New Study Shows Saltwater Intrusion is Wreaking Havoc on Farms Throughout Delmarva

Newark, DE (July 31, 2023) — With continued sea-level rise, coastal waters are reaching farther inland causing changes in soil salinity and water quality, leading to permanent land loss and ecosystem alterations1,2,3,4,5,6. In coastal counties of the United States, which hosts about 9% of all US farmlands7, saltwater intrusion (SWI) into coastal ground- and surface-water results from a combination of natural sea-level variability and sea-level rise, land subsidence, drought and storm surges, the connectivity of the landscape to tidal channels and groundwater extraction4,8,9,10. Furthermore, SWI increasingly results from frequent far-reaching seasonal high tides, as opposed to solely from infrequent powerful storms, as was the more dominant driver a few decades ago11. With reported rates of sea-level rise twice the global average9, the Mid-Atlantic region of the United States deserves special attention.

SWI is leading to a suite of ecological changes including increased soil salinity, visible salt patches on the soil, the formation of ghost forests and expansion of salt-tolerant invasive species2,12. Moreover, SWI can directly reduce crop yield, as most crops are highly sensitive to saline soil13,14,15,16. Yet, it is challenging to quantify agricultural losses due to SWI, since marginal yield losses are difficult to detect or quantify. Visible salt patches on farm fields, that typically occur near field edges, close to agricultural ditches and tidal creeks at the lowest lying points on the field (Fig. 1), can be used as a proxy for SWI mapping. However, documenting these salt signatures is challenging due to their fine spatial scale and patchiness, ranging from a few to hundreds of metres. Measuring the extent and severity of the impacts of SWI through field-based methods is labour-intensive, time-consuming and expensive. A more direct and cost-effective approach would be to use remotely sensed images (aerial and satellite data) and machine-learning approaches to identify white, reflective patches on the soil as salt signatures (Fig. 1). Combining field-based knowledge of salt patches in the study area and remote sensing techniques, we have developed a method that is efficient in identifying fine-scale salt patch features over a large geographic region.

We provide high-resolution mapping of visual evidence of salt patches on farmlands in the Delmarva Peninsula covering 1.54 million ha over 14 coastal counties in Delaware, Maryland and Virginia; an area that hosts 28.4% of the total harvested farmlands in those three states. At least 35% of the land on the Delmarva Peninsula is within 5 m of the high tideline17. Using a Random Forest (RF) algorithm trained and tested with 94,240 reference points, we mapped and quantified farmland in 14 counties that displayed transient or persistent salt patches between 2011 and 2017 (Fig. 2) and estimated the loss in profit from these salt-impacted farmlands. The ‘salt patch’ class in this study includes bright white salt patches along the farm fringes (Fig. 1), mostly found in elevation <1 m, as well as bright white patches scattered anywhere in a field. We further quantified the area that converted from salt patch, farmlands and bare soil into marsh (Fig. 3) that often represents the permanent loss in productive farmlands due to increasing soil salinity.

Our results show that the effect of salt patches on agricultural productivity extends far beyond what is currently mappable. For example, while the acreage of land with visible salt patches may be small, its presence denotes that the entire field is at risk of conversion to saline soil unsuitable for traditional farming. Further, in some instances, soil salinity in nearby areas of the field may be high enough to reduce crop yield but not enough to leave bare patches. To identify farmlands at the greatest risk of SWI, we calculated the acreage of corn and soybeans within 50, 100 and 200 m buffers around the existing salt patches using the United States Department of Agriculture (USDA) Cropland Data Layer18. We selected these distances on the basis of measurements of soil electrical conductivity (EC) collected from 36 farm sites (Supplementary Fig. 1; Methods). Our choice of corn and soybeans as preferred crops is because the Delmarva economy is dominated by corn–soybean farming.

