although desirable, that climate projection be done based on modeling systems thatincorporate synchronous coupling between climate and agricultural land use. The equilibrium approach ofapplying an asynchronously coupled modeling framework [Wang et al., 2017] can work as a feasible alterna-tive. This, however, does not mean that transient dynamics is not important. It could significantly modifythe projected land use dynamics depending on national policies and is especially important at shorter timescales. To support impact assessment and to provide actionable information for the development of climateadaptation strategies, transient dynamics of the agricultural system needs to be accounted for.

ReferencesAhmed, K. F., et al. (2013), Statistical downscaling and bias correction of climate model outputs for climate change impact assessment in

the U.S. northeast, Global Planet. Change, 100, 320–332, doi:10.1016/j.gloplacha.2012.11.003.Ahmed, K. F., et al. (2015), Potential impact of climate change on cereal crop yield in West Africa, Clim. Change, 133, 321–334.Ahmed, K. F., et al. (2016), Potential impact of climate and socioeconomic changes on future agricultural land use in West Africa, Earth Syst.

Dyn. Discuss., 6, 1129–1162.Brovkin, V., T. Raddatz, C. H. Reick, M. Claussen, and V. Gayler (2009), Global biogeophysical interactions between forest and climate, Geo-

phys. Res. Lett.,36, L07405, doi:10.1029/2009GL037543.Brovkin, V., et al. (2013), Effect of anthropogenic land-use and land-cover changes on climate and land carbon storage in CMIP5 projec-

tions for the twenty-first century, J. Clim., 26(18), 6859–6881.Burney, J. A., S. J. Davis, and D. B. Lobell (2010), Greenhouse gas mitigation by agricultural intensification, Proc. Natl. Acad. Sci. U. S. A., 107,

12,052–12,057.Charney, J. G., W. J. Quirk, S. H. Chow, and J. Kornfield (1977), A comparative study of the effects of albedo change on drought in semi-arid

regions, J. Atmos. Sci., 34, 1366–1385.Claussen, M., V. Brovkin, and A. Ganopolski (2001), Biogeophysical versus biogeochemical feedbacks of large-scale land cover change,

Geophys. Res. Lett., 28, 1011–1014.

AcknowledgmentsThis study was supported by fundingfrom NSF for a collaborative EaSMproject (AGS-1049017, AGS-1048967,AGS-1049186). Computational supportwas provided by NCAR Yellowstone(UCNN0001). We thank the twoanonymous reviewers for theirconstructive comments on an earlierversion of this paper. All data arearchived at the Hydroclimatology andBiosphere-Atmosphere Interactionsgroup at the University of Connecticut,and are available for download uponrequest (guiling.wang@uconn.edu).

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 387

Emanuel, K. A. (1991), A scheme for representing cumulus convection in large-scale models, J. Atmos. Sci., 48, 2313–2335.Frieler, K., et al. (2015), A framework for the cross-sectoral integration of multi-model impact projections: Land use decisions under climate

impacts uncertainties, Earth Syst. Dyn., 6, 447–460.Giorgi, F., et al. (2012), RegCM4: Model description and preliminary tests over multiple CORDEX domains, Clim. Res., 52, 7–29.Hagos, S., L. R. Leung, Y. Xue, A. Boone, F. de Sales, N. Neupane, M. Huang, and J. H. Yoon (2014), Assessment of uncertainties in the

response of the African monsoon precipitation to land use change simulated by a regional model, Clim. Dyn., 43, 2765–2775.Hurtt, G. C., et al. (2011), Harmonization of land-use scenarios for the period 1500–2100: 600 years of global gridded annual land-use transi-

tions, wood harvest, and resulting secondary lands, Clim. Change, 109, 117–161.Jones, J. W., et al. (2003), DSSAT Cropping System Model, Eur. J. Agron., 18, 235–265.Koster, R. D., et al. (2004), Regions of strong coupling between soil moisture and precipitation, Science, 305, 1138–1140.Lawrence, P. J., and T. N. Chase (2007), Representing a new MODIS consistent land surface in the Community Land Model (CLM 3.0), J. Geo-

phys. Res., 112, G01023, doi:10.1029/2006JG000168.Lawrence, P. J., et al. (2011), Parameterization improvements and functional and structural advances in Version 4 of the Community Land

