Short Paper
Area of land degradation in arid regions from 2000 to 2023
2025 Volume 81 Issue 4 Pages 219-224
Details
2025 Volume 81 Issue 4 Pages 219-224
The majority of recent land uses in drylands have been grasslands (32.2% of all drylands) and dryland forests (typical forest and sparse forest including shrub and savanna, 23%). If these lands are appropriately managed, they have the potential to sequester 0.84 Gt of soil organic carbon per year. However, climate change associated with global warming has led to an increase in temperature and irregular rainfall, which can both exacerbate the damage of droughts and desertification, and there is an urgent need to develop sustainable land management in arid regions. This study examined the interannual changes in the degraded land area in arid regions derived from a threshold value of the normalized difference vegetation index (NDVI) and the satellite-based aridity index (SbAI) from 2000 to 2023. Here, degraded land includes existing deserts and land having both permanent and temporal aeolian desertified areas. The total area of degraded lands was found to have exhibited a decreasing trend since 2000, but has gradually increased since about 2015. A method based on climo-vegetation regions is presented that uses NDVI and SbAI along with the climatological land use and classification of arid regions to understand current vegetation conditions as well as areas that deviated from their climatic potential.
Arid regions have already been affected by climate changes associated with global warming, with some regions experiencing increases in drought and desertification (Mirzabaev et al., 2022). These arid regions are highly vulnerable to climate change through increased temperatures and irregular rainfall, and they will likely face an increasing number of extreme weather events as global warming progresses in the future (Dai, 2011; Koutroulis, 2019). In addition, a simultaneous increase in environmental stress associated with human activities will exacerbate the damage caused by disasters such as desertification, drought, and heat waves (UNEP, 1997; Millennium Ecosystem Assessment (MA), 2005; Cherlet et al., 2018; Mirzabaev et al., 2022). Because drylands account for about half of all land area, developing sustainable land management (SLM) will be a crucially important issue in mitigating and adapting to ongoing climate change (MA, 2005; IPCC, 2021).
To develop mitigation and adaptation for climate change, it is important to know both the current situation and future changes of drylands (Cowie et al., 2011). Using a numerical simulation and Representative Concentration Pathway emissions scenarios, Mirzabaev et al. (2022) indicated that the future expansion of arid regions is probable in southwest North America, the northern fringe of Africa, Southern Africa, and Australia. The main areas of semi-arid expansion are likely to occur on the north side of the Mediterranean, Southern Africa, and North and South America (e.g., Feng and Fu, 2013). Kimura (2020) compared the satellite-based aridity index (SbAI) and the climatic traditional aridity index (AI) from 2001 to 2013 and concluded that land areas corresponding to the transition zone toward dryness included North and South America, Southern Africa, countries around the Mediterranean, Central Asia, and Northeast Asia.
In a study of current conditions in drylands, Kimura and Moriyama (2019a) defined areas where the annual maximum Normalized Difference Vegetation Index (NDVImax) was <0.2 and the annual averaged SbAI was >0.025 as degraded land (i.e., aeolian desertified areas). In addition, Wang (2011, 2014) defined aeolian desertification as land degradation characterized by wind erosion in arid regions. Although desertification has generally been assessed using complex methods based on various factors in individual dryland regions, aeolian desertification should be considered as a common “yardstick” to monitor the progress of desertification (Shinoda, 2002).
Kimura and Moriyama (2019a) indicated that annual changes in the global area of degraded land decreased slightly from 2000 to 2017. However, they did not specify whether the land degradation occurred in drylands or elsewhere. Because “desertification” has been defined as land degradation in drylands (UNEP, 1997; MA, 2005), an objective of this study was to extract degraded land areas in drylands vulnerable to climate change, as well as to continuously monitor their secular changes.
In the study of degraded areas, the lifetime of satellites is an important issue. The TERRA satellite containing mounted MODIS (Moderate Resolution Imaging Spectroradiometer) instruments are nearing the end of their services lives (NASA, 2025). MODIS has collected valuable data from 2000 to 2020, which have been used in various studies. For example, Chen et al. (2019) indicated that MODIS data from 2000 to 2017 showed increasing vegetation globally due to direct factors (e.g., human land-use management) and indirect factors (e.g., climate change). To achieve our goal of continuously monitoring changes in degraded dryland areas over time, it is necessary to transition to data from next-generation satellites without losing any valuable continuous data.
In this study, we estimated SbAI and NDVI from 2018 to 2023 by using SGLI (Second-generation GLobal Imager) data from the GCOM-C (Global Change Observation Mission-Climate) satellite. By comparing and calibrating these values with values obtained with MODIS data from 2018 to 2020, we attempted to transition monitoring to SGLI starting in 2021. We then made a continuous estimation of the area of land degradation in arid regions from 2000 to 2023. Finally, we indicated the relationship between annual averaged SbAI and yearly maximum NDVI from 2000 to 2023 with the climatological land use and classification of arid regions to understand current vegetation conditions as well as areas that deviated from their climatic potential to obtain information that would contribute to improving SLM in drylands.