Most farmlands on the Delmarva Peninsula are planted in grain crop rotations during summer (for example, corn–soybean). While soybean is more tolerant to salt13, corn is a more profitable crop. Due to the ongoing and increasing SWI effects, crop yields in affected sites are expected to decline over the coming years15,19,20. In an attempt to quantify the range in potential losses in profit from increased soil salinization, we used an ‘enterprise budget’ (Methods) that considers yields, land rental, fertilizer prices and harvested grain prices for three scenarios: (1) ‘business-as-usual’ where the at-risk farmlands have the current corn and soybean acreage, (2) ‘corn counterfactual’ (assuming at-risk farmlands with corn and soybean combined have 100% corn, thus 100% annual profit from corn) and (3) ‘soybean counterfactual’ (assuming at-risk farmlands with corn and soybean combined have 100% soybean, thus 100% annual profit from soybean). We estimated potential losses assuming zero profits (not zero yield) on salt patches and surrounding farmland within 50, 100 and 200 m buffers, relative to a counterfactual of average profits under high yields. In addition, we estimated potential losses assuming a reduced 80% yield for the two counterfactual scenarios. Counterfactual profits (potential losses) were calculated on the basis of average 10 year input, crop prices and the highest reported annual county-based crop yields21.

Results

Visible salt patches expanding at an alarmingly high rate

Salt patches, associated with very little to no plant growth, represent a complete loss of productive land. About 472 ha of land across the Delmarva Peninsula, mostly near field edges, had visible salt patches during 2011–2013 (Fig. 2). This area nearly doubled to 905 ha during 2016–2017 and varied greatly by county. The nine coastal Maryland counties experienced a 79% increase in salt patch area. In Delaware and the Eastern Shore of Virginia, the area of salt patches increased 81% and 243%, respectively, between 2011 and 2017. While the expansion rate is alarming, the absolute area of these identified salt patches remained small in 2017: about 445 ha in Maryland; 339 ha in Delaware; and 122 ha in Virginia. The rate of change between time-steps varied across the counties with numbers ranging from a 7.6% increase in Kent County to a 450.5% increase in Caroline County, both in Maryland (Fig. 2). Salt patches remained a small fraction of total land cover across Delmarva, ranging between 0.01% and 0.18% of total farmlands in a given county in 2011–2013 and between 0.01% and 0.39% in 2016–2017 (Fig. 4a). Moreover, appearance and disappearance of salt patches varied over time, with only 24 ha of visible salt patches identified in the period 2011–2013 remaining visible in 2016–2017 (8.7 ha in Maryland, 15 ha in Delaware and 0.6 ha in Virginia). The overall expansion of salt patches was largely due to 436, 323 and 121 ha of new salt patches that appeared in 2016–2017 in Maryland, Delaware and Virginia, respectively. About 36% and 32% of the salt patch area are located on sites with elevation <2 m during 2011–2013 and 2016–2017, respectively.

Farmlands at risk from further saltwater intrusion

Increasing soil salinity might result in gradual conversion of farmlands to marsh22. We estimated that about 36.5 ha of land was converted from salt patch to marsh and about 1,007 ha of land was converted from bare soil to marsh between 2011 and 2017 across our study area. Over 188 ha and 275 ha of bare soil converted to marsh between 2011 and 2017 within the 100 and 200 m buffers, respectively. In addition, over 8,096 ha of farmland was converted to marsh across the 14 coastal Delmarva counties between 2011 and 2017 (Fig. 3). The three Delaware counties have the largest share of such conversions at 3,824 ha, followed by the nine counties in Maryland (3,488 ha) and two counties in Virginia (784 ha). These converted lands are more suitable for salt-tolerant species14,16,23, including both native marsh species and salt-tolerant invasive species such as the common reed (the Eurasian lineage of Phragmites australis).

We estimated about 13,732 ha of at-risk farmland across the Delmarva Peninsula during 2011–2013 that are located within 50 m of a visible salt patch. By 2016–2017, that number grew to about 28,022 ha or about three-quarters the size of Philadelphia. The increase in at-risk farmlands varied by state—from 4,726 to 9,150 ha in the three Delaware counties (94% increase), from 1,321 to 2,636 ha in the two Virginia counties (99% increase) and from 7,684 to 16,236 ha in the nine Maryland counties (111% increase). Between the two time-steps, this represents a change from 2.5% to 5.2%, 2.3% to 4.3% and 1.9% to 4.1% of total farmlands in the study counties in Virginia, Delaware and Maryland, respectively. In 2011–2013, these at-risk farmlands represented between as little as 0.4% of all farmlands in Cecil County, Maryland, to up to 4.7% of all farmlands in Somerset County, Maryland (Fig. 4b). In 2016–2017, the range of at-risk farmlands increased to a minimum of 1.6% in Kent, Maryland, and a maximum of 8% in Somerset, Maryland (Fig. 4b).