Model, J. Adv. Model. Earth Syst., 3, 1, doi:10.1029/2011MS00045.Lee, X., et al. (2011), Observed increase in local cooling effect of deforestation at higher latitudes, Nature, 479, 384–387.Oleson K, et al. (2013), Technical description of version 4.5 of the Community Land Model (CLM), NCAR Tech. Note NCAR/TN-5031STR, 420

pp., doi:10.5065/D6RR1W7M.Pielke, R. A., et al. (2011), Land use/land cover changes and climate: Modeling analysis and observational evidence, WIREs Rev. Clim.

Change, 2, 828–850.Pitman, A. J., et al. (2009), Uncertainties in climate responses to past land cover change: First results from the LUCID intercomparison study,

Geophys. Res. Lett., 36, L14814, doi:10.1029/2009GL039076.Pongratz, J., et al. (2010), Biogeophysical versus biogeochemical climate response to historical anthropogenic land cover change, Geophys.

Res. Lett., 37, L08702, doi:10.1029/2010GL043010.Rounsevell, M. D. A., et al. (2014), Towards decision-based global land use models for improved understanding of the Earth system, Earth

Syst. Dyn., 5, 117–137.Rosegrant, M. W. (2012), International Model for Policy Analysis of Agricultural Commodities and Trade (IMPACT ) Model Description, Int. Food

Policy Res. Inst., Washington, D. C.Sylla, M. B., J. S. Pal, G. L. Wang, and P. J. Lawrence (2016), Impact of land cover characterization on regional climate modeling over West

Africa, Clim. Dyn., 46, 637–650.Sheffield, J., G. Goteti, and E. F. Wood (2006), Development of a 50-yr, high resolution global dataset of meteorological forcings for land

surface modeling, J. Clim., 19(13), 3088–3111.Taylor, C. M., et al. (2002), The influence of land use change on climate in the Sahel, J. Clim., 15, 3615–3629.Wang, G., and E. A. Eltahir (2000), Ecosystem dynamics and the Sahel drought, Geophys. Res. Lett., 27, 795–798.Wang, G., et al. (2015), On the development of a coupled regional climate-vegetation model RCM-CLM-CN-DV and its validation in Tropical

Africa, Clim. Dyn., 46, 515–539, doi:10.1007/s00382-015-2596-z.Wang, G., M. Yu, and Y. Xue (2016), Modeling the potential contribution of land cover changes to the late twentieth century Sahel drought

using a regional climate model: Impact of lateral boundary conditions, Clim. Dyn., 47, 3457–3477, doi:10.1007/s00382-015-2812-x.Wang, G., K. Farzan Ahmed, L. You, M. Yu, J. Pal, and Z. Ji (2017), Projecting regional climate and cropland changes using a linked biogeo-

physical-socioeconomic modeling framework: 1. Model description and an equilibrium application over West Africa, J. Adv. Model. EarthSyst., 9, 354–376, doi:10.1002/2016MS000712.

Xue, Y., et al. (2010), Intercomparison and analyses of the climatology of the West African Monsoon in the West African Monsoon Modelingand Evaluation project (WAMME) first model intercomparison experiment, Clim. Dyn., 35, 3–27, doi:10.1007/s00382-010-0778-2.

Zeng, N. et al. (1999), Enhancement of interdecadal climate variability in the Sahel by vegetation interaction, Science, 286, 1537–1540.

Journal of Advances in Modeling Earth Systems 10.1002/2016MS000721

AHMED ET AL. TRANSIENT CLIMATE-CROPLAND PROJECTIONS 388