We targeted arid regions between latitudes of 55°N and 55°S to exclude subarctic tundra areas, where the climate is subarctic. Arid regions classified as hyper-arid, arid, semi-arid, and dry sub-humid were defined by using the result of AI distributions averaged from 2000 to 2020 (the total arid regions was 57,902,038 km2) (Fig. 1; Kimura and Moriyama, 2024).
Fig. 1. Spatial distribution of the four dryland types averaged from 2000 to 2020 (data from Kimura and Moriyama, 2024) determined by the calculation of the climatic aridity index (AI) (= ratio of annual precipitation to annual potential evaporation).
The period of analysis was from 2000 to 2023. MODIS has provided data that have enabled the calculation of NDVI and SbAI since 2000, but since 2021, the data were impossible to use because of orbit deviations (NASA, 2025). Therefore, beginning in 2021, next-generation satellite data (i.e., data from an SGLI mounted on the GCOM-C satellite) were used to continue the monitoring of land degradation.
Daily NDVI was calculated on the basis of near-infrared (NIR) and visible (RED) wavelength channels of the MODIS and SGLI data products, that is, MOD09CMG (Vermote, 2015) on MODIS which has a 0.05° resolution, and Level-3 on SGLI (Murakami et al., 2022) which has a 0.05° resolution. Yearly maximum NDVI (NDVImax) was selected as the highest daily value of NDVI during the year.
This study relied on daily SbAI values, the physical meaning of which is the opposite of heat capacity and/or thermal inertia determined by land surface wetness (Kimura and Moriyama, 2014). The values are derived from MOD09CMG and MOD11C1 on MODIS (Wan et al., 2015) and Level-3 on SGLI for broadband albedo and land surface temperature (LST); both datasets have a 0.05° resolution. Our SbAI has been used in previous studies of arid regions of various countries to monitor drought conditions and land surface wetness (Bakhtiari et al., 2021; Niu et al., 2022; Casañas et al., 2024).
Daily SbAI is defined as follows (Kimura and Moriyama, 2014):
where
and ∆Ts is the difference in LST between day and night; Rs is the absorbed solar radiation as calculated from broadband albedo r, the solar constant S0 (1367 W m-2), and the solar zenith angle at the Sun’s apex θc; and ri is the spectral reflectance of channels 1 through 7 for MODIS (Liang, 2004) and VN8, VN11, and SW3 for SGLI (Susaki, 2020). Poor pixels in the Quality Assurance dataset of day or night LST and in the reflectance products were eliminated from computation. Dry surfaces have low thermal inertia (Jones and Vaughan, 2010), which leads to large values of ∆Ts and hence high SbAI. Daily SbAI values were averaged to derive annual values for our analysis.
We examined the distribution of annual maximum NDVI (hereafter, NDVImax) < 0.2 and annual averaged SbAI (hereafter, SbAI) > 0.025 in arid regions and identified areas that meet both these criteria as degraded land (aeolian desertified areas). These areas include existing desert and land with both permanent and temporal aeolian desertified areas (Kimura and Moriyama, 2019a; 2021). Because the grid scale for the AI was 1.0° × 1.0° (Kimura and Moriyama, 2024), the spatial resolution of the AI was changed to 0.05°; that is, the AI value within a 1.0° grid was the same as it is in the 20 × 20 grids.
The MODIS Land Cover Climate Modeling Grid Product (MCD12C1) was used for the classification of land use in arid regions in 2020 (Friedl and Sulla-Menashe, 2015), especially typical forest, sparse forest (shrub and savanna), and grasslands (Kimura and Moriyama, 2024).
2.2. Classification of arid regions using SbAIIn this study, SbAI was also used to classify arid regions. Kimura and Moriyama (2019b) classified arid regions from 2001 to 2013 according to the relationship between SbAI and AI:
Hyper-arid: SbAI > 0.025 (AI < 0.05),
Arid: 0.022 ≤ SbAI ≤ 0.025 (0.05 ≤ AI < 0.2),
Semi-arid: 0.017 ≤ SbAI < 0.022 (0.2 ≤ AI < 0.5),
Dry sub-humid: 0.015 ≤ SbAI < 0.017 (0.5 ≤ AI < 0.65).
The AI ranges given in parentheses correspond to the definitions given in UNEP (1997), MA (2005), and Cherlet et al. (2018).
