Editing IPCC:AR6/SRCCL/Chapter-4 (section)

=== 4.2.4 Approaches to assessing land degradation ===

<div id="section-4-2-4-approaches-to-assessing-land-degradation-block-1"></div>

In a review of different approaches and attempts to map global land degradation, Gibbs and Salmon (2015) <sup>[[#fn:r335|335]]</sup> identified four main approaches to map the global extent of degraded lands: expert opinions (Oldeman and van Lynden 1998 <sup>[[#fn:r336|336]]</sup> ; Dregne 1998 <sup>[[#fn:r337|337]]</sup> ; Reed 2005 <sup>[[#fn:r338|338]]</sup> ; Bot et al. 2000 <sup>[[#fn:r339|339]]</sup> ); satellite observation of vegetation greenness – for example, remote sensing of Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Plant Phenology Index (PPI) – (Yengoh et al. 2015 <sup>[[#fn:r340|340]]</sup> ; Bai et al. 2008c <sup>[[#fn:r341|341]]</sup> ; Shi et al. 2017 <sup>[[#fn:r342|342]]</sup> ; Abdi et al. 2019 <sup>[[#fn:r343|343]]</sup> ; JRC 2018 <sup>[[#fn:r344|344]]</sup> ); biophysical models (biogeographical/ topological) (Cai et al. 2011b <sup>[[#fn:r345|345]]</sup> ; Hickler et al. 2005 <sup>[[#fn:r346|346]]</sup> ; Steinkamp and Hickler 2015 <sup>[[#fn:r347|347]]</sup> ; Stoorvogel et al. 2017 <sup>[[#fn:r348|348]]</sup> ); and inventories of land use/ condition. Together they provide a relatively complete evaluation, but none on its own assesses the complexity of the process (Vogt et al. 2011 <sup>[[#fn:r349|349]]</sup> ; Gibbs and Salmon 2015 <sup>[[#fn:r350|350]]</sup> ). There is, however, a robust consensus that remote sensing and field-based methods are critical to assess and monitor land degradation, particularly over large areas (such as global, continental and sub-continental) although there are still knowledge gaps to be filled (Wessels et al. 2007 <sup>[[#fn:r351|351]]</sup> , 2004 <sup>[[#fn:r352|352]]</sup> ; Prince 2016 <sup>[[#fn:r353|353]]</sup> ; Ghazoul and Chazdon 2017 <sup>[[#fn:r354|354]]</sup> ) as well as the problem of baseline values (Section 4.1.3).

Remote sensing can provide meaningful proxies of land degradation in terms of severity, temporal development, and areal extent. These proxies of land degradation include several indexes that have been used to assess land conditions, and monitoring changes of land conditions – for example, extent of gullies, severe forms of rill and sheet erosion, and deflation. The presence of open-access, quality controlled and continuously updated global databases of remote sensing data is invaluable, and is the only method for consistent monitoring of large areas over several decades (Sedano et al. 2016 <sup>[[#fn:r355|355]]</sup> ; Brandt et al. 2018b <sup>[[#fn:r356|356]]</sup> ; Turner 2014 <sup>[[#fn:r357|357]]</sup> ).The NDVI, as a proxy for Net Primary Production (NPP) (see Glossary), is one of the most commonly used methods to assess land degradation, since it indicates land cover, an important factor for soil protection. Although NDVI is not a direct measure of vegetation biomass, there is a close coupling between NDVI integrated over a season and in situ NPP ( ''high agreement, robust evidence'' ) (see Higginbottom et al. 2014 <sup>[[#fn:r358|358]]</sup> ; Andela et al. 2013 <sup>[[#fn:r359|359]]</sup> ; Wessels et al. 2012 <sup>[[#fn:r360|360]]</sup> ).