We found that ~35,032 ha of farmland were within 100 m of a salt patch during 2011–2013, which increased to 68,475 ha, or about twice the size of Philadelphia, during 2016–2017. Delaware counties experienced a rise in at-risk farmlands from 12,368 to 22,416 ha (81%), whereas the counties in Virginia and Maryland had an increase from 3,250 to 6,283 ha (93%) and from 19,414 to 39,775 ha (105%), respectively. This represents an increase from 6% to 12.2%, 6% to 10.5% and 4.7% to 10.1% of the total farmlands in the study counties in Virginia, Delaware and Maryland, respectively. The distribution pattern remains the same as was seen for at-risk farmlands within a 50 m buffer. Somerset County, Maryland, had the largest share for both time-steps (11.7% and 16.9%; Fig. 4c).

Using a more liberal estimate of the area around visible salt patches in which crop yields may be affected, we found that about 91,073 and 166,930 ha of at-risk farmland was within 200 m of a visible salt patch during 2011–2013 and 2016–2017, respectively. The study counties in Delaware, Virginia and Maryland witnessed a rise in at-risk farmlands from 33,064 to 55,511 ha (68%), 8,095 to 14,898 ha (84%) and 49,914 to 96,521 ha (93%), respectively (Fig. 4d). This represents an increase from 16.1% to 26.1%, 15% to 29.4% and 12.2% to 24.5% in the study counties in Delaware, Virginia and Maryland, respectively. It is noteworthy that crop stress due to soil salinization does not decline linearly with distance from a visible salt patch. In other words, the deleterious effects experienced by crops do not depend on their exact location within these buffers per se, as we recorded similar soil EC values in all these buffers (Supplementary Fig. 1).

Substantial profit loss from saltwater intrusion on Delmarva

Within the buffers around the salt patches, the sources of financial losses stem from both observed salt patches and potential salinization of the adjacent farmlands, thereby reducing yield and profit. It is not possible to estimate the exact percentage of loss in profit due to the varied levels of salinity in farmlands where salt patches are not visible yet. Hence, we first estimated an upper bound in these losses with the assumption of zero profit on the salt-impacted farmlands, both salt patches and farmlands within these buffers, planted in corn or soybean. Loss estimates are based on the assumption that yields on salt-affected lands generate revenue that only just covers input costs, as opposed to assuming zero yields, which would induce financial losses if crops were planted. We calculated foregone profits (potential losses) on the basis of high yield and average per bushel profits over 10 years (2011–2020) in each county (Fig. 5). Then we estimated subscenarios within our two counterfactuals (100% profit coming from corn and soybeans, respectively) assuming a 20% yield loss, that is 80% crop yield potential. It should be noted that the same level of salinity would result in different yield decline for corn and soybeans.

Discussion

The mid-Atlantic region of the United States has been witnessing rapid landscape-level changes over the last few decades6,24,25. This study documents visible salt patches and their spatiotemporal evolution across the Delmarva Peninsula. Our results show that the rapid growth of salt patch area across the Delmarva during the last decade is notable. Bare areas in farm fields displaying the distinct signature of SWI nearly doubled (92% increase) during the 6 years of the study period (2011–2017), as did the potential losses in profit in at-risk farmlands.

While visible salt patches are a good indicator of the spatial distribution of salt-impacted farmlands, their absence may not necessarily indicate a productive farmland or absence of SWI; high salinity areas may not be equally visible at all times. Various factors ranging from farming practices to regional climate and weather events may modify identifiable salt patches on the ground. For example, following a large rain event, farm abandonment or a fallow period26,27, salt patches may not be visible in aerial images. Compared to drier climates, the Delmarva Peninsula receives an average rainfall of 1,140 mm rainfall annually28. This is usually enough to dilute and remove salts from the soil surface and to allow plants to germinate16; yet, the water table in our study region is often close enough to the surface that the saline water has nowhere else to go. Moreover, soil salinity may increase incidentally following nuisance flooding29 or weather events such as northeasters or hurricanes, which push salts inland and may increase the visibility of salt signatures in remotely sensed data. The addition of salts at later crop growth stages may reduce yields30, without causing bare patches. As such, visible salt patches and the approach of estimating at-risk farmlands using buffers around them should be considered a highly conservative estimate of farmlands affected by SWI.