2.3. Climo-vegetation regions in arid regionsClimo-vegetation regions are based upon the concept that the world’s climate and vegetation are closely related to each other and form identifiable regions (Shinoda, 2021). Kimura and Moriyama (2024) examined the relationship between AI (climate) and NDVImax (vegetation) and defined the distributed range of the AI value in each land use category of arid regions, especially in typical forest, sparse forest, and grasslands (what we call “climatological land use”). The distributed ranges of AI and NDVImax for each climatological land use were as follows:
Typical forests: 0.45 ≤ AI ≤ 0.63, 0.82 ≤ NDVImax ≤ 0.90,
Sparse forests: 0.12 ≤ AI ≤ 0.46, 0.27 ≤ NDVImax ≤ 0.71,
Grasslands: 0.18 ≤ AI ≤ 0.48, 0.37 ≤ NDVImax ≤ 0.71.
The concept of climo-vegetation regions provides useful information that contributes to SLM in drylands. For example, for grasslands and sparse forests, which account for the majority of the arid land, an AI of 0.3 was the threshold for their stable existence (Kimura and Moriyama, 2024). In this study, because SbAI was used to classify arid regions, the distributed range of SbAI using Eq. (5) was modified as follows for each climatological land use:
Typical forests: 0.015 ≤ SbAI ≤ 0.018,
Sparse forests: 0.018 ≤ SbAI ≤ 0.024,
Grasslands: 0.018 ≤ SbAI ≤ 0.023.
The relationship between NDVImax obtained from SGLI (NDVImaxSGLI) and NDVImax from MODIS (NDVImaxMODIS) from 2018 through 2020 in arid regions is shown in Fig. 2 and can be represented by the following regression equation:
Fig. 2. Relationship between NDVImaxMODIS and NDVImaxSGLI. The color bar denotes the number of pixels.
When the NDVI value was <0.7, NDVImaxSGLI was slightly lower than NDVImaxMODIS. However, NDVImaxSGLI values were consistent with NDVImaxMODIS values, with a bias of －0.0033, a root mean square error (RMSE) of 0.047, and an R2 of 0.968 over the arid regions. Bayarsaikhan et al. (2022) also indicated that NDVISGLI and NDVIMODIS were highly consistent throughout Mongolia in 2019 (R2 = 0.93, RMSE = 0.079).
The relationship between SbAI obtained from SGLI (SbAISGLI) and SbAI from MODIS (SbAIMODIS) from 2018 through 2020 is shown in Fig. 3 and can be represented by the following regression equation:
Fig. 3. Relationship between SbAIMODIS and SbAISGLI. The color bar denotes the number of pixels.
SbAISGLI values were consistent with those of SbAIMODIS, with a bias of 5.4×10-5, RMSE of 0.003, and R2 of 0.816. However, SbAISGLI was lower than SbAIMODIS when SbAISGLI < 0.025 and higher when SbAISGLI > 0.025. This is the first comparison of SbAI between SGLI and MODIS, and the main factor contributing to the difference between them has been thought to be due to the algorithm for estimating LST, but consideration of the estimation accuracy of the LST itself is beyond the scope of this study. Eq. (7) was therefore used to convert SbAI monitoring data starting in 2021.
3.2. Areas of degraded land over arid regionsAreas of degraded land over the arid regions were estimated using MODIS from 2000 to 2020 and SGLI from 2021 to 2023 (Fig. 4). The RMSE between the degraded land areas estimated by SGLI and MODIS from 2018 to 2020 was 527,290 km2 (0.9% of the total arid regions of 57,902,038 km2). There was no statistically significant increase or decrease in the areas of degraded land throughout the period from 2000 to 2023. However, when divided into intervals of approximately a decade, clear decreasing or increasing trends could be identified.
Fig. 4. Annual change (2000–2023) in the area of degraded land (left axis) and the respective percent change relative to the total arid regions (57,902,038 km2, right axis).
According to Kimura and Moriyama (2019a), the annual change of the global area of degraded land from 2000 to 2017 decreased slightly, and a similar trend was obtained for only drylands (R2 = 0.224; p < 0.05) (Fig. 4). The RMSE between the degraded land area in the global region and that in dryland regions was 366,211 km2 from 2000 to 2017, and most of the land degradation area was in drylands (96%). The 4% discrepancy was identified in the border region between India and China, which was not classified as arid land.