Distinction between land degradation/improvement and the effects of climate variation is an important and contentious issue (Murthy and Bagchi 2018 <sup>[[#fn:r361|361]]</sup> ; Ferner et al. 2018 <sup>[[#fn:r362|362]]</sup> ).There is no simple and straightforward way to disentangle these two effects. The interaction of different determinants of primary production is not well understood. A key barrier to this is a lack of understanding of the inherent interannual variability of vegetation (Huxman et al. 2004 <sup>[[#fn:r363|363]]</sup> ; Knapp and Smith 2001 <sup>[[#fn:r364|364]]</sup> ; Ruppert et al. 2012 <sup>[[#fn:r365|365]]</sup> ; Bai et al. 2008a <sup>[[#fn:r366|366]]</sup> ; Jobbágy and Sala 2000 <sup>[[#fn:r367|367]]</sup> ). One possibility is to compare potential land productivity modelled by vegetation models and actual productivity measured by remote sensing (Seaquist et al. 2009 <sup>[[#fn:r368|368]]</sup> ; Hickler et al. 2005 <sup>[[#fn:r369|369]]</sup> ; van der Esch et al. 2017 <sup>[[#fn:r370|370]]</sup> ), but the difference in spatial resolution, typically 0.5 degrees for vegetation models compared to 0.25–0.5 km for remote sensing data, is hampering the approach. The Moderate Resolution Imaging Spectroradiometer (MODIS) provides higher spatial resolution (up to 0.25 km), delivers data for the EVI, which is calculated in the same way as NDVI, and has showed a robust approach to estimate spatial patterns of global annual primary productivity (Shi et al. 2017 <sup>[[#fn:r371|371]]</sup> ; Testa et al. 2018 <sup>[[#fn:r372|372]]</sup> ).

Another approach to disentangle the effects of climate and land use/ management is to use the Rain Use Efficiency (RUE), defined as the biomass production per unit of rainfall, as an indicator (Le Houerou 1984 <sup>[[#fn:r373|373]]</sup> ; Prince et al. 1998 <sup>[[#fn:r374|374]]</sup> ; Fensholt et al. 2015 <sup>[[#fn:r375|375]]</sup> ). A variant of the RUE approach is the residual trend (RESTREND) of a NDVI time series, defined as the fraction of the difference between the observed NDVI and the NDVI predicted from climate data (Yengoh et al. 2015 <sup>[[#fn:r376|376]]</sup> ; John et al. 2016 <sup>[[#fn:r377|377]]</sup> ). These two metrics aim to estimate the NPP, rainfall and the time dimensions. They are simple transformations of the same three variables: RUE shows the NPP relationship with rainfall for individual years, while RESTREND is the interannual change of RUE; also, both consider that rainfall is the only variable that affects biomass production. They are legitimate metrics when used appropriately, but in many cases they involve oversimplifications and yield misleading results (Fensholt et al. 2015 <sup>[[#fn:r378|378]]</sup> ; Prince et al. 1998 <sup>[[#fn:r379|379]]</sup> ).

Furthermore, increases in NPP do not always indicate improvement in land condition/reversal of land degradation, since this does not account for changes in vegetation composition. It could, for example, result from conversion of native forest to plantation, or due to bush encroachment, which many consider to be a form of land degradation (Ward 2005 <sup>[[#fn:r380|380]]</sup> ). Also, NPP may be increased by irrigation, which can enhance productivity in the short to medium term while increasing risk of soil salinisation in the long term (Niedertscheider et al. 2016 <sup>[[#fn:r381|381]]</sup> ).

Recent progress and expanding time series of canopy characterisations based on passive microwave satellite sensors have offered rapid progress in regional and global descriptions of forest degradation and recovery trends (Tian et al. 2017 <sup>[[#fn:r382|382]]</sup> ). The most common proxy is vertical optical depth (VOD) and has already been used to describe global forest/savannah carbon stock shifts over two decades, highlighting strong continental contrasts (Liu et al. 2015a <sup>[[#fn:r383|383]]</sup> ) and demonstrating the value of this approach to monitor forest degradation at large scales. Contrasting with NDVI, which is only sensitive to vegetation ‘greenness’, from which primary production can be modelled, VOD is also sensitive to water in woody parts of the vegetation and hence provides a view of vegetation dynamics that can be complementary to NDVI. As well as the NDVI, VOD also needs to be corrected to take into account the rainfall variation (Andela et al. 2013 <sup>[[#fn:r384|384]]</sup> ).