On the basis of our findings, it is a reasonable assumption that all visible salt patches indicate land presently affected by SWI or the very frontlines of coastal changes due to sea-level rise. Visible salt patches are ephemeral, often occurring before farm abandonment and land use change. However, widespread marsh conversion in the study region denotes a strong directional change in land covers as a SWI consequence. About 36.5 ha of salt patch area converted to marsh, while 8,096 ha of farmland in the Delmarva Peninsula converted to marsh during the 6 year study period, which exceeds the area of fields exhibiting salt patches and suggests that sea-level rise is a substantial source of land cover change in this region. Moreover, the extent of sea-level rise impacts appears to be growing; we estimated that in 2016–2017 between 28,022 (using a 50 m buffer) and 166,930 (using a 200 m buffer) additional hectares of farmland were at-risk on the Peninsula due to their proximity to the visible salt patches. Evidently, the effects of sea-level rise and SWI are far more extensive than what is visible at the surface.

These changes are of great economic concern and having a visible signal of contemporary sea-level rise presents an opportunity to detect the geographic distribution of sea-level impacts in near real-time, which holds promise for a number of future applications. From a basic science perspective, real-time tracking will enable greater mechanistic understanding and improved ability to model SWI. For example, empirical models based on spatial correlations could help us to better understand the role of ditch and canal networks31, soil characteristics and geologic features32, legacies of land reclamation33,34 or water table levels and recharge35,36,37. Greater understanding of how human activities exacerbate or reduce salt patches and marsh, such as the construction or removal of tide gates and berms, should be of particular interest to policy-makers. Real-time tracking could also be used to better understand the distribution and inequity of economic impacts in the coastal zone, knowledge of which could be used to design and target new incentive programmes to the landowners who most need them in the changing coastal landscape. Our high-resolution geospatial datasets provide a finer spatial resolution compared to global datasets, such as the Global Map of Salt-affected Soils or GSASmap38. This level of detail is critical for the farm-level decision-making that is often required to design and implement state-level policies.

Our work provides evidence for an immediate policy attention required to protect the coastal lands against increasing soil salinization. Due to its sensitivity to salinity, the corn-focused agricultural economy is not suitable for many SWI-affected coastal fields across Delmarva. Increasing the share of farmlands under more salt-tolerant crops (for example, soybeans, sorghum or barley), reducing inputs, adding gypsum to the salt-affected lands or using crop insurance as a strategy to limit losses and delay transitions might result in lower economic losses in the immediate future. However, landowners might be forced to abandon these farmlands once the soil becomes salinized beyond the tolerance of any traditionally farmed food crops. Recent studies have examined alternative crops, such as barley, quinoa and sorghum that might be more suitable for these landscapes16. Other adaptation strategies might include a controlled conversion of these landscapes into marsh that can support wildlife or act as a barrier to encroaching seawater39. While such transitions are vital to sustainable solutions, the fate of such coastal frontier zones will be shaped by the salinity gradient across these evolving landscapes. In highly salinized regions, halophytes might contribute to further soil salinization through continued and efficient water uptake in brackish soils—an example of a positive feedback loop40,41. Conversely, marsh vegetation might protect comparatively less salinized regions from further salt accumulation42.

SWI is rampant across the North American Coastal Plain, from Massachusetts, United States, in the north to Northern Mexico in the south, with documented coastal forest loss6. This study highlights another SWI consequence that has far-reaching implications for the US economy as well as coastal ecosystems, by drawing attention to the gradual loss of productive Delmarva farmlands from SWI. Due to the reliance on freely available aerial and satellite images and well-established machine-learning methods, our geospatial method is transferable to other coastal regions across and beyond the mid-Atlantic with known SWI issues1. While other long-term SWI consequences, such as the expansion of ghost forests along the US coasts, have been documented in recent studies1,6, the elusive nature of salt patches posed a challenge in estimating agricultural losses. Our high-resolution datasets not only address that challenge but also provide a baseline and a reproducible approach that can be used to track the spread and cost of SWI in the Delmarva Peninsula and beyond.