The area of land degradation has increased significantly since 2015 (R2 = 0.946; p < 0.001). Fig. 5 shows the distributions of degraded land in 2015 and 2023. The area of degraded land increased in North America, the Sahel (Mauritania, Mali, Niger, Chad, Sudan), Central Asia (Turkmenistan, Uzbekistan, Kazakhstan), Southern Africa (Namibia, South Africa), China-Mongolia border, and Australia. The increases in the Sahel, Western Sahara, and Central Asia were particularly noticeable. There have been no similar studies of degraded land area, so it is not possible to compare our results with those of other studies. However, drought conditions occurred in the Sahel (Ndehedehe et al., 2020; Rauch et al., 2025), Central Asia (UNCCD, 2023), and North America (EPA, 2025) during this time period. Although this is a small sample, annual changes of degraded land areas in drylands may be partly influenced by climate change due to global warming, the El Niño and La Niña phenomena, and blocking high (Cook, 2019; IPCC, 2021). In contrast, the area of degraded land declined near the border between Tibet and the Taklamakan Desert in 2023. Between 2000 and 2020, approximately 57.7% of the Kunlun Mountains area exhibited significant greening trends. The greening was primarily attributed to climatic factors, notably increased temperatures and precipitation (Liu et al, 2021; Zheng et al., 2021).
Fig. 5. Global distributions of degraded land in 2015 and 2023. Purple and blue show degraded land in 2015, and purple and red show degraded land in 2023. The area outlined by the black rectangle is the Sahel belt (12°N to 20°N), as defined by Seaquist et al., (2009).
Fig. 6 was developed based on the concept of climo-vegetation regions, but it uses an actual degree of aridity (SbAI), rather than a climatic index such as AI (Gamo et al., 2013; Kimura and Moriyama, 2024). The average NDVImax (from 2000 to 2023) decreased with increased average annual SbAI (from 2000 to 2023), following a logistic curve (R2 = 0.480):
Fig. 6. Relationship between SbAI and NDVImax averaged from 2000 to 2023. Each solid outlined area represents the respective climatological land use and the dashed outlined areas are transition zones.
Fig. 6 also includes the range of respective climatological land use and class of arid regions classified by SbAI. Forests were found in the wetter parts of dry sub-humid regions, most grasslands were found in semi-arid regions, and sparse forests were found in a wide range of arid to semi-arid regions. The area that satisfies both SbAI > 0.025 and NDVImax < 0.2 was defined as the degraded land area (outlined by the orange rectangle in Fig. 6).
The transition zone (outlined by the dashed red lines) on the border between the area of degraded land and sparse forest has the potential to transition to degraded land or sparse forest. The transition zone on the border between the area of grassland and typical forest has the potential to transition to a typical forest or grassland. Because lands in the drier transition zone are sensitive to climate change and human activities, continuous monitoring is essential to prevent increasing dryness or desertification in zones such as the Sahel belt (Fig. 5) (Seaquist et al., 2009; Leroux et al., 2017; Tsubo et al., 2022).
Causes of desertification can be divided into natural and anthropogenic factors (UNEP, 1997; MA, 2005). Of the natural factors, drought is the most devastating, and it inflicts heavy damage on arid regions. As an area becomes drier, soil degradation progresses, and wind erosion occurs (Wang, 2011, 2014). We defined land prone to wind erosion as degraded land using NDVI and SbAI thresholds obtained from satellite data and monitored interannual changes (2000 to 2023) in the degraded land area in arid regions. MODIS data from 2000 to 2020 and SGLI data from 2021 to 2023 were used to calculate NDVI and SbAI.
The trend in the area of degraded land decreased since 2000, but it tended to gradually increase after around 2015. It is not possible to conclude whether the increasing trend is due to natural or anthropogenic factors. However, given the similarity with the drought trends in the Sahel, Central Asia, and Western United States, the increase may have been dominated by natural factors such as global climate change.
We presented a method based on the climo-vegetation regions. The method compared SbAI and NDVImax to understand the relationship between current vegetation conditions and land use relative to climatological conditions. We also identified areas that deviated from their climatic potential. For example, if satellite monitoring of SbAI and NDVImax at target sites indicate a degraded land category (or hyper arid region) in Fig. 6 and an arid climatic condition is confirmed on Fig. 1, we can judge that the present land use at this site is not suitable for climatic potential. The degraded land area in the Sahel belt in 2023 fits this example. Even if satellite monitoring of SbAI and NDVImax at target sites coincide with the identified climatic potential, if the NDVImax is small, the vegetation will be under stress as the result of some kind of environmental condition.
As shown in Figs. 4 and 5, the area of degraded land fluctuated in arid regions even in the period of 24 years. Given that both the SbAI and NDVI reflect short- and long-term drought conditions (Kimura and Moriyama, 2017; Niu et al., 2022), uninterrupted monitoring, particularly through satellite-based methods, is essential to assess the impacts of climate change on arid regions. And also, daily monitoring of SbAI and NDVI will make it feasible to diagnose hot spots or periods of drought on a seasonal or more frequent basis. It is our hope that the usefulness of this method will be confirmed generally in future studies and that it will become useful for assessing sustainability in arid regions.
This work was partly supported by JSPS KAKENHI (grant number: 25K02131). We appreciate the valuable comments from two reviewers and managing editor of this paper.