Even though remote sensing offers much potential, its application to land degradation and recovery remains challenging as structural changes often occur at scales below the detection capabilities of most remote-sensing technologies. Additionally, if the remote sensing is based on vegetation index data, other forms of land degradation, such as nutrient depletion, changes of soil physical or biological properties, loss of values for humans, among others, cannot be inferred directly by remote sensing. The combination of remotely sensed images and field-based approach can give improved estimates of carbon stocks and tree biodiversity (Imai et al. 2012 <sup>[[#fn:r385|385]]</sup> ; Fujiki et al. 2016 <sup>[[#fn:r386|386]]</sup> ).

Additionally, the majority of trend techniques employed would be capable of detecting only the most severe of degradation processes, and would therefore not be useful as a degradation early-warning system (Higginbottom et al. 2014 <sup>[[#fn:r387|387]]</sup> ; Wessels et al. 2012 <sup>[[#fn:r388|388]]</sup> ). However, additional analyses using higher-resolution imagery, such as the Landsat and SPOT satellites, would be well suited to providing further localised information on trends observed (Higginbottom et al. 2014 <sup>[[#fn:r389|389]]</sup> ). New approaches to assess land degradation using high spatial resolution are developing, but the need for time series makes progress slow. The use of synthetic aperture radar (SAR) data has been shown to be advantageous for the estimation of soil surface characteristics, in particular, surface roughness and soil moisture (Gao et al. 2017 <sup>[[#fn:r390|390]]</sup> ; Bousbih et al. 2017 <sup>[[#fn:r391|391]]</sup> ), and detecting and quantifying selective logging (Lei et al. 2018 <sup>[[#fn:r392|392]]</sup> ). Continued research effort is required to enable full assessment of land degradation using remote sensing.

Computer simulation models can be used alone or combined with the remote sensing observations to assess land degradation. The Revised Universal Soil Loss Equation (RUSLE) can be used, to some extent, to predict the long-term average annual soil loss by water erosion. RUSLE has been constantly revisited to estimate soil loss based on the product of rainfall–runoff erosivity, soil erodibility, slope length and steepness factor, conservation factor, and support practice parameter (Nampak et al. 2018 <sup>[[#fn:r393|393]]</sup> ). Inherent limitations of RUSLE include data-sparse regions, inability to account for soil loss from gully erosion or mass wasting events, and that it does not predict sediment pathways from hillslopes to water bodies (Benavidez et al. 2018 <sup>[[#fn:r394|394]]</sup> ). Since RUSLE models only provide gross erosion, the integration of a further module in the RUSLE scheme to estimate the sediment yield from the modelled hillslopes is needed. The spatially distributed sediment delivery model, WaTEM/SEDEM, has been widely tested in Europe (Borrelli et al. 2018 <sup>[[#fn:r395|395]]</sup> ). Wind erosion is another factor that needs to be taken into account in the modelling of soil erosion (Webb et al. 2017a <sup>[[#fn:r396|396]]</sup> , 2016 <sup>[[#fn:r397|397]]</sup> ). Additional models need to be developed to include the limitations of the RUSLE models.

Regarding the field-based approach to assess land degradation, there are multiple indicators that reflect functional ecosystem processes linked to ecosystem services and thus to the value for humans. These indicators are a composite set of measurable attributes from different factors, such as climate, soil, vegetation, biomass, management, among others, that can be used together or separately to develop indexes to better assess land degradation (Allen et al. 2011 <sup>[[#fn:r398|398]]</sup> ; Kosmas et al. 2014 <sup>[[#fn:r399|399]]</sup> ).

Declines in vegetation cover, changes in vegetation structure, decline in mean species abundances, decline in habitat diversity, changes in abundance of specific indicator species, reduced vegetation health and productivity, and vegetation management intensity and use, are the most common indicators in the vegetation condition of forest and woodlands (Stocking et al. 2001 <sup>[[#fn:r400|400]]</sup> ; Wiesmair et al. 2017 <sup>[[#fn:r401|401]]</sup> ; Ghazoul and Chazdon 2017 <sup>[[#fn:r402|402]]</sup> ; Alkemade et al. 2009 <sup>[[#fn:r403|403]]</sup> ).

Several indicators of the soil quality (SOM, depth, structure, compaction, texture, pH, C:N ratio, aggregate size distribution and stability, microbial respiration, soil organic carbon, salinisation, among others) have been proposed (Schoenholtz et al. 2000 <sup>[[#fn:r404|404]]</sup> ) (Section 2.2). Among these, SOM directly and indirectly drives the majority of soil functions. Decreases in SOM can lead to a decrease in fertility and biodiversity, as well as a loss of soil structure, causing reductions in water-holding capacity, increased risk of erosion (both wind and water) and increased bulk density and hence soil compaction (Allen et al. 2011 <sup>[[#fn:r405|405]]</sup> ; Certini 2005 <sup>[[#fn:r406|406]]</sup> ; Conant et al. 2011a <sup>[[#fn:r407|407]]</sup> ). Thus, indicators related with the quantity and quality of the SOM are necessary to identify land degradation (Pulido et al. 2017 <sup>[[#fn:r408|408]]</sup> ; Dumanski and Pieri 2000 <sup>[[#fn:r409|409]]</sup> ). The composition of the microbial community is very likely to be positive impacted by both climate change and land degradation processes (Evans and Wallenstein 2014 <sup>[[#fn:r410|410]]</sup> ; Wu et al. 2015 <sup>[[#fn:r411|411]]</sup> ; Classen et al. 2015 <sup>[[#fn:r412|412]]</sup> ), thus changes in microbial community composition can be very useful to rapidly reflect land degradation (e.g., forest degradation increased the bacterial alpha-diversity indexes) (Flores-Rentería et al. 2016 <sup>[[#fn:r413|413]]</sup> ; Zhou et al. 2018 <sup>[[#fn:r414|414]]</sup> ). These indicators might be used as a set of site-dependent indicators, and in a plant-soil system (Ehrenfeld et al. 2005 <sup>[[#fn:r415|415]]</sup> ).

Useful indicators of degradation and improvement include changes in ecological processes and disturbance regimes that regulate the flow of energy and materials and that control ecosystem dynamics under a climate change scenario. Proxies of dynamics include spatial and temporal turnover of species and habitats within ecosystems (Ghazoul et al. 2015 <sup>[[#fn:r416|416]]</sup> ; Bahamondez and Thompson 2016 <sup>[[#fn:r417|417]]</sup> ). Indicators in agricultural lands include crop yield decreases and difficulty in maintaining yields (Stocking et al. 2001 <sup>[[#fn:r418|418]]</sup> ). Indicators of landscape degradation/improvement in fragmented forest landscapes include the extent, size and distribution of remaining forest fragments, an increase in edge habitat, and loss of connectivity and ecological memory (Zahawi et al. 2015 <sup>[[#fn:r419|419]]</sup> ; Pardini et al. 2010 <sup>[[#fn:r420|420]]</sup> ).

In summary, as land degradation is such a complex and global process, there is no single method by which land degradation can be estimated objectively and consistently over large areas ( ''very high confidence'' ). However, many approaches exist that can be used to assess different aspects of land degradation or provide proxies of land degradation. Remote sensing, complemented by other kinds of data (i.e., field observations, inventories, expert opinions), is the only method that can generate geographically explicit and globally consistent data over time scales relevant for land degradation (several decades).

<span id="status-and-current-trends-of-land-degradation"></